research papers of animal sciences

• Search by keyword
• Search by citation

Page 1 of 25

Altered microRNA composition in the uterine lumen fluid in cattle ( Bos taurus ) pregnancies initiated by artificial insemination or transfer of an in vitro produced embryo

MicroRNAs (miRNAs) are presented in the uterine lumen of many mammals, and in vitro experiments have determined that several miRNAs are important for the regulation of endometrial and trophoblast functions. Ou...

• View Full Text

Effects of replacing soybean meal with enzymolysis-fermentation compound protein feed on growth performance, apparent digestibility of nutrients, carcass traits, and meat quality in growing-finishing pigs

Addressing the shortage of high-quality protein resources, this study was conducted to investigate the effects of replacing soybean meal (SBM) with different levels of enzymolysis-fermentation compound protein...

Higher abundance of DLD protein in buffalo bull spermatozoa causes elevated ROS production leading to early sperm capacitation and reduction in fertilizing ability

Before fertilization, spermatozoa undergo a crucial maturation step called capacitation, which is a unique event regulates the sperm’s ability for successful fertilization. The capacitation process takes place...

Dietary supplementation of valine, isoleucine, and tryptophan may overcome the negative effects of excess leucine in diets for weanling pigs containing corn fermented protein

Diets with high inclusion of corn co-products such as corn fermented protein (CFP) may contain excess Leu, which has a negative impact on feed intake and growth performance of pigs due to increased catabolism ...

Host genetics and gut microbiota synergistically regulate feed utilization in egg-type chickens

Feed efficiency is a crucial economic trait in poultry industry. Both host genetics and gut microbiota influence feed efficiency. However, the associations between gut microbiota and host genetics, as well as ...

Dietary processed former foodstuffs for broilers: impacts on growth performance, digestibility, hematobiochemical profiles and liver gene abundance

The present experiment aimed to evaluate the effects of commercially processed former foodstuffs (cFF) as dietary substitutes of corn, soybean meal and soybean oil on the growth performance, apparent total tra...

Hepatoprotective effects of magnolol in fatty liver hemorrhagic syndrome hens through shaping gut microbiota and tryptophan metabolic profile

Magnolol (MAG) exhibits hepatoprotective activity, however, whether and how MAG regulates the gut microbiota to alleviate fatty liver hemorrhagic syndrome (FLHS) remains unclear. Therefore, we investigated the...

Quercetin ameliorates oxidative stress-induced apoptosis of granulosa cells in dairy cow follicular cysts by activating autophagy via the SIRT1/ROS/AMPK signaling pathway

Follicular cysts contribute significantly to reproductive loss in high-yield dairy cows. This results from the death of follicular granulosa cells (GCs) caused by oxidative stress. Quercetin is known to have s...

Early-life milk replacer feeding mediates lipid metabolism disorders induced by colonic microbiota and bile acid profiles to reduce body weight in goat model

Dysregulation of lipid metabolism and its consequences on growth performance in young ruminants have attracted attention, especially in the context of alternative feeding strategies. This study aims to elucida...

Polystyrene nanoplastic exposure actives ferroptosis by oxidative stress-induced lipid peroxidation in porcine oocytes during maturation

Polystyrene nanoplastics (PS-NPs) are becoming increasingly prevalent in the environment with great advancements in plastic products, and their potential health hazard to animals has received much attention. S...

Methionine deficiency inhibited pyroptosis in primary hepatocytes of grass carp ( Ctenopharyngodon idella ): possibly via activating the ROS-AMPK-autophagy axis

Methionine (Met) is the only sulfur-containing amino acid among animal essential amino acids, and methionine deficiency (MD) causes tissue damage and cell death in animals. The common modes of cell death inclu...

Dietary Zn proteinate with moderate chelation strength alleviates heat stress-induced intestinal barrier function damage by promoting expression of tight junction proteins via the A20/NF-κB p65/MMP-2 pathway in the jejunum of broilers

The aim of this study was to determine whether and how Zn proteinate with moderate chelation strength (Zn-Prot M) can alleviate heat stress (HS)-induced intestinal barrier function damage of broilers. A comple...

Dietary bile acids supplementation decreases hepatic fat deposition with the involvement of altered gut microbiota and liver bile acids profile in broiler chickens

High-fat diets (HFD) are known to enhance feed conversion ratio in broiler chickens, yet they can also result in hepatic fat accumulation. Bile acids (BAs) and gut microbiota also play key roles in the formati...

Role of immunomodulatory probiotics in alleviating bacterial diarrhea in piglets: a systematic review

Diarrhea is a common enteric disease in piglets that leads to high mortality and economic losses in swine production worldwide. Antibiotics are commonly used to prevent or treat diarrhea in piglets. However, i...

The influence of iron nutrition on the development of intestine and immune cell divergency in neonatal pigs

Appropriate iron supplementation is essential for neonatal growth and development. However, there are few reports on the effects of iron overload on neonatal growth and immune homeostasis. Thus, the aim of thi...

Recent advances in essential oils and their nanoformulations for poultry feed

Antibiotics in poultry feed to boost growth performance are becoming increasingly contentious due to concerns over antimicrobial resistance development. Essential oils (EOs), as natural, plant-derived compound...

Epigallocatechin-3-gallate protects bovine ruminal epithelial cells against lipopolysaccharide-induced inflammatory damage by activating autophagy

Subacute ruminal acidosis (SARA) causes an increase in endotoxin, which can induce immune and inflammatory responses in the ruminal epithelium of dairy cows. In non-ruminants, epigallocatechin-3-gallate (EGCG)...

Insights into the interplay between gut microbiota and lipid metabolism in the obesity management of canines and felines

Obesity is a prevalent chronic disease that has significant negative impacts on humans and our companion animals, including dogs and cats. Obesity occurs with multiple comorbidities, such as diabetes, hyperten...

Melatonin alleviates palmitic acid-induced mitochondrial dysfunction by reducing oxidative stress and enhancing autophagy in bovine endometrial epithelial cells

Negative energy balance (NEB) typically occurs in dairy cows after delivery. Cows with a high yield are more likely to experience significant NEB. This type of metabolic imbalance could cause ketosis, which is...

Effects of bacteriocin-producing Lactiplantibacillus plantarum on bacterial community and fermentation profile of whole-plant corn silage and its in vitro ruminal fermentation, microbiota, and CH 4 emissions

Silage is widely used to formulate dairy cattle rations, and the utilization of antibiotics and methane emissions are 2 major problems for a sustainable and environmentally beneficial ruminant production syste...

Astragalus polysaccharides-induced gut microbiota play a predominant role in enhancing of intestinal barrier function of broiler chickens

The intestinal barrier is the first line of defense against intestinal invasion by pathogens and foreign antigens and is closely associated with the gut microbiota. Astragalus polysaccharides (APS) have a long hi...

CAMKK2-AMPK axis endows dietary calcium and phosphorus levels with regulatory effects on lipid metabolism in weaned piglets

In the realm of swine production, optimizing body composition and reducing excessive fat accumulation is critical for enhancing both economic efficiency and meat quality. Despite the acknowledged impact of die...

From follicle to blastocyst: microRNA-34c from follicular fluid-derived extracellular vesicles modulates blastocyst quality

Within the follicular fluid, extracellular vesicles (EVs) guide oocyte growth through their cargo microRNAs (miRNAs). Here, we investigated the role of EVs and their cargo miRNAs by linking the miRNAs found in...

Lipolysis pathways modulate lipid mediator release and endocannabinoid system signaling in dairy cows’ adipocytes

As cows transition from pregnancy to lactation, free fatty acids (FFA) are mobilized from adipose tissues (AT) through lipolysis to counter energy deficits. In clinically healthy cows, lipolysis intensity is r...

Advances in single-cell transcriptomics in animal research

Understanding biological mechanisms is fundamental for improving animal production and health to meet the growing demand for high-quality protein. As an emerging biotechnology, single-cell transcriptomics has ...

Postbiotics from Saccharomyces cerevisiae fermentation stabilize microbiota in rumen liquid digesta during grain-based subacute ruminal acidosis (SARA) in lactating dairy cows

Subacute ruminal acidosis (SARA) is a common metabolic disorder of high yielding dairy cows, and it is associated with dysbiosis of the rumen and gut microbiome and host inflammation. This study evaluated the ...

Dietary silymarin improves performance by altering hepatic lipid metabolism and cecal microbiota function and its metabolites in late laying hens

Liver lipid dysregulation is one of the major factors in the decline of production performance in late-stage laying hens. Silymarin (SIL), a natural flavonolignan extracted from milk thistle, is known for its ...

Lysine 2-hydroxyisobutyrylation levels determined adipogenesis and fat accumulation in adipose tissue in pigs

Excessive backfat deposition lowering carcass grade is a major concern in the pig industry, especially in most breeds of obese type pigs. The mechanisms involved in adipogenesis and fat accumulation in pigs re...

Effects of different energy levels in low-protein diet on liver lipid metabolism in the late-phase laying hens through the gut-liver axis

The energy/protein imbalance in a low-protein diet induces lipid metabolism disorders in late-phase laying hens. Reducing energy levels in the low-protein diet to adjust the energy-to-protein ratio may improve...

Whole-genome analysis reveals distinct adaptation signatures to diverse environments in Chinese domestic pigs

Long-term natural and artificial selection has resulted in many genetic footprints within the genomes of pig breeds across distinct agroecological zones. Nevertheless, the mechanisms by which these signatures ...

Visfatin (NAMPT) affects global gene expression in porcine anterior pituitary cells during the mid-luteal phase of the oestrous cycle

The pituitary belongs to the most important endocrine glands involved in regulating reproductive functions. The proper functioning of this gland ensures the undisturbed course of the oestrous cycle and affects...

Nano-Se exhibits limited protective effect against heat stress induced poor breast muscle meat quality of broilers compared with other selenium sources

At present, heat stress (HS) has become a key factor that impairs broiler breeding industry, which causes growth restriction and poor meat quality of broilers. Selenium (Se) is an excellent antioxidant and pla...

Effects of altering the ratio of C16:0 and cis -9 C18:1 in rumen bypass fat on growth performance, lipid metabolism, intestinal barrier, cecal microbiota, and inflammation in fattening bulls

C16:0 and cis -9 C18:1 may have different effects on animal growth and health due to unique metabolism in vivo. This study was investigated to explore the different effects of altering the ratio of C16:0 and cis -9...

Source and level of dietary iron influence semen quality by affecting inflammation, oxidative stress and iron utilization levels in boars

Boars fed a mixed form of inorganic and organic iron in excess of the NRC recommended levels still develop anemia, which suggested that the current level and form of iron supplementation in boar diets may be i...

Heat stress affects mammary metabolism by influencing the plasma flow to the glands

Environmental heat stress (HS) can have detrimental effects on milk production by compromising the mammary function. Mammary plasma flow (MPF) plays a crucial role in nutrient supply and uptake in the mammary ...

Transcriptomic and epigenomic landscapes of muscle growth during the postnatal period of broilers

Broilers stand out as one of the fastest-growing livestock globally, making a substantial contribution to animal meat production. However, the molecular and epigenetic mechanisms underlying the rapid growth an...

Progesterone regulates tissue non-specific alkaline phosphatase (TNSALP) expression and activity in ovine utero-placental tissues

Tissue non-specific alkaline phosphatase (TNSALP; encoded by the ALPL gene) has a critical role in the postnatal regulation of phosphate homeostasis, yet how TNSALP activity and expression are regulated during pr...

Probiotic Lactobacillus rhamnosus GG improves insulin sensitivity and offspring survival via modulation of gut microbiota and serum metabolite in a sow model

Sows commonly experience insulin resistance in late gestation and lactation, causing lower feed intake and milk production, which can lead to higher mortality rates in newborn piglets. The probiotic Lactobacillus...

Improving the accuracy of genomic prediction in dairy cattle using the biologically annotated neural networks framework

Biologically annotated neural networks (BANNs) are feedforward Bayesian neural network models that utilize partially connected architectures based on SNP-set annotations. As an interpretable neural network, BA...

Prebiotic galactooligosaccharide improves piglet growth performance and intestinal health associated with alterations of the hindgut microbiota during the peri-weaning period

Weaning stress reduces growth performance and health of young pigs due in part to an abrupt change in diets from highly digestible milk to fibrous plant-based feedstuffs. This study investigated whether dietar...

The walnut-derived peptide TW-7 improves mouse parthenogenetic embryo development of vitrified MII oocytes potentially by promoting histone lactylation

Previous studies have shown that the vitrification of metaphase II (MII) oocytes significantly represses their developmental potential. Abnormally increased oxidative stress is the probable factor; however, th...

UCHL1 promotes the proliferation of porcine granulosa cells by stabilizing CCNB1

The proliferation of porcine ovarian granulosa cells (GCs) is essential to follicular development and the ubiquitin–proteasome system is necessary for maintaining cell cycle homeostasis. Previous studies found...

Effects of methionine supplementation in a reduced protein diet on growth performance, oxidative status, intestinal health, oocyst shedding, and methionine and folate metabolism in broilers under Eimeria challenge

This study investigated effects of different methionine (Met) supplementation levels in a reduced protein diet on growth performance, intestinal health, and different physiological parameters in broilers under Ei...

Combining genetic markers, on-farm information and infrared data for the in-line prediction of blood biomarkers of metabolic disorders in Holstein cattle

Various blood metabolites are known to be useful indicators of health status in dairy cattle, but their routine assessment is time-consuming, expensive, and stressful for the cows at the herd level. Thus, we e...

Current status and challenges for cell-cultured milk technology: a systematic review

Cellular agriculture is an innovative technology for manufacturing sustainable agricultural products as an alternative to traditional agriculture. While most cellular agriculture is predominantly centered on t...

Effect of atractylenolide III on zearalenone-induced Snail1-mediated epithelial–mesenchymal transition in porcine intestinal epithelium

The intestinal epithelium performs essential physiological functions, such as nutrient absorption, and acts as a barrier to prevent the entry of harmful substances. Mycotoxins are prevalent contaminants found ...

Hypoxia-mediated programmed cell death is involved in the formation of wooden breast in broilers

Wooden breast (WB) myopathy is a common myopathy found in commercial broiler chickens worldwide. Histological examination has revealed that WB myopathy is accompanied by damage to the pectoralis major (PM) mus...

Bacteria colonization and gene expression related to immune function in colon mucosa is associated with growth in neonatal calves regardless of live yeast supplementation

As Holstein calves are susceptible to gastrointestinal disorders during the first week of life, understanding how intestinal immune function develops in neonatal calves is important to promote better intestina...

Embryonic thermal manipulation: a potential strategy to mitigate heat stress in broiler chickens for sustainable poultry production

Due to high environmental temperatures and climate change, heat stress is a severe concern for poultry health and production, increasing the propensity for food insecurity. With climate change causing higher t...

Myostatin gene role in regulating traits of poultry species for potential industrial applications

The myostatin ( MSTN ) gene is considered a potential genetic marker to improve economically important traits in livestock, since the discovery of its function using the MSTN knockout mice. The anti-myogenic functi...

• Editorial Board
• Manuscript editing services
• Instructions for Editors
• Sign up for article alerts and news from this journal
• Follow us on Twitter

Annual Journal Metrics

Citation Impact 2023 Journal Impact Factor: 6.3 5-year Journal Impact Factor: 6.4 Source Normalized Impact per Paper (SNIP): 1.647 SCImago Journal Rank (SJR): 1.453

Speed 2023 Submission to first editorial decision (median days): 7 Submission to acceptance (median days): 119

Usage 2023 Downloads: 1,188,041 Altmetric mentions: 583

• More about our metrics

Journal of Animal Science and Biotechnology

ISSN: 2049-1891

• Submission enquiries: Access here and click Contact Us
• General enquiries: [email protected]

Information

• Author Services

Initiatives

You are accessing a machine-readable page. In order to be human-readable, please install an RSS reader.

All articles published by MDPI are made immediately available worldwide under an open access license. No special permission is required to reuse all or part of the article published by MDPI, including figures and tables. For articles published under an open access Creative Common CC BY license, any part of the article may be reused without permission provided that the original article is clearly cited. For more information, please refer to https://www.mdpi.com/openaccess .

Feature papers represent the most advanced research with significant potential for high impact in the field. A Feature Paper should be a substantial original Article that involves several techniques or approaches, provides an outlook for future research directions and describes possible research applications.

Feature papers are submitted upon individual invitation or recommendation by the scientific editors and must receive positive feedback from the reviewers.

Editor’s Choice articles are based on recommendations by the scientific editors of MDPI journals from around the world. Editors select a small number of articles recently published in the journal that they believe will be particularly interesting to readers, or important in the respective research area. The aim is to provide a snapshot of some of the most exciting work published in the various research areas of the journal.

Original Submission Date Received: .

• Active Journals
• Find a Journal
• Proceedings Series
• For Authors
• For Reviewers
• For Editors
• For Librarians
• For Publishers
• For Societies
• For Conference Organizers
• Open Access Policy
• Institutional Open Access Program
• Special Issues Guidelines
• Editorial Process
• Research and Publication Ethics
• Article Processing Charges
• Testimonials
• Preprints.org
• SciProfiles
• Encyclopedia

Journal Description

• Open Access — free for readers, with article processing charges (APC) paid by authors or their institutions.
• High Visibility:  indexed within Scopus , SCIE (Web of Science) , PubMed , PMC , Embase , PubAg , AGRIS , Animal Science Database , CAB Abstracts , and other databases .
• Journal Rank:  JCR - Q1 (Veterinary Sciences) / CiteScore - Q1 (General Veterinary )
• Rapid Publication: manuscripts are peer-reviewed and a first decision is provided to authors approximately 16.1 days after submission; acceptance to publication is undertaken in 2.6 days (median values for papers published in this journal in the first half of 2024).
• Recognition of Reviewers: reviewers who provide timely, thorough peer-review reports receive vouchers entitling them to a discount on the APC of their next publication in any MDPI journal, in appreciation of the work done.
• Companion journals for Animals include: Birds ,   Ruminants and Zoonotic Diseases .

Latest Articles

Graphical abstract

Journal Menu

• Animals Home
• Aims & Scope
• Editorial Board
• Reviewer Board
• Topical Advisory Panel
• Instructions for Authors

Special Issues

• Sections & Collections
• Article Processing Charge
• Indexing & Archiving
• Editor’s Choice Articles
• Most Cited & Viewed
• Journal Statistics
• Journal History
• Journal Awards
• Society Collaborations

Conferences

• Editorial Office

Journal Browser

• arrow_forward_ios Forthcoming issue arrow_forward_ios Current issue
• Vol. 14 (2024)
• Vol. 13 (2023)
• Vol. 12 (2022)
• Vol. 11 (2021)
• Vol. 10 (2020)
• Vol. 9 (2019)
• Vol. 8 (2018)
• Vol. 7 (2017)
• Vol. 6 (2016)
• Vol. 5 (2015)
• Vol. 4 (2014)
• Vol. 3 (2013)
• Vol. 2 (2012)
• Vol. 1 (2011)

Highly Accessed Articles

Latest books, e-mail alert.

Topical Collections

Further information, mdpi initiatives, follow mdpi.

Subscribe to receive issue release notifications and newsletters from MDPI journals

• Search Menu
• Sign in through your institution
• Advance Articles
• High-Impact Collection
• Infographics
• Author Guidelines
• Submission Site
• Open Access
• Self-Archiving Policy
• Why Publish with Us?
• About Journal of Animal Science
• About the American Society of Animal Science
• Editorial Board
• Advertising & Corporate Services
• Journals Career Network
• Journals on Oxford Academic
• Books on Oxford Academic

High-Impact Research from Journal of Animal Science

Explore a collection of the most read and most cited articles making an impact in  Journal of Animal Science  published within the past two years. This collection will be continuously updated with the journal's leading articles so be sure to revisit periodically to see what is being read and cited.

Also discover the articles being discussed the most on digital media by  exploring this Altmetric report  pulling the most discussed articles from the past year.

Recognizing and Celebrating Women in Science

Oxford University Press is proud to support diverse voices across our publishing. In this collection, we shine a spotlight on the representation of women in scientific fields, the gains that have been made in their fields, from research and major discoveries to advocacy and outreach, and amplify the voices of women who have made a career in scientific research.

Delve into the Women in Science collection

• Recommend to Your Librarian
• Advertising and Corporate Services

Affiliations

• Online ISSN 1525-3163
• Copyright © 2024 American Society of Animal Science
• About Oxford Academic
• Publish journals with us
• University press partners
• What we publish
• New features  
• Open access
• Institutional account management
• Rights and permissions
• Get help with access
• Accessibility
• Advertising
• Media enquiries
• Oxford University Press
• Oxford Languages
• University of Oxford

Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide

• Copyright © 2024 Oxford University Press
• Cookie settings
• Cookie policy
• Privacy policy
• Legal notice

This Feature Is Available To Subscribers Only

Sign In or Create an Account

This PDF is available to Subscribers Only

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

Loading metrics

Open Access

Essays articulate a specific perspective on a topic of broad interest to scientists.

See all article types »

A guide to open science practices for animal research

Contributed equally to this work with: Kai Diederich, Kathrin Schmitt

Affiliation German Federal Institute for Risk Assessment, German Centre for the Protection of Laboratory Animals (Bf3R), Berlin, Germany

* E-mail: [email protected]

• Kai Diederich, 
• Kathrin Schmitt, 
• Philipp Schwedhelm, 
• Bettina Bert, 
• Céline Heinl

Published: September 15, 2022

• https://doi.org/10.1371/journal.pbio.3001810
• Reader Comments

Translational biomedical research relies on animal experiments and provides the underlying proof of practice for clinical trials, which places an increased duty of care on translational researchers to derive the maximum possible output from every experiment performed. The implementation of open science practices has the potential to initiate a change in research culture that could improve the transparency and quality of translational research in general, as well as increasing the audience and scientific reach of published research. However, open science has become a buzzword in the scientific community that can often miss mark when it comes to practical implementation. In this Essay, we provide a guide to open science practices that can be applied throughout the research process, from study design, through data collection and analysis, to publication and dissemination, to help scientists improve the transparency and quality of their work. As open science practices continue to evolve, we also provide an online toolbox of resources that we will update continually.

Citation: Diederich K, Schmitt K, Schwedhelm P, Bert B, Heinl C (2022) A guide to open science practices for animal research. PLoS Biol 20(9): e3001810. https://doi.org/10.1371/journal.pbio.3001810

Copyright: © 2022 Diederich et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: The authors received no specific funding for this work.

Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: All authors are employed at the German Federal Institute for Risk Assessment and part of the German Centre for the Protection of Laboratory Animals (Bf3R) which developed and hosts animalstudyregistry.org , a preregistration platform for animal studies and animaltestinfo.de, a database for non-technical project summaries (NTS) of approved animal study protocols within Germany.

Abbreviations: CC, Creative Commons; CIRS-LAS, critical incident reporting system in laboratory animal science; COVID-19, Coronavirus Disease 2019; DOAJ, Directory of Open Access Journals; DOI, digital object identifier; EDA, Experimental Design Assistant; ELN, electronic laboratory notebook; EU, European Union; IMSR, International Mouse Strain Resource; JISC, Joint Information Systems Committee; LIMS, laboratory information management system; MGI, Mouse Genome Informatics; NC3Rs, National Centre for the Replacement, Refinement and Reduction of Animals in Research; NTS, non-technical summary; RRID, Research Resource Identifier

Introduction

Over the past decade, the quality of published scientific literature has been repeatedly called into question by the failure of large replication studies or meta-analyses to demonstrate sufficient translation from experimental research into clinical successes [ 1 – 5 ]. At the same time, the open science movement has gained more and more advocates across various research areas. By sharing all of the information collected during the research process with colleagues and with the public, scientists can improve collaborations within their field and increase the reproducibility and trustworthiness of their work [ 6 ]. Thus, the International Reproducibility Networks have called for more open research [ 7 ].

However, open science practices have not been adopted to the same degree in all research areas. In psychology, which was strongly affected by the so-called reproducibility crisis, the open science movement initiated real practical changes leading to a broad implementation of practices such as preregistration or sharing of data and material [ 8 – 10 ]. By contrast, biomedical research is still lagging behind. Open science might be of high value for research in general, but in translational biomedical research, it is an ethical obligation. It is the responsibility of the scientist to transparently share all data collected to ensure that clinical research can adequately evaluate the risks and benefits of a potential treatment. When Russell and Burch published “The Principles of Humane Experimental Technique” in 1959, scientists started to implement their 3Rs principle to answer the ethical dilemma of animal welfare in the face of scientific progress [ 11 ]. By replacing animal experiments wherever possible, reducing the number of animals to a strict minimum, and refining the procedures where animals have still to be used, this ethical dilemma was addressed. However, in recent years, whether the 3Rs principle is sufficient to fully address ethical concerns about animal experiments has been questioned [ 12 ].

Most people tolerate the use of animals for scientific purposes only under the basic assumption that the knowledge gained will advance research in crucial areas. This implies that performed experiments are reported in a way that enables peers to benefit from the collected data. However, recent studies suggest that a large proportion of animal experiments are never actually published. For example, scientists working within the European Union (EU) have to write an animal study protocol for approval by the competent authorities of the respective country before performing an animal experiment [ 13 ]. In these protocols, scientists have to describe the planned study and justify every animal required for the project. By searching for publications resulting from approved animal study protocols from 2 German University Medical Centers, Wieschowski and colleagues found that only 53% of approved protocols led to a publication after 6 years [ 14 ]. Using a similar approach, Van der Naald and colleagues determined a publication rate of 60% at the Utrecht Medical Center [ 15 ]. In a follow-up survey, the respective researchers named so-called “negative” or null-hypothesis results as the main cause for not publishing outcomes [ 15 ]. The current scientific system is shaped by publishers, funders, and institutions and motivates scientists to publish novel, surprising, and positive results, revealing one of the many structural problems that the numerous efforts towards open science initiatives are targeting. Non-publication not only strongly contradicts ethical values, but also it compromises the quality of published literature by leading to overestimation of effect sizes [ 16 , 17 ]. Furthermore, publications of animal studies too often show poor reporting that strongly impairs the reproducibility, validity, and usefulness of the results [ 18 ]. Unfortunately, the idea that negative or equivocal findings can also contribute to the gain of scientific knowledge is frequently neglected.

So far, the scientific community using animals has shown limited resonance to the open science movement. Due to the strong controversy surrounding animal experiments, scientists have been reluctant to share information on the topic. Additionally, translational research is highly competitive and researchers tend to be secretive about their ideas until they are ready for publication or patent [ 19 , 20 ]. However, this missing openness could also point to a lack of knowledge and training on the many open science options that are available and suitable for animal research. Researchers have to be convinced of the benefits of open science practices, not only for science in general, but also for the individual researcher and each single animal. Yet, the key players in the research system are already starting to value open science practices. An increasing number of journals request open sharing of data, funders pay for open access publications and institutions consider open science practices in hiring decisions. Open science practices can improve the quality of work by enabling valuable scientific input from peers at the early stages of research projects. Furthermore, the extended communication that open science practices offer can draw attention to research and help to expand networks of collaborators and lead to new project opportunities or follow-up positions. Thus, open science practices can be a driver for careers in academia, particularly those of early career researchers.

Beyond these personal benefits, improving transparency in translational biomedical research can boost scientific progress in general. By bringing to light all the recorded research outputs that until now have remained hidden, the publication bias and the overestimation of effect sizes can be reduced [ 17 ]. Large-scale sharing of data can help to synthesize research outputs in preclinical research that will enable better decision-making for clinical research. Disclosing the whole research process will help to uncover systematic problems and support scientists in thoroughly planning their studies. In the long run, we predict that the implementation of open science practices will lead to the use of fewer animals in unintentionally repeated experiments that previously showed unreported negative results or in the establishment of methods by avoiding experimental dead ends that are often not published. More collaborations and sharing of materials and methods can further reduce the number of animal experiments used for the implementation of new techniques.

Open science can and should be implemented at each step of the research process ( Fig 1 ). A vast number of tools are already provided that were either directly conceptualized for animal research or can be adapted easily. In this Essay, we provide an overview of open science tools that improve transparency, reliability, and animal welfare in translational in vivo biomedical research by supporting scientists to clearly communicate their research and by supporting collaborative working. Table 1 lists the most prominent open science tools we discuss, together with their respective links. We have structured this Essay to guide you through which tools can be used at each stage of the research process, from planning and conducting experiments, through to analyzing data and communicating the results. However, many of these tools can be used at many different steps. Table 1 has been deposited on Zenodo and will be updated continuously [ 21 ].

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

Application of open science practices at each step of the research process can maximize the impact of performed animal experiments. The implementation of these practices will lead to less time pressure at the end of a project. Due to the connection of most of these open science practices, spending more time in the planning phase and during the conduction of experiments will save time during the data analysis and publication of the study. Indeed, consulting reporting guidelines early on, preregistering a statistical plan, and writing down crucial experimental details in an electronic lab notebook, will strongly accelerate the writing of a manuscript. If protocols or even electronic lab notebooks were made public, just citing these would simplify the writing of publications. Similarly, if a data management plan is well designed before starting data collection, analyzing, and depositing data in a public repository, as is increasingly required, will be fast. NTS, non-technical summary.

https://doi.org/10.1371/journal.pbio.3001810.g001

https://doi.org/10.1371/journal.pbio.3001810.t001

Planning the study

Transparent practices can be adopted at every stage of the research process. However, to ensure full effectivity, it is highly recommended to engage in detailed planning before the start of the experiment. This can prevent valuable time from being lost at the end of the study due to careless decisions being made at the beginning. Clarifying data management at the start of a project can help avoiding filing chaos that can be very time consuming to untangle. Keeping clear track of a project and study design will also help if new colleagues are included later on in the project or if entire project parts are handed over. In addition, all texts written on the rationale and hypothesis of the study or method descriptions, or design schemes created during the planning phase can be used in the final publications ( Fig 1 ). Similarly, information required for preregistration of animal studies or for reporting according to the ARRIVE guidelines are an extension of the details required for ethical approval [ 22 , 23 ]. Thus, the time burden within the planning phase is often overestimated. Furthermore, the thorough planning of experiments can avoid the unnecessary use of animals by preventing wrong avenues from being pursued.

Implementing open scientific practices at the beginning of a project does not mean that the idea and study plan must be shared immediately, but rather is critical for making the entire workflow transparent at the end of the project. However, optional early sharing of information can enable peers to give feedback on the study plan. Studies potentially benefit more from this a priori input than they would from the classical a posteriori peer-review process.

Most people perceive guidelines as advice that instructs on how to do something. However, it is sometimes useful to consider the term in its original meaning; “the line that guides us”. In this sense, following guidelines is not simply fulfilling a duty, but is a process that can help to design a sound research study and, as such, guidelines should be consulted at the planning stage of a project. The PREPARE guidelines are a list of important points that should be thought-out before starting a study involving animal experiments in order to reduce the waste of animals, promote alternatives, and increase the reproducibility of research and testing [ 24 ]. The PREPARE checklist helps to thoroughly plan a study and focuses on improving the communication and collaboration between all involved participants of the study (i.e., animal caretakers and scientists). Indeed, open science begins with the communication within a research facility. It is currently available in 33 languages and the responsible team from Norecopa, Norway’s 3R-center, takes requests for translations into further languages.

The UK Reproducibility Network has also published several guiding documents (primers) on important topics for open and reproducible science. These address issues such as data sharing [ 25 ], open access [ 26 ], open code and software [ 27 ], and preprints [ 28 ], as well as preregistration and registered reports [ 27 ]. Consultation of these primers is not only helpful in the relevant phases of the experiment but is also encouraged in the planning phase.

Although the ARRIVE guidelines are primarily a reporting guideline specifically designed for preparing a publication containing animal data, they can also support researchers when planning their experiments [ 22 , 23 ]. Going through the ARRIVE website, researchers will find tools and explanations that can support them in planning their experiments [ 29 ]. Consulting the ARRIVE checklist at the beginning of a project can help in deciding what details need to be documented during conduction of the experiments. This is particularly advisable, given that compliance to ARRIVE is still poor [ 18 ].

Experimental design

To maximize the validity of performed experiments and the knowledge gained, designing the study well is crucial. It is important that the chosen animal species reflects the investigated disease well and that basic characteristics of the animal, such as sex or age, are considered carefully [ 30 ]. The Canadian Institutes of Health Research provides a collection of resources on the integration of sex and gender in biomedical research with animals, including tips and tools for researchers and reviewers [ 31 ]. Additionally, it is advisable to avoid unnecessary standardization of biological and environmental factors that can reduce the external validity of results [ 32 ]. Meticulous statistical planning can further optimize the use of animals. Free to use online tools for calculating sample sizes such as G*Power or the inVivo software package for R can further support animal researchers in designing their statistical plan [ 33 , 34 ]. Randomization for the allocation of groups can be supported with specific tools for scientists like Research Randomizer, but also by simple online random number generators [ 35 ]. Furthermore, it might be advisable when designing the study to incorporate pathological analyses into the experimental plan. Optimal planning of tissue collection, performance of pathological procedures according to accepted best practices, and use of optimal pathological analysis and reporting methods can add some extra knowledge that would otherwise be lost. This can improve the reproducibility and quality of translational biomedicine, especially, but not exclusively, in animal studies with morphological endpoints. In all animal studies, unexpected deaths in experimental animals can occur and be the cause of lost data or missed opportunities to identify health problems [ 36 , 37 ].

To support researchers in designing their animal research, the National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC3Rs) has also developed the Experimental Design Assistant (EDA) [ 38 , 39 ]. This online tool helps researchers to better structure in vivo research by creating detailed schemes of the study design. It provides feedback on the entered design, drawing researcher’s attention to crucial decisions in the project. The resulting schemes can be used to transparently share the study design by uploading it into a study preregistration, enclosing it in a grant application, or submitting it with a final manuscript. The EDA can be used for different study designs in diverse scenarios and helps to communicate researcher plans to others [ 40 ]. The EDA might be particularly of interest to clarify very complex study designs involving multiple experimental groups. Working with the EDA might appear rather complex in the beginning, but the NC3R provides regular webinars that can help to answer any questions that arise.

Preregistration

Preregistration is an effective tool to improve the quality and transparency of research. To preregister their work, scientists must determine crucial details of the study before starting any experiment. Changes occurring during a study can be outlined at the end. A preregistered study plan should include at least the hypothesis and determine all the parameters that are known in advance. A description of the planned study design and statistical analysis will enable reviewers and peers to better retrace the workflow. It can prevent the intentional use of the flexibility of analysis to reach p -values under a certain significance level (e.g., p-hacking or HARKing (Hypothesizing After Results are Known)). With preregistration, scientists can also claim their idea at an early stage of their research with a citable individual identifier that labels the idea as their own. Some open preregistration platforms also provide a digital object identifier (DOI), which makes the registered study citable. Three public registries actively encourage the preregistration of animal studies conducted around the world: OSF registry, preclinicaltrials.eu, and animalstudyregistry.org [ 41 – 45 ]. Scientists can choose the registry according to their needs. Preregistering a study in a public registry supports scientists in planning their study and later to critically reevaluate their own work and assess its limitations and potentials.

As an alternative to public registries, researchers can also submit their study plan to one of hundreds of journals already publishing registered reports, including many journals open to animal research [ 8 ]. A submitted registered report passes 2 steps of peer review. In the first step, reviewers comment on the idea and the study design. After an “in-principle-acceptance,” researchers can conduct their study as planned. If the authors conduct the experiments as described in the accepted study protocol, the journal will publish the final study regardless of the outcome. This might be an attractive option, especially for early career researchers, as a manuscript is published at the beginning of a project with the guarantee of a future final publication.

The benefits of preregistration can already be observed in clinical research, where registration has been mandatory for most trials for more than 20 years. Preregistration in clinical research has helped to make known what has been tested and not just what worked and was published, and the implementation of trial registration has strongly reduced the number of publications reporting significant treatment effects [ 46 ]. In animal research, with its unrealistically high percentage of positive results, preregistration seems to be particularly worthwhile.

Research data management

To get the most out of performed animal experiments, effective sharing of data at the end of the study is essential. Sharing research data optimally is complex and needs to be prepared in advance. Thus, data management can be seen as one part of planning a study thoroughly. Many funders have recognized the value of the original research data and request a data management plan from applicants in advance [ 25 , 47 ]. Various freely available tools such as DMPTool or DMPonline already exist to design a research data management plan that complies to the requirements of different funders [ 48 , 49 ]. The data management plan defines the types of data collected and describes the handling and names responsible persons throughout the data lifecycle. This includes collecting the data, analyzing, archiving, and sharing it. Finally, a data management plan enables long-term access and the possibility for reuse by peers. Developing such a plan, whether it is required by funders or not, will later simplify the application of the FAIR data principle (see section on the FAIR data principle). The Longwood Medical Area Research Data Management Working Group from the Harvard Medical School developed a checklist to assist researchers in optimally managing their data throughout the data lifecycle [ 50 ]. Similarly, the Joint Information Systems Committee (JISC) provides a great research data management toolkit including a checklist for researchers planning their project [ 51 ]. Consulting this checklist in the planning phase of a project can prevent common errors in research data management.

Non-technical project summary

One instrument specifically conceived to create transparency on animal research for the general public is the so-called non-technical project summary (NTS). All animal protocols approved within the EU must be accompanied by these comprehensible summaries. NTSs are intended to inform the public about ongoing animal experiments. They are anonymous and include information on the objectives and potential benefits of the project, the expected harm, the number of animals, the species, and a statement of compliance with the requirements of the 3Rs principle. However, beyond simply informing the public, NTSs can also be used for meta-research to help identify new research areas with an increased need for new 3R technologies [ 52 , 53 ]. NTSs become an excellent tool to appropriately communicate the scientific value of the approved protocol and for meta-scientists to generate added value by systematically analyzing theses summaries if they fulfill a minimum quality threshold [ 54 , 55 ]. In 2021, the EU launched the ALURES platform ( Table 1 ), where NTSs from all member states are published together, opening the opportunities for EU-wide meta-research. NTSs are, in contrast to other open science practices, mandatory in the EU. However, instead of thinking of them as an annoying duty, it might be worth thoroughly drafting the NTS to support the goals of more transparency towards the public, enabling an open dialogue and reducing extreme opinions.

Conducting the experiments

Once the experiments begin, documentation of all necessary details is essential to ensure the transparency of the workflow. This includes methodological details that are crucial for replicating experiments, but also failed attempts that could help peers to avoid experiments that do not work in the future. All information should be stored in such a way that it can be found easily and shared later. In this area, many new tools have emerged in recent years ( Table 1 ). These tools will not only make research transparent for colleagues, but also help to keep track of one’s own research and improve internal collaboration.

Electronic laboratory notebooks

Electronic laboratory notebooks (ELNs) are an important pillar of research data management and open science. ELNs facilitate the structured and harmonized documentation of the data generation workflow, ensure data integrity, and keep track of all modifications made to the original data based on an audit trail option. Moreover, ELNs simplify the sharing of data and support collaborations within and outside the research group. Methodological details and research data become searchable and traceable. There is an extensive amount of literature providing advice on the selection and the implementation process of an ELN depending on the specific needs and research area and its discussion would be beyond the scope of this Essay [ 56 – 58 ]. Some ELNs are connected to a laboratory information management system (LIMS) that provides an animal module supporting the tracking of animal details [ 59 ]. But as research involving animals is highly heterogeneous, this might not be the only decision point and we cannot recommend a specific ELN that is suitable for all animal research.

ELNs are already established in the pharmaceutical industry and their use is on the rise among academics as well. However, due to concerns around costs for licenses, data security, and loss of flexibility, many research institutions still fear the expenses that the introduction of such a system would incur [ 56 ]. Nevertheless, an increasing number of academic institutions are implementing ELNs and appreciating the associated benefits [ 60 ]. If your institution already has an ELN, it might be easiest to just use the option available in the research environment. If not, the Harvard Medical School provides an extensive and updated overview of various features of different ELNs that can support scientists in choosing the appropriate one for their research [ 61 ]. There are many commercial ELN products, which may be preferred when the administrative workload should be outsourced to a large extent. However, open-source products such as eLabFTW or open BIS provide a greater opportunity for customization to meet specific needs of individual research institutions [ 62 – 64 ]. A huge number of options are available depending on the resources and the features required. Some scientists might prefer generic note taking tools such as Evernote or just a simple Word document that offers infinite flexibility, but specific ELNs can further support good record keeping practice by providing immutability, automated backups, standardized methods, and protocols to follow. Clearly defining the specific requirements expected might help to choose an adequate system that would improve the quality of the record compared to classical paper laboratory notebooks.

Sharing protocols

Adequate sharing of methods in translational biomedical sciences is key to reproducibility. Several repositories exist that simplify the publication and exchange of protocols. Writing down methods at the end of the project bears the risk that crucial details might be missing [ 65 ]. On protocols.io, scientists can note all methodological details of a procedure, complete them with uploaded documents, and keep them for personal use or share them with collaborators [ 66 ]. Authors can also decide at any point in time to make their protocol public. Protocols published on protocols.io receive a DOI and become citable; they can be commented on by peers and adapted according to the needs of the individual researcher. Protocol.io files from established protocols can also be submitted together with some context and sample datasets to PLOS ONE , where it can be peer-reviewed and potentially published [ 67 , 68 ]. Depending on the affiliation of the researchers to academia or industry and on an internal or public sharing of files, protocols.io can be free of charge or come with costs. Other journals also encourage their authors to deposit their protocols in a freely accessible repository, such as protocol exchange from Nature portfolio [ 69 ]. Another option might be to separately submit a protocol that was validated by its use in an already published research article to an online and peer-reviewed journal specific for research protocols, such as Bio-Protocol. A multitude of journals, including eLife and Science already collaborate with Bio-Protocol and recommend authors to publish the method in Bio-Protocol [ 70 ]. Bio-Protocol has no submission fees and is freely available to all readers. Both protocols.io and Bio-Protocol allow the illustration of complex scientific methods by uploading videos to published protocols. In addition, protocols can be deposited in a general research repository such as the Open Science Framework (OSF repository) and referenced in appropriate publications.

Sharing critical incidents

Sharing critical or even adverse events that occur in the context of animal experimentation can help other scientists to avoid committing the same mistakes. The system of sharing critical incidents is already established in clinical practice and helps to improve medical care [ 71 , 72 ]. The online platform critical incident reporting system in laboratory animal science (CIRS-LAS) represents the first preclinical equivalent to these clinical systems [ 73 ]. With this web-based tool, critical incidents in animal research can be reported anonymously without registration. An expert panel helps to analyze the incident to encourage an open dialogue. Critical incident reporting is still very marginal in animal research and performed procedures are very variable. These factors make a systemic analysis and a targeted search of incidence difficult. However, it may be of special interest for methods that are broadly used in animal research such as anesthesia. Indeed, a broad feed of this system with data on errors occurring in standard procedures today could help avoid critical incidences in the future and refine animal experiments.

Sharing animals, organs, and tissue

When we think about open science, sharing results and data are often in focus. However, sharing material is also part of a collaborative and open research culture that could help to greatly reduce the number of experimental animals used. When an animal is killed to obtain specific tissue or organs, the remainder is mostly discarded. This may constitute a wasteful practice, as surplus tissue can be used by other researchers for different analyses. More animals are currently killed as surplus than are used in experiments, demonstrating the potential for sharing these animals [ 74 , 75 ].

Sharing information on generated surplus is therefore not only economical, but also an effective way to reduce the number of animals used for scientific purposes. The open-source software Anishare is a straightforward way for breeders of genetically modified lines to promote their surplus offspring or organs within an institution [ 76 ]. The database AniMatch ( Table 1 ) connects scientists within Europe who are offering tissue or organs with scientists seeking this material. Scientists already sharing animal organs can support this process by describing it in publications and making peers aware of this possibility [ 77 ]. Specialized research communities also allow sharing of animal tissue or animal-derived products worldwide that are typically used in these fields on a collaborative basis via the SEARCH-framework [ 78 , 79 ]. Depositing transgenic mice lines into one of several repositories for mouse strains can help to further minimize efforts in producing new transgenic lines and most importantly reduce the number of surplus animals by supporting the cryoconservation of mouse lines. The International Mouse Strain Resource (IMSR) can be used to help find an adequate repository or to help scientists seeking a specific transgenic line find a match [ 80 ].

Analyzing the data

Animal researchers have to handle increasingly complex data. Imaging, electrophysiological recording, or automated behavioral tracking, for example, produce huge datasets. Data can be shared as raw numerical output but also as images, videos, sounds, or other forms from which raw numerical data can be generated. As the heterogeneity and the complexity of research data increases, infinite possibilities for analysis emerge. Transparently reporting how the data were processed will enable peers to better interpret reported results. To get the most out of performed animal experiments, it is crucial to allow other scientists to replicate the analysis and adapt it to their research questions. It is therefore highly recommended to use formats and tools during the analysis that allow a straightforward exchange of code and data later on.

Transparent coding

The use of non-transparent analysis codes have led to a lack of reproducibility of results [ 81 ]. Sharing code is essential for complex analysis and enables other researchers to reproduce results and perform follow-up studies, and citable code gives credit for the development of new algorithms ( Table 1 ). Jupyter Notebooks are a convenient way to share data science pipelines that may use a variety of coding languages, including like Python, R or Matlab, and also share the results of analyses in the form of tables, diagrams, images, and videos. Notebooks contain source code and can be published or collaboratively shared on platforms like GitHub or GitLab, where version control of source code is implemented. The data-archiving tool Zenodo can be used to archive a repository on GitHub and create a DOI for the archive. Thereby contents become citable. Using free and open-source programming language like R or Python will increase the number of potential researchers that can work with the published code. Best practice for research software is to publish the source code with a license that allows modification and redistribution.

Choice of data visualization

Choosing the right format for the visualization of data can increase its accessibility to a broad scientific audience and enable peers to better judge the validity of the results. Studies based on animal research often work with very small sample sizes. Visualizing these data in histograms may lead to an overestimation of the outcomes. Choosing the right dot plots that makes all recorded points visible and at the same time focusses on the summary instead of the individual points can further improve the intuitive understanding of a result. If the sample size is too low, it might not be meaningful to visualize error bars. A variety of freely available tools already exists that can support scientists in creating the most appropriate graphs for their data [ 82 ]. In particular, when representing microscopy results or heat maps, it should be kept in mind that a large part of the population cannot perceive the classical red and green representation [ 83 ]. Opting for the color-blind safe color maps and checking images with free tools such as color oracle ( Table 1 ) can increase the accessibility of graphs. Multiple journals have already addressed flaws in data visualization and have introduced new policies that will accelerate the uptake of transparent representation of results.

Publication of all study outcomes

Open science practices have received much attention in the past few years when it comes to publication of the results. However, it is important to emphasize that although open science tools have their greatest impact at the end of the project, good study preparation and sharing of the study plan and data early on can greatly increase the transparency at the end.

The FAIR data principle

To maximize the impact and outcome of a study, and to make the best long-term use of data generated through animal experiments, researchers should publish all data collected during their research according to the FAIR data principle. That means the data should be findable, accessible, interoperable, and reusable. The FAIR principle is thus an extension of open access publishing. Data should not only be published without paywalls or other access restrictions, but also in such a manner that they can be reused and further processed by others. For this, legal as well as technical requirements must be met by the data. To achieve this, the GoFAIR initiative has developed a set of principles that should be taken into account as early as at the data collection stage [ 49 , 84 ]. In addition to extensively described and machine-readable metadata, these principles include, for example, the application of globally persistent identifiers, the use of open file formats, and standardized communication protocols to ensure that humans and machines can easily download the data. A well-chosen repository to upload the data is then just the final step to publish FAIR data.

FAIR data can strongly increase the knowledge gained from performed animal experiments. Thus, the same data can be analyzed by different researchers and could be combined to obtain larger sample sizes, as already occurs in the neuroimaging community, which works with comparable datasets [ 85 ]. Furthermore, the sharing of data enables other researchers to analyze published datasets and estimate measurement reliabilities to optimize their own data collection [ 86 , 87 ]. This will help to improve the translation from animal research into clinics and simultaneously reduce the number of animal experiment in future.

Reporting guidelines

In preclinical research, the ARRIVE guidelines are the current state of the art when it comes to reporting data based on animal experiments [ 22 , 23 ]. The ARRIVE guidelines have been endorsed by more than 1,000 journals who ask that scientists comply with them when reporting their outcomes. Since the ARRIVE guidelines have not had the expected impact on the transparency of reporting in animal research publications, a more rigorous update has been developed to facilitate their application in practice (ARRIVE 2.0 [ 23 ]). We believe that the ARRIVE guidelines can be more effective if they are implemented at a very early stage of the project (see section on guidelines). Some more specialized reporting guidelines have also emerged for individual research fields that rely on animal studies, such as endodontology [ 88 ]. The equator network collects all guidelines and makes them easily findable with their search tool on their website ( Table 1 ). MERIDIAN also offers a 1-stop shop for all reporting guidelines involving the use of animals across all research sectors [ 89 ]. It is thus worth checking for new reporting guidelines before preparing a manuscript to maximize the transparency of described experiments.

Identifiers

Persistent identifiers for published work, authors, or resources are key for making public data findable by search engines and are thus a prerequisite for compliance to FAIR data principles. The most common identifier for publications will be a DOI, which makes the work citable. A DOI is a globally unique string assigned by the International DOI Foundation to identify content permanently and provide a persistent link to its location on the Internet. An ORCID ID is used as a personal persistent identifier and is recommendable to unmistakably identify an author ( Table 1 ). This will avoid confusions between authors with the same name or in the case of name changes or changes of affiliation. Research Resource Identifiers (RRID) are unique ID numbers that help to transparently report research resources. RRID also apply to animals to clearly identify the species used. RRID help avoid confusion between different names or changing names of genetic lines and, importantly, make them machine findable. The RRID Portal helps scientists find a specific RRID or create one if necessary ( Table 1 ). In the context of genetically altered animal lines, correct naming is key. The Mouse Genome Informatics (MGI) Database is the authoritative source of official names for mouse genes, alleles, and strains ([ 90 ]).

Preprint publication

Preprints have undergone unprecedented success, particularly during the height of the Coronavirus Disease 2019 (COVID-19) pandemic when the need for rapid dissemination of scientific knowledge was critical. The publication process for scientific manuscripts in peer-reviewed journals usually requires a considerable amount of time, ranging from a few months to several years, mainly due to the lengthy review process and inefficient editorial procedures [ 91 , 92 ]. Preprints typically precede formal publication in scientific journals and, thus, do not go through a peer review process, thus, facilitating the prompt open dissemination of important scientific findings within the scientific community. However, submitted papers are usually screened and checked for plagiarism. Preprints are assigned a DOI so they can be cited. Once a preprint is published in a journal, its status is automatically updated on the preprint server. The preprint is linked to the publication via CrossRef and mentioned accordingly on the website of the respective preprint platform.

After initial skepticism, most publishers now allow papers to be posted on preprint servers prior to submission. An increasing number of journals even allow direct submission of a preprint to their peer review process. The US National Institutes of Health and the Wellcome Trust, among other funders, also encourage prepublication and permit researchers to cite preprints in their grant applications. There are now numerous preprint repositories for different scientific disciplines. BioASAP provides a searchable database for preprint servers that can help in identifying the one that best matches an individual’s needs [ 93 ]. The most popular repository for animal research is bioRxiv, which is hosted by the Cold Spring Harbor Laboratory ( Table 1 ).

The early exchange of scientific results is particularly important for animal research. This acceleration of the publication process can help other scientists to adapt their research or could even prevent animal experiments if other scientists become aware that an experiment has already been done before starting their own. In addition, preprints can help to increase the visibility of research. Journal articles that have a corresponding preprint publication have higher citation and Altmetric counts than articles without preprint [ 94 ]. In addition, the publication of preprints can help to combat publication bias, which represents a major problem in animal research [ 16 ]. Since journals and readers prioritize cutting-edge studies with positive results over inconclusive or negative results, researchers are reluctant to invest time and money in a manuscript that is unlikely to be accepted in a high-impact journal.

In addition to the option of publishing as preprint, other alternative publication formats have recently been introduced to facilitate the publication of research results that are hard to publish in traditional peer-reviewed journals. These include micro publications, data repositories, data journals, publication platforms, and journals that focus on negative or inconclusive results. The tool fiddle can support scientists in choosing the right publication format [ 95 , 96 ].

Open access publication

Publishing open access is one of the most established open science strategies. In contrast to the FAIR data principle, the term open access publication refers usually to the publication of a manuscript on a platform that is accessible free of charge—in translational biomedical research, this is mostly in the form of a scientific journal article. Originally, publications accessible free of charge were the answer to the paywalls established by renowned publishing houses, which led to social inequalities within and outside the research system. In translational biomedical research, the ethical aspect of urgently needed transparency is another argument in favor of open access publication, as these studies will not only be findable, but also internationally readable.

There are different ways of open access publishing; the 2 main routes are gold open access and green open access. Numerous journals offer now gold open access. It refers to the immediate and fully accessible publication of an article. The Directory of Open Access Journals (DOAJ) provides a complete and updated list for high-quality, open access, and peer-reviewed journals [ 97 ]. Charité–Universitätsmedizin Berlin offers a specific tool for biomedical open access journals that supports animal researchers to choose an appropriate journal [ 49 ]. In addition, the Sherpa Romeo platform is a straightforward way to identify publisher open access policies on a journal-by-journal basis, including information on preprints, but also on licensing of articles [ 51 ]. Hybrid open access refers to openly accessible articles in otherwise paywalled journals. By contrast, green open access refers to the publication of a manuscript or article in a repository that is mostly operated by institutions and/or universities. The publication can be exclusively on the repository or in combination with a publisher. In the quality-assured, global Directory of Open Access Repositories (openDOAR), scientists can find thousands of indexed open access repositories [ 49 ]. The publisher often sets an embargo during which the authors cannot make the publication available in the repository, which can restrict the combined model. It is worth mentioning that gold open access is usually more expensive for the authors, as they have to pay an article processing charge. However, the article’s outreach is usually much higher than the outreach of an article in a repository or available exclusively as subscription content [ 98 ]. Diamond open access refers to publications and publication platforms that can be read free of charge by anyone interested and for which no costs are incurred by the authors either. It is the simplest and fairest form of open access for all parties involved, as no one is prevented from participating in scientific discourse by payment barriers. For now, it is not as widespread as the other forms because publishers have to find alternative sources of revenue to cover their costs.

As social media and the researcher’s individual public outreach are becoming increasingly important, it should be remembered that the accessibility of a publication should not be confused with the licensing under which the publication is made available. In order to be able to share and reuse one’s own work in the future, we recommend looking for journals that allow publications under the Creative Commons licenses CC BY or CC BY-NC. This also allows the immediate combination of gold and green open access.

Creative commons licenses

Attributing Creative Commons (CC) licenses to scientific content can make research broadly available and clearly specifies the terms and conditions under which people can reuse and redistribute the intellectual property, namely publications and data, while giving the credit to whom it deserves [ 49 ]. As the laws on copyright vary from country to country and law texts are difficult to understand for outsiders, the CC licenses are designed to be easily understandable and are available in 41 languages. This way, users can easily avoid accidental misuse. The CC initiative developed a tool that enables researchers to find the license that best fits their interests [ 49 ]. Since the licenses are based on a modular concept ranging from relatively unrestricted licenses (CC BY, free to use, credit must be given) to more restricted licenses (CC BY-NC-ND, only free to share for non-commercial purposes, credit must be given), one can find an appropriate license even for the most sensitive content. Publishing under an open CC license will not only make the publication easy to access but can also help to increase its reach. It can stimulate other researchers and the interested public to share this article within their network and to make the best future use of it. Bear in mind that datasets published independently from an article may receive a different CC license. In terms of intellectual property, data are not protected in the same way as articles, which is why the CC initiative in the United Kingdom recommends publishing them under a CC0 (“no rights reserved”) license or the Public Domain Mark. This gives everybody the right to use the data freely. In an animal ethics sense, this is especially important in order to get the most out of data derived from animal experiments.

Data and code repositories

Sharing research data is essential to ensure reproducibility and to facilitate scientific progress. This is particularly true in animal research and the scientific community increasingly recognizes the value of sharing research data. However, even though there is increasing support for the sharing of data, researchers still perceive barriers when it comes to doing so in practice [ 99 – 101 ]. Many universities and research institutions have established research data repositories that provide continuous access to datasets in a trusted environment. Many of these data repositories are tied to specific research areas, geographic regions, or scientific institutions. Due to the growing number and overall heterogeneity of these repositories, it can be difficult for researchers, funding agencies, publishers, and academic institutions to identify appropriate repositories for storing and searching research data.

Recently, several web-based tools have been developed to help in the selection of a suitable repository. One example is Re3data, a global registry of research data repositories that includes repositories from various scientific disciplines. The extensive database can be searched by country, content (e.g., raw data, source code), and scientific discipline [ 49 ]. A similar tool to help find a data archive specific to the field is FAIRsharing, based at Oxford University [ 102 ]. If there is no appropriate subject-specific data repository or one seems unsuitable for the data, there are general data repositories, such as Open Science Framework, figshare, Dryad, or Zenodo. To ensure that data stored in a repository can be found, a DOI is assigned to the data. Choosing the right license for the deposited code and data ensures that authors get credit for their work.

Publication and connection of all outcomes

If scientists have used all available open science tools during the research process, then publishing and linking all outcomes represents the well-deserved harvest ( Fig 2 ). At the end of a research process, researchers will not just have 1 publication in a journal. Instead, they might have a preregistration, a preprint, a publication in a journal, a dataset, and a protocol. Connecting these outcomes in a way that enables other scientists to better assess the results that link these publications will be key. There are many examples of good open science practices in laboratory animal science, but we want to highlight one of them to show how this could be achieved. Blenkuš and colleagues investigated how mild stress-induced hyperthermia can be assessed non-invasively by thermography in mice [ 103 ]. The study was preregistered with animalstudyregistry.org , which is referred to in their publication [ 104 ]. A deviation from the originally preregistered hypothesis was explained in the manuscript and the supplementary material was uploaded to figshare [ 105 ].

Application of open science practices can increase the reproducibility and visibility of a research project at the same time. By publishing different research outputs with more detailed information than can be included in a journal article, researchers enable peers to replicate their work. Reporting according to guidelines and using transparent visualization will further improve this reproducibility. The more research products that are generated, the more credit can be attributed. By communicating on social media or additionally publishing slides from delivered talks or posters, more attention can be raised. Additionally, publishing open access and making the work machine-findable makes it accessible to an even broader number of peers.

https://doi.org/10.1371/journal.pbio.3001810.g002

It might also be helpful to provide all resources from a project in a single repository such as Open Science Framework, which also implements other, different tools that might have been used, like GitHub or protocols.io.

Communicating your research

Once all outcomes of the project are shared, it is time to address the targeted peers. Social media is an important instrument to connect research communities [ 106 ]. In particular, Twitter is an effective way to communicate research findings or related events to peers [ 107 ]. In addition, specialized platforms like ResearchGate can support the exchange of practical experiences ( Table 1 ). When all resources related to a project are kept in one place, sharing this link is a straightforward way to reach out to fellow scientists.

With the increasing number of publications, science communication has become more important in recent years. Transparent science that communicates openly with the public contributes to strengthening society’s trust in research.

Conclusions

Plenty of open science tools are already available and the number of tools is constantly growing. Translational biomedical researchers should seize this opportunity, as it could contribute to a significant improvement in the transparency of research and fulfil their ethical responsibility to maximize the impact of knowledge gained from animal experiments. Over and above this, open science practices also bear important direct benefits for the scientists themselves. Indeed, the implementation of these tools can increase the visibility of research and becomes increasingly important when applying for grants or in recruitment decisions. Already, more and more journals and funders require activities such as data sharing. Several institutions have established open science practices as evaluation criteria alongside publication lists, impact factor, and h-index for panels deciding on hiring or tenure [ 108 ]. For new adopters, it is not necessary to apply all available practices at once. Implementing single tools can be a safe approach to slowly improve the outreach and reproducibility of one’s own research. The more open science products that are generated, the more reproducible the work becomes, but also the more the visibility of a study increases ( Fig 2 ).

As other research fields, such as social sciences, are already a step ahead in the implementation of open science practices, translational biomedicine can profit from their experiences [ 109 ]. We should thus keep in mind that open science comes with some risks that should be minimized early on. Indeed, the more open science practices become incentivized, the more researchers could be tempted to get a transparency quality label that might not be justified. When a study is based on a bad hypothesis or poor statistical planning, this cannot be fixed by preregistration, as prediction alone is not sufficient to validate an interpretation [ 110 ]. Furthermore, a boom of data sharing could disconnect data collectors and analysts, bearing the risk that researchers performing the analysis lack understanding of the data. The publication of datasets could also promote a “parasitic” use of a researcher’s data and lead to scooping of outcomes [ 111 ]. Stakeholders could counteract such a risk by promoting collaboration instead of competition.

During the COVID-19 pandemic, we have seen an explosion of preprint publications. This unseen acceleration of science might be the adequate response to a pandemic; however, the speeding up science in combination with the “publish or perish” culture could come at the expense of the quality of the publication. Nevertheless, a meta-analysis comparing the quality of reporting between preprints and peer-reviewed articles showed that the quality of reporting in preprints in the life sciences is at most slightly lower on average compared to peer-reviewed articles [ 112 ]. Additionally, preprints and social media have shown during this pandemic that a premature and overconfident communication of research results can be overinterpreted by journalists and raise unfounded hopes or fears in patients and relatives [ 113 ]. By being honest and open about the scope and limitations of the study and choosing communication channels carefully, researchers can avoid misinterpretation. It should be noted, however, that by releasing all methodological details and data in research fields such as viral engineering, where a dual use cannot be excluded, open science could increase biosecurity risk. Implementing access-controlled repositories, application programming interfaces, and a biosecurity risk assessment in the planning phase (i.e., by preregistration) could mitigate this threat [ 114 ].

Publishing in open access journals often involves higher publication costs, which makes it more difficult for institutes and universities from low-income countries to publish there [ 115 ]. Equity has been identified as a key aim of open science [ 116 ]. It is vital, therefore, that existing structural inequities in the scientific system are not unintentionally reinforced by open science practices. Early career researchers have been the main drivers of the open science movement in other fields even though they are often in vulnerable positions due to short contracts and hierarchical and strongly networked research environments. Supporting these early career researchers in adopting open science tools could significantly advance this change in research culture [ 117 ]. However, early career researchers can already benefit by publishing registered reports or preprints that can provide a publication much faster than conventional journal publications. Communication in social media can help them establish a network enabling new collaborations or follow-up positions.

Even though open science comes with some risks, the benefits easily overweigh these caveats. If a change towards more transparency is accompanied by the implementation of open science in the teaching curricula of the universities, most of the risks can be minimized [ 118 ]. Interestingly, we have observed that open science tools and infrastructure that are specific to animal research seem to mostly come from Europe. This may be because of strict regulations within Europe for animal experiments or because of a strong research focus in laboratory animal science along with targeted research funding in this region. Whatever the reason might be, it demonstrates the important role of research policy in accelerating the development towards 3Rs and open science.

Overall, it seems inevitable that open science will eventually prevail in translational biomedical research. Scientists should not wait for the slow-moving incentive framework to change their research habits, but should take pioneering roles in adopting open science tools and working towards more collaboration, transparency, and reproducibility.

Acknowledgments

The authors gratefully acknowledge the valuable input and comments from Sebastian Dunst, Daniel Butzke, and Nils Körber that have improved the content of this work.

• View Article
• PubMed/NCBI
• Google Scholar
• 6. Cary Funk MH, Brian Kennedy, Courtney Johnson. Americans say open access to data and independent review inspire more trust in research findings. Pew Research Center Website: Pew Research Center; 2019. Available from: https://www.pewresearch.org/science/2019/08/02/americans-say-open-access-to-data-and-independent-review-inspire-more-trust-in-research-findings/ .
• 7. International Reproducibility Networks. International Networks Statement UK Reproducibility Network Website: UK Reproducibility Network. 2021. Available from: https://cpb-eu-w2.wpmucdn.com/blogs.bristol.ac.uk/dist/b/631/files/2021/09/International-Networks-Statement-v1.0.pdf .
• 13. Article 36 of Directive 2010/63/EU of the European Parliament and of the Council of 22 September 2010 amended by Regilation (EU) 2019/1010 of the European Parliament and of the Council of 5 June 2019. OJEU. 2010;L276:36.
• 19. American Association for Cancer Research. Editorial Policies. 2021. Available from: https://aacrjournals.org/content/authors/editorial-policies .
• 21. Diederich K, Schmitt K, Schwedhelm P, Bert B, Heinl C. Open Science Toolbox for Animal Research. Zenodo. 2022. Available from: https://zenodo.org/record/6497560 .
• 29. NC3R. ARRIVE guidelines. NC3R Website. Available from: https://arriveguidelines.org/ .
• 31. Canadian Institutes of Health Research. How to integrate sex and gender into research. Website of the Canadian Institutes of Health Research: Canadian Institutes of Health Research. 2019 [cited 2019 Aug 21]. Available from: https://cihr-irsc.gc.ca/e/50836.html .
• 33. Simon T, Bate RAC. InVivoStat. Available from: https://invivostat.co.uk/ .
• 35. Urbaniak G, Plous S. Research randomizer (version 4.0) [computer software]. 2013.
• 47. Medical Research Council’s. Data sharing policy. UK Research and Innovation Website 2021. Available from: https://www.ukri.org/publications/mrc-data-sharing-policy/ .
• 48. University of California Curation Center. DMPTool. 2021. Available from: https://dmptool.org/ .
• 49. Digital Curation Centre. DMPOnline. Available from: https://dmponline.dcc.ac.uk/ . Digital Curation Centre; 2021.
• 50. Harvard Longwood Medical Area Research Data Management Working Group. Biomedical Data Lifecycle. Harvard Medical School Website: Harvard Medical School; 2021. Available from: https://datamanagement.hms.harvard.edu/about/what-research-data-management/biomedical-data-lifecycle .
• 51. Joint Information Systems Committee. Research data management toolkit JISC Website: JISC; 2018. Available from: https://www.jisc.ac.uk/guides/rdm-toolkit .
• 54. German Centre for the Protection of Laboratory Animals (Bf3R). NTPs—Nicht Technische Projektzusammenfassungen 3R-SMART; 2020. Available from: https://www.3r-smart.de/index.php?id=6895 .
• 55. Understanding Animal Research. Guide to writing non-technical summaries concordat on openness on animal research in the UK2018. Available from: https://concordatopenness.org.uk/guide-to-writing-non-technical-summaries .
• 56. Gerlach B, Untucht C, Stefan A. Electronic Lab Notebooks and Experimental Design Assistants. In: Bespalov A, Michel MC, Steckler T, editors. Good Research Practice in Non-Clinical Pharmacology and Biomedicine. Cham: Springer International Publishing; 2020. p. 257–75.
• 58. Adam BL, Birte L. ELN Guide: electronic laboratory notebooks in the context of research data management and good research practice–a guide for the life sciences. Cologne, Germany: ZB MED–Information Centre for Life Sciences; 2021.
• 59. AgileBio. LabCollector Website https://labcollector.com/labcollector-lims/features/modules/animals-module/2022 . Available from: https://labcollector.com/labcollector-lims/features/modules/animals-module/ .
• 61. Harvard Longwood Medical Area Research Data Management Working Group. Electronic Lab Notebook Comparison Matrix. Zenodo. 2021.
• 70. Bio-protocol. Collaborating Journals bio-protocol website2021. Available from: https://bio-protocol.org/default.aspx?dw=Collaborating .
• 76. Dinkel H. anishare: GitHub; [updated June 2018]. Available from: https://github.com/hdinkel/anishare .
• 89. O’Connor AM. MERIDIAN: Menagerie of Reporting guidelines Involving Animals. Iowa State University; 2022. Available from: https://meridian.cvm.iastate.edu/ .
• 90. The Jackson Laboratory. Mouse Nomenclature Home Page at the Mouse Genome Informatics website World Wide Web: The Jackson Laboratory,Bar Harbor, Maine. Available from: http://www.informatics.jax.org/mgihome/nomen/index.shtml .
• 97. Directory of Open Access Journals. Find open access journals & articles. Available from: https://doaj.org/ . Directory of Open Access Journals, [DOAJ]; 2021.
• 98. Gold Open Access research has greater societal impact as used more outside of academia [press release]. Springer Nature Website: Springer. Nature. 2020;30:2020.
• 104. Franco NH. Can we use infrared thermography for assessing emotional states in mice?—A comparison between handling-induced stress by different techniques. Available from: animalstudyregistry.org . German Federal Institute for Risk Assessment (BfR); 2020. https://doi.org/10.17590/asr.0000224

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals

Animal behaviour articles from across Nature Portfolio

Animal behaviour is the scientific study of the behaviour of animals. The discipline covers study under experimental conditions, behaviourism, or natural conditions, ethology.

Latest Research and Reviews

Mammals show faster recovery from capture and tagging in human-disturbed landscapes

In wildlife tagging, stress from capture and handling can alter post- release behavior and potentially study interpretations. This study of 42 mammal species shows that these effects diminish within 4–7 days, and quicker for animals in high human activity areas indicating adaptation to disturbance.

• Jonas Stiegler
• Cara A. Gallagher
• Niels Blaum

Cleaner fish with mirror self-recognition capacity precisely realize their body size based on their mental image

• Taiga Kobayashi
• Masanori Kohda
• Shumpei Sogawa

Early impact of domestication on aggressiveness, activity, and stress behaviors in zebrafish ( Danio rerio ) using mirror test and automated videotracking

• C. Chevalier

Environmental stress reduces shark residency to coral reefs

Environmental stress on coral reefs is increasing. Authors show a stress-related reduction in residency of grey reef sharks on the remote reefs of the Chagos Archipelago, promoting more diffuse space use and increased time away from the shallow forereefs.

• Michael J. Williamson
• Emma J. Tebbs
• David M. P. Jacoby

Dogs with prior experience of a task still overimitate their caregiver

• Louise Mackie
• Ludwig Huber

Cetacean passive acoustic network in the Belgian part of the North sea

• Arienne Calonge
• Patricia Navarro González
• Elisabeth Debusschere

News and Comment

Lonely? Playful? Why are dolphin attacks rising in Japan?

Dolphin ecologist Tadamichi Morisaka discusses common dolphin behaviours that could explain the encounters with people.

• Anna Ikarashi

Camera-carrying sea lions map uncharted areas of the ocean

Sensors mounted on the animals’ backs gather data on unexplored sea floor habitats.

• Helena Kudiabor

Fish can tell the direction of sounds — here’s how

New research finally unravels the mystery of how fish can hear directionally.

• Emily Bates

Dogs might have evolved to read your emotions

Emotional contagion between humans and canines comes from centuries of selective breeding, suggests a comparison with pet pigs.

• Freda Kreier

The plastic that biodegrades in your home compost

Enzymes embedded directly into the material allows PLA plastic to completely break down — plus a gel that can safely store proteins for shipping.

• Benjamin Thompson

Massive cicada emergence prompted raccoons to run wild

Meanwhile, deer kept a low profile — perhaps because the deafening noise of the insects made it hard to hear predators.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

• Open access
• Published: 06 March 2021

Research perspectives on animal health in the era of artificial intelligence

• Pauline Ezanno   ORCID: orcid.org/0000-0002-0034-8950 1 ,
• Sébastien Picault 1 ,
• Gaël Beaunée 1 ,
• Xavier Bailly 2 ,
• Facundo Muñoz 3 ,
• Raphaël Duboz 3 , 4 ,
• Hervé Monod 5 &
• Jean-François Guégan 3 , 6 , 7  

Veterinary Research volume  52 , Article number:  40 ( 2021 ) Cite this article

16k Accesses

38 Citations

10 Altmetric

Metrics details

Leveraging artificial intelligence (AI) approaches in animal health (AH) makes it possible to address highly complex issues such as those encountered in quantitative and predictive epidemiology, animal/human precision-based medicine, or to study host × pathogen interactions. AI may contribute (i) to diagnosis and disease case detection, (ii) to more reliable predictions and reduced errors, (iii) to representing more realistically complex biological systems and rendering computing codes more readable to non-computer scientists, (iv) to speeding-up decisions and improving accuracy in risk analyses, and (v) to better targeted interventions and anticipated negative effects. In turn, challenges in AH may stimulate AI research due to specificity of AH systems, data, constraints, and analytical objectives. Based on a literature review of scientific papers at the interface between AI and AH covering the period 2009–2019, and interviews with French researchers positioned at this interface, the present study explains the main AH areas where various AI approaches are currently mobilised, how it may contribute to renew AH research issues and remove methodological or conceptual barriers. After presenting the possible obstacles and levers, we propose several recommendations to better grasp the challenge represented by the AH/AI interface. With the development of several recent concepts promoting a global and multisectoral perspective in the field of health, AI should contribute to defract the different disciplines in AH towards more transversal and integrative research.

1 Introduction

Artificial intelligence (AI) encompasses a large range of theories and technologies used to solve problems of high logical or algorithmic complexity. It crosses many disciplines, including mechanistic modelling, software engineering, data science, and statistics (Figure  1 ). Introduced in the 1950s, many AI methods have been developed or extended recently with the improvement of computer performance. Recent developments have been fuelled by the interfaces created between AI and other disciplines, such as bio-medicine, as well as massive data from different fields, particularly those associated with healthcare [ 1 , 2 ].

Interactions between animal health (AH), artificial intelligence (AI), and closely related research domains. This illustration is pinpointing only the links between AH (in blue), AI and its main subfields (in red), and other related fields of research (in black). It can be naturally complexified through the interactions between AH and other research topics (e.g., human medicine) or between core disciplines (e.g., statistics and physics).

AI addresses three challenges that also make sense in animal health (AH): (1) understanding a situation and its dynamics, e.g., epidemic spread; (2) the perception of the environment, which corresponds in AH to the detection of patterns (e.g., repeated sequence of observations), forms (e.g., of a protein) and signals (e.g., increased mortality compared to a baseline) at different scales; (3) computer-based decision making, or, more realistically, human decision support (e.g., expert systems, diagnostic support, resource allocation).

To answer these challenges, a wide range of concepts and methods are developed in AI. This includes machine learning (ML), a widely known AI method nowadays, which has been developing since the 1980s [ 3 ]. Since the 2000s, deep learning is developing with the rise of big data and the continuous increasing of computing capacities, enabling the exploration of massive amount of information that cannot be processed by conventional statistical methods. In addition, this also includes methods and algorithms for solving complex problems, automating tasks or reasoning, integrating information from heterogeneous sources, or decision support (Figure  1 ). These methods are now uprising in the human health sector, but are still rarely used to study animal health issues that they would help to revisit.

Part of the scientific challenges faced in AH can be approached from a new perspective by using some of these AI methods to analyse the ever-increasing data collected on animals, pathogens, and their environment. AH research benefits from advances in machine and deep learning methods, e.g., in predictive epidemiology, individual-based precision medicine, and to study host–pathogen interactions [ 2 , 4 ]. These methods contribute to disease diagnosis and individual case detection, to more reliable predictions and reduced errors, to speed-up decisions and improved accuracy in risk analysis, and to better targeting interventions in AH [ 5 ]. AH research also benefits from scientific advances in other domains of AI. Knowledge representation and modelling of reasoning [ 6 ] allow more realistic representations of complex socio-biological systems such as those encountered in AH. Examples include processes related to decision-making and uncertainty management [ 7 , 8 ], as well as of patient life courses like in human epidemiology [ 9 ]. This contributes to making them more readable by noncomputer experts. In addition, advances in problem solving under constrained resource allocation [ 10 ], in autonomous agents [ 11 ], multi-agent systems [ 12 ], and multi-level systems [ 13 ], as well as on automatic computer code generation [ 14 ] can be mobilised to enhance efficient and reliable epidemiological models. Interestingly, this may aid to anticipate the effect of control and management decisions at different spatial and temporal scales (animal, herd, country…).

Conducting research at the AH/AI interface also leads to identify new challenges for AI, on themes common with human health but considering different contexts and perspectives [ 15 ]. First, taking into account the particular agro- and socio-economic conditions of production systems is crucial when dealing with AH. Animal production systems depend on human activities and decisions. They can be a source of income (e.g., livestock) or labour forces and source of food in family farming. Citizens have also high expectations in terms of ethics and animal welfare [ 16 ]. Conventional measures to control animal diseases may no longer be acceptable by society (e.g., mass culling during outbreaks [ 17 ], antimicrobial usage, [ 18 ]). Alternatives must be identified and assessed, and AI can contribute. For example, individual-based veterinarian medicine is emerging, mobilising both AI methods and new AH data streams, these data differing from data in human health [ 19 ]. The integration of data from deep sequencing in AH, including emerging technologies for studying the metabolome and epigenome, is also a challenge [ 20 , 21 ]. Second, interactions between animal species, in particular between domestic animals and wildlife, lead to specific infectious disease risks (e.g., multi-host pathogens such as for African swine fever, pathogens crossing the species barrier facilitated by frequent contacts and promiscuity). The intensity of such interactions could increase due to separate or synergistic actions of environmental (e.g., landscape homogenisation, land use change for agriculture development, climate change), demographic (e.g., increasing global demand for animal production) and societal (e.g., outdoor livestock management) pressures. In addition, working on multi-species disease networks provides crucial information on the underlying molecular mechanisms favouring interspecific transmission [ 22 ]. Third, animal populations are governed by recurrent decision-making that also impacts health management (e.g., trade, control measures). Economic criteria as consequences on livestock farmers’ incomes are therefore essential indicators for evaluating AH control strategies, which can sometimes be misunderstood or may be at odds with societal expectations. These specificities make the AH/AI interface a theme of interest to stimulate new methodological work and to solve some of old and current locks faced by AH research today. With the development of new concepts in health such as One Health, Ecohealth and Planetary Health, promoting multidisciplinarity, stakeholders’ participation, data sharing, and tackling the complexity of health issues (e.g., multi-host pathogen transmission, short and long-term climatic impacts on disease patterns [ 23 ]), AI could participate in this new development by making it possible to technically solve some of the complex problems posed.

Mobilising the literature published at the AH/AI interface between 2009 and 2019 (Additional file 1 A), focusing our literature search on mainly livestock and wildlife, as well as interviews conducted with French researchers positioned at this interface (Additional file 1 B), we identified the main research areas in AH in which AI is currently involved country-wide. We explored how AI methods contribute to revisiting AH questions and may help remove methodological or conceptual barriers within the field. We also analysed how AH questions interrogate and stimulate new AI technical or scientific developments. In this paper, we first discuss issues related to data collection, organisation and access (Section  1 ), then we discuss how AI methods contribute to revisiting our understanding of animal epidemiological systems (Section  2 ), to improving case detection and diagnosis at different scales (Section  3 ), and to anticipating pathogen spread and control in a wide range of scenarios in order to improve AH management, facilitate decision-making and stimulate innovation (Section  4 ). Finally, we present the possible obstacles and levers to the development of AI in modern AH (Section  5 ), before making recommendations to best address the new challenges represented by this AH/IA interface (Section  6 ).

2 Collect, organise and make accessible quality data

A central point for research in AH remains the quality and availability of data, at the different organization levels of living systems and therefore at different spatial and temporal scales [ 24 ]. Data of interest are diverse. They can be obtained thanks to molecular analysis (e.g., genomic, metagenomics, or metabolic data), from observational data on individuals (e.g., body temperature, behaviour, milk production and composition, weight, feed intake), or from the production system (e.g., herd structure, breeding, management of sanitary issues). They can also be obtained at larger scales, beyond herds or local groups of animals (e.g., epidemiological data, demographic events, commercial movements, meteorological data, land-use occupation).

Even though the acquisition of these massive and heterogeneous data remain challenging (e.g., metabolome data), a large and diverse amount is already collected: (i) through mandatory reporting in accordance with regulations (e.g., commercial movements of cattle, epidemio-surveillance platform), (ii) by automatic devices (e.g., sensors, video surveillance systems), and (iii) on an ad hoc basis as part of research programs. This leads to a very wide diversity of data properties, and therefore of their management, access and possible uses. These data can be specifically obtained for certain animals or herds (e.g., during cohort monitoring programs) or by private companies (e.g., pig trade movements such as in France, milk collection). This can limit accessibility to academics and public research. Globalisation and large-scale animal trade may generate the need to use data obtained at worldwide scale in AH, especially for quantitative epidemiology (e.g., transcontinental spread of pathogens, animal genetics and breed management) leading to standardisation issues [ 25 ].

Consideration should be given to future systems for observing, collecting and managing these data [ 26 ], and to practices aimed at better collaboration between stakeholders. While data management has always been an important element in applied research to facilitate their use and valorisation, it is now a strategic issue both in theoretical and more applied research, coupled with a technical and algorithmic challenge [ 27 , 28 , 29 ]. Indeed, producing algorithms to manage massive data flows and stocks, by optimising calculations, is a challenge, particularly in real time. It seems also necessary to make heterogeneous data sources interoperable, requiring dedicated methodological developments [ 25 ]. In addition, much of the data is private, with ownership often heterogeneous (e.g., multiple owners, non-centralised data, closed data) and sometimes unclear (e.g., lack of knowledge of the real owner of the data between, for example, farmers, the data collector or the farmers’ union). All this tends to considerably complicate access to the data, raises questions about intellectual property, and raises questions in relation to regulations with regards to data protection, e.g., the adaptation of regulation to AH while respecting the confidentiality of the personal data mobilised. What is the relevant business model for data collection or access to existing databases? What about the openness of AH data (e.g., duality between the notion of public good and the private nature of certain data) to make it possible to experiment in real situations and compare the performance of AI algorithms? Answering these questions would facilitate the collection and sharing of ad hoc data. AI, particularly when combining a participatory framework with expert systems and multi-agent systems, helps to build realistic representations of complex socio-biological systems. Thus, it proves to be an effective tool to promote the collaboration of different stakeholders in collective and optimised decision-making, and to assess of the impact of changes in uses and practices [ 30 ].

Encouraging experimentation of AI technologies at a territorial scale becomes crucial to favour their development, validate their performance, and measure their predictive quality. In AH, simplified access to data-generating facilities would allow innovative solutions to be tested on a larger scale and would accelerate their development and evaluation. Substantial expertise exists (e.g., epidemiological data platform, large cohorts, experimental farms) that could be put to good use. In addition, AI could help to revisit sampling methods for field data collection in AH and epidemiological surveillance, by better and more dynamically targeting the data to be collected while avoiding redundant collinear, non-necessary data.

3 Contribution of AI to better understand animal epidemiological systems

Recent technological advances involving AI approaches have made it possible to obtain vast quantities of measurements and observations, as well as to store and share these data more efficiently. This has resulted in an increasing requirement for appropriate data analytical methods. AI methods emerged as the response of the computer-science community to these requirements, leveraging the exponential improvements in computational power. In parallel, statistical methods have greatly evolved in the last few decades as well, e.g., with regards to dimensionality-reduction in the spaces of variables and parameters, variable selection, and model comparison and combination. The rise in computational power has unleashed the development of Bayesian inference through simulation or approximate methods [ 31 ]. Bayesian methods have, in turn, facilitated the integration of data from diverse sources, the incorporation of prior knowledge and allowed for inference on more complex and realistic models while changing the paradigm of statistical inference [ 32 , 33 , 34 ].

3.1 Better understanding the evolution of AH and socio-ecological systems in a One Health context

Learning methods can be used to do phylogenetic reconstructions, contributing in particular to new evolutionary scenarios of pathogens and their transmission pathways. For example, phylogenetic models offer an interesting perspective for identifying environmental bacterial strains with high infectious potentiality [ 35 ], or for predicting the existence of putative host reservoirs or vectors [ 36 ]. The analysis of pathogen sharing among hosts has been used to classify the potential reservoirs of zoonotic diseases using machine learning [ 37 ]. The analysis of pathogen genomes can also be used to identify genotypes of animal pathogens that are more likely to infect humans [ 38 ].

Using phenomenological niche models that rely on data distribution more than on hypotheses about ecological processes at play, disease occurrence data or retrospective serological data coupled with environmental variables can be related to the risk of being exposed to a pathogen. Thus, they can help monitor potential spillovers and emerging risks and anticipate the epidemic pathogen spread [ 39 ]. For instance, Artificial Neural Networks (ANN) have identified the level of genetic introgression between wild and domesticated animal populations in a spatialized context [ 40 ], which may help to understand gene diffusion in host × pathogen systems involving multiple host species, and characterise specimen pools at higher risk to act as pathogen spreaders or sinks. Other AI approaches such as multi-agent models, a more mechanistic approach, have been used in an explicit spatial context for vector-borne pathogen transmissions, and proved to be sufficiently versatile to be adapted to several other particular contexts [ 12 ].

It should be noted here that several studies reveal the relatively ancient nature of AI research in AH. Such AI methods have often made it possible to identify signals (e.g., genetic introgression) or even particular patterns or properties (e.g., importance of density-dependence in the vector-borne transmission) that are less visible or hardly detectable by more conventional statistical treatments.

All these approaches contribute to better understand pathogen transmission in complex system networks as generally observed for emerging infections in tropical, developing regions of the world. On this matter, an improved knowledge is key for protecting humans against these new threats, and AI/AH interfaces development and training in cooperation with the poorest countries would facilitate synergistic effects and actions to predict and tackle new disease threats.

3.2 Reliability, reproducibility and flexibility of mechanistic models in AH

Better understanding and predicting pathogen spread often requires an explicit and integrative representation of the mechanisms involved in the dynamics of AH systems, irrespective of the scale (within-host: [ 41 ]; along a primary production chain: [ 42 ]; in a territory: [ 43 , 44 ]; over a continent: [ 45 ]).

Mathematical (equations) or computer-based (simulations) models can be used. Such mechanistic models (i.e., which represent the mechanisms involved in the infection dynamics), when sufficiently modular to represent contrasted situations, make it possible to anticipate the effects of conventional but also innovative control measures (e.g., new candidate molecules, sensors, genomic selection; [ 46 ]).

However, to assess realistic control measures, mechanistic epidemiological models require the integration of observational data and knowledge from biology, epidemiology, evolution, ecology, agronomy, sociology or economics. Their development can rapidly face challenges of reliability, transparency, reproducibility, and usage flexibility. Moreover, these models are often developed de novo, making little use of previous models from other systems. Finally, these models, even based on realistic biological hypotheses, may be considered negatively as black boxes by end users (health managers), because the underlying assumptions often became hidden in the code or equations.

The integration of multiple modelling perspectives (e.g., disciplines, points of view, spatio-temporal scales) is an important question in the modelling-simulation field. Epidemiological modelling could benefit from existing tools and methods developed in this field [ 47 , 48 , 49 ]. Although essential, good programming practices alone [ 50 ] cannot meet these challenges [ 51 ]. Scientific libraries and platforms accelerate the implementation of the complex models often needed in AH. For example, the R library SimInf [ 52 ] helps integrate observational data into mechanistic models. The BROADWICK framework [ 53 ] provides reusable software components for several scales and modelling paradigms, but still requires modellers to write large amounts of computer code.

New methods at the crossroads between software engineering and AI can enhance transparency and reproducibility in mechanistic modelling, fostering communication between software scientists, modellers and AH researchers throughout the modelling process (e.g., assumption formulation, assessment, and revision). Knowledge representation methods from symbolic AI, formalised using advanced software engineering methods such as domain-specific languages (DSL, e.g., in KENDRICK for differential equation models: [ 54 ]), makes model components accessible in a readable structured text file instead of computer code. Hence, scientists from various disciplines and field managers can be more involved in the model design and evaluation. Scenario exploration and model revision also no longer require rewriting the model code.

Other AI methods can improve model flexibility and modularity. Autonomous software agents enable to represent various levels of abstraction and organisation [ 55 ], helping modellers go more easily back and forth within small and larger scales, and ensure that all relevant mechanisms are adequately formalised at proper scales (i.e., scale-dependency of determinants and drivers in hierarchical living systems). Combining knowledge representation (through a DSL) and such a multi-level agent-based simulation architecture (e.g., in EMULSION, Figure  2 , [ 56 ]) enables to encompass several types of models (e.g., compartmental, individual-based) and scales (e.g., individual, population, territory), and it tackles simultaneously the recurring needs for transparency, reliability and flexibility in modelling contagious diseases. This approach should also facilitate in the future the production of support decision tools for veterinary and public health managers and stakeholders.

AI at the service of mechanistic epidemiological modelling (adapted from [ 51 ] ) . A . Modellers develop each epidemiological model de novo, producing specific codes not easily readable by scientists from other disciplines or by model end-users. B . Using AI approaches to combine a domain-specific language and an agent-based software architecture enhances reproducibility, transparency, and flexibility of epidemiological models. A simulation engine reads text files describing the system to automatically produce the model code. Complementary add-ons can be added if required. Models are easier to transfer to animal health managers as decision support tools.

3.3 Extracting knowledge from massive data in basic AH biology

Supervised, unsupervised and semi-supervised learning methods facilitate basic research development in biology and biomedicine, for example by using morphological analyses to study cell mobility [ 57 ]. The use of classification approaches and smart filters allows nowadays to sort massive molecular data (e.g., data from high throughput sequencing and metagenomics). Metabolic, physiological and immunological signalling pathways are explored, and metabolites are identified and quantified in complex biological mixtures, which was before a major challenge [ 58 ]. In addition, diagnostic time may be reduced by developing image analysis processing (e.g., accelerated detection of clinical patterns; [ 59 , 60 ]), often necessary to study host–pathogen interactions in animal pathology. For example, the use of optimisation methods has improved the understanding of the fragmentation of prion assemblages, contributing to a significant reduction in the time required to diagnose neurodegenerative animal diseases, thus paving the way for identifying potential therapeutic targets [ 61 ]. In livestock breeding, there is a methodological transition underway from traditional prediction strategies to more advanced machine learning approaches including artificial neural networks, deep learning and Bayesian networks which are being used to improve the reliability of genetic predictions and further the understanding of phenotypes biology. [ 62 ].

In human health, new disciplines have emerged in the second half of the 20 th century at the interface between AI and flagship disciplines, such as cell biology and immunology. Interface disciplines have developed, e.g., computational biology and immunology, which today must spread to AH. Current human immunology is based on the description of fine molecular and cellular mechanisms (e.g., the number of known interleukins has increased considerably compared to the 1970s). The desire to understand the processes underlying immune responses has led to a revolution by inviting this discipline to focus on complex systems biology and AI-based approaches [ 63 ]. However, the imbalance between the numbers of immunologists and immunology modellers is hampering the fantastic growth of this new discipline.

As an additional level of complexity, the hierarchical nature of biological systems makes that at the individual level, animals including humans must be considered as holobionts made of myriads of hosted microbial forms that form discrete ecological units (i.e., infracommunities). The potential of AI to grasp such diversity and complexity (e.g., tissue-specific microbiotes) and to scaling-up to higher levels of organization (e.g., component and compound communities of microbes, including pathogens, circulating in herd and in a given region) is certainly tremendous and should be studied with the same vigour as recent development in computational biology and immunology [ 40 ].

4 Revisiting AH case detection methods at different scales

Managing livestock health issues requires effective case detection methods, at the individual or even infra-individual (organ) scale, at the group/herd scale, or at larger scales (e.g., territories, countries). Machine learning methods allow detecting patterns and signals in massive data, e.g., in spatial data or time-series of health syndromes and disease cases, contributing to the development of smart agriculture and telemedicine (Figure  3 ). Alerts can be produced, and contribute to management advice in numerical agriculture [ 64 ] and veterinary practices [ 65 ]. AI may contribute to an earlier detection of infected cases and the rationalisation of treatments (including antimicrobials) in farm animals, by analysing data collected from connected sensors [ 66 ], by targeting individuals or groups of animals [ 59 ], or even by using mechanistic models to predict the occurrence of case detections and their treatment [ 67 ]. Also, machine learning methods enable to discriminate pathogen strains and thus to better understand their respective transmission pathways if different [ 68 ]. Finally, therapeutic strategies can be reasoned through multi-criteria optimisation, by identifying whom to treat in a herd, when, according to what protocol and for how long, in order to maximise the probability of cure while minimising both the risk of drug resistance and the volume or number of doses that are necessary (i.e., individual-based and precision medicine).

Extracting information from massive data to monitor animal health and better rationalise treatments.

Nevertheless, alert quality depends on the quality and representativeness of the datasets used by the learning algorithms. Numerous biases (e.g., hardware, software, human) can affect prediction accuracy. Moreover, alerts produced after training necessarily reflect the specificities of the system from which the data originates (e.g., area, period, rearing practices). Thus, result transposition to other epidemiological systems or to the same system subjected to environmental or regulatory changes remains risky. Furthermore, while machine learning methods (e.g., classification, image analysis, pattern recognition, data mining) provide solutions for a wide range of biomedical and bio-health research questions, it is crucial to demonstrate the performance of these methods by measuring their predictive quality and comparing them to alternative statistical methods whenever possible [ 69 ].

At the population level, case detection is based on direct (detection of syndromes) or indirect surveillance, mobilising syndrome proxies. Hence, the emergence of some animal diseases can be detected by syndromic surveillance, by detecting abnormal or rare signals in routine data (e.g., mortality, reproduction, abortion, behaviour, milk production, increased drug use; [ 70 ]). Also, serological data can be used retrospectively to identify individual characteristics related to a risk of being exposed to a pathogen, and thus orientate management efforts (e.g., in wildlife; [ 71 ]). Statistics and AI are largely complementary to address such issues. Both mobilise the wide range of available data, which are highly heterogeneous, massive and mostly sparse, to detect signals that are often weak or scarce [ 28 , 72 , 73 ]. Such signals can be proxy records (e.g., emergence of infectious diseases following environmental disturbances), health symptoms and syndromes, or even metabolic pathways in cascades which can be precursors of chronic or degenerative diseases. AI also includes methods to mobilise information available on the web. For example, semi-automatic data mining methods enabled to identify emerging signals for international surveillance of epizooties [ 74 ] or to analyse veterinary documents such as necropsy reports [ 75 , 76 ]. Methods from the field of natural language processing can compensate the scarcity of data by extracting syntactic and semantic information from textual records, triggering alerts on new emerging threats that could have been missed otherwise.

On a large to very large scale (i.e., territory, country, continent, global), data analysis of commercial animal movements between farms makes it possible to predict the associated epidemic risk [ 77 , 78 ]. These movements are difficult to predict, particularly since animal trade is based on many factors associated with human activities and decisions. Methods for recognising spatio-temporal patterns and methodological developments for the analysis of oriented and weighted dynamic relational graphs are required in this field because very few of the existing methods allow large-scale systems to be studied, whereas datasets are often very large (e.g., several tens or even hundreds of thousands of interacting operations).

On this topic, the specific frontier between learning methods of AI and statistics is relatively blurred, lying most on the relative prominence of the computational performance of algorithms versus mathematics, probability and rigorous statistical inference. While machine learning methods are more empirical, focused on improving their predictive performance, statistics is more concerned with the quantification and modelling of uncertainties and errors [ 79 , 80 ]. In the last decade, both communities have started to communicate and to mix together. Methods have cross-fertilised, giving birth to statistical models using synthetic variables generated by AI methods, or AI algorithms optimising statistical measures of likelihood or quality. New research areas, such as Probabilistic Machine Learning, have emerged at the interface between the two domains [ 1 , 80 , 81 ]. Meanwhile, machine learning and statistics have kept their specific interests and complementarity; machine learning methods are especially well-suited to processing non-standard data types (e.g., images, sounds), while statistics can draw inference and model processes for which only few data are available, or where the quantities of interest are extreme events.

5 Targeted interventions, model of human decisions, and support of AH decisions

5.1 choosing among alternatives.

A challenge for animal health managers is to identify the most relevant combinations of control measures according to local (e.g., farm characteristics, production objectives) and territorial (e.g., available resources, farm location, management priorities) specificities. They have to anticipate the effects of health, environmental and regulatory changes, and deliver quality health advice. The question also arises of how to promote innovation in AH, such as to anticipate the required characteristics of candidate molecules in vaccine strategies or drug delivery [ 82 , 83 ], or to assess the competitive advantage of new strategies (e.g., genomic selection of resistant animals, new vaccines) over more conventional ones. Private (e.g., farmers, farmers’ advisors) and collective managers (e.g., farmer groups, public authorities) need support decision tools to better target public incentives, identify investments to be favoured by farmers [ 46 ] and target the measures as effectively as possible: who to target (which farms, which animals)?; with which appropriate measure(s)?; when and for how long? These questions become essential to reasoning about input usage (e.g., antimicrobials, pesticides, biocides) within the framework of the agro-ecological transition.

The use of mechanistic modelling is a solution to assess, compare and prioritise ex ante a wide range of options (Figure  4 ; [ 84 ]). However, most of the available models do not explicitly integrate human decision-making, while control decisions are often made by farmers (e.g., unregulated diseases), with sometimes large-scale health and decision-making consequences (e.g., pathogen spread, dissemination of information and rumours, area of influence). Recent work aims to integrate humans and their decisions by mobilising optimal control and adaptive strategies from AI [ 7 , 85 ] or health economics methods [ 86 , 87 ]. A challenge is to propose clear and context-adapted control policies [ 88 ]. Such research is just starting in AH [ 46 ] and must be extended as part of the development of agro-ecology, facing current societal demand for product quality and respect for ecosystems and their biodiversity on one side, animal well-being and ethics on the other side, and more generally international health security.

Identifying relevant strategies to control bovine paratuberculosis at a regional scale (adapted from [ 76 ] ) . Classically, identifying relevant strategies means defining them a priori and comparing them, e.g., by modelling. Only a small number of alternatives can be considered. If all alternatives are considered as in the figure, it results in a multitude of scenarios whose analysis becomes challenging. Here, each point corresponds to the epidemiological situation after 9 years of pathogen spread over a network of 12 500 dairy cattle herds for a given strategy (asterisk: no control). Initially, 10% of the animals are infected on average in 30% of the herds. The blue dots correspond to the most favourable strategies. Mobilizing AI approaches in such a framework, especially optimization under constraints, would facilitate the identification of relevant strategies by exploring the space of possibilities in a more targeted manner.

5.2 Accounting for expectations and fears of animal health managers

Animal health managers should have access to model predictions in a time frame compatible with management needs, which is problematic in the face of unpredictable emerging events (e.g., new epidemiological systems, transmission pathways, trade patterns, control measures). Developing a library of models included in a common framework would strengthen the responsiveness of modellers in animal epidemiology. Relevant models would be developed more quickly and would gain accuracy from real-time modelling as epidemics progress [ 89 , 90 ]. However, if this makes move more quickly from concepts (knowledge and assumptions) to simulations and support decision tools, a gain in performance is still required to perform analyses at a very large scale. The automatic generation of high-performance computer code could be a relevant solution, which however remains a crucial methodological lock to be addressed in AI. In addition, it is often required to perform a very large number of calculations or to analyse very large datasets, which call for a rational use of computing resources. Software transferred to health managers sometimes require the use of private cloud resources (i.e., it does not run on simple individual computers), highlighting the trade-offs between simulation cost, service continuity (e.g., failure management) and time required to obtain simulation results [ 91 ]. These questions are currently related to computer science research, and collaborations are desirable between these researchers and those from AH.

Managers also wish to rely on accurate predictions from realistic representations of the biological systems. Before being used, model behaviour should be analysed, which raises the questions of exploring the space of uncertainties and data, and of optimization under constraints. This often requires intensive simulations, which would benefit from optimization algorithms to explore more efficiently the space of possibilities. In turn, this would allow, for example, the automatic identification of how to achieve a targeted objective (e.g., reducing the prevalence of a disease below an acceptable threshold) while being constrained in resource allocation. While this issue finds solutions in modern statistics for relatively simple systems, it represents a science front for complex systems (e.g., large scale, multi-host/multi-pathogen systems) that are becoming the norm. In addition, optimization goals specific to AH may generate ad hoc methodological needs [ 92 ]. The needs in abstraction and analysis capacity are massive and could benefit from complementarities between AI (e.g., reasoned exploration, intelligent use of computer resources, optimized calculations) and statistics to extract as much information as possible from the data: (1) explore, analyse, predict; (2) infer processes and emergent properties. Methodological developments are still required and would benefit many health issues, particularly in relation to the currently evolving concepts of reservoir-host, edge-host and species barrier [ 93 ]. Furthermore, methodological developments and dissemination of existing methods should be reinforced.

Finally, three barriers have been identified to the development of support decision tools for health managers, related to the societal issue of the acceptability of AI sensu lato, as a major factor of progress. First, ethical issues, which are obvious when it comes to human health, are just as important to consider in AH. Which AI-based tools do we want for modern animal husbandries and trades, and for which objectives? Are these tools not likely to lead to discrimination against farms according to their health status, even when this status cannot be managed by the farmer alone? Second, in AH too, there is a fear that AI-tools may replace human expertise. However, automating does not mean replacing human, his expertise and decision [ 94 ], but rather supporting his capacities for abstraction and analysis, accelerating the global process, making predictions more reliable, guiding complementary research. Nevertheless, a significant development of computer resources and equipment is not without impacting the environment in terms of carbon footprint (e.g., energy-intensive servers, recycling of sensors), which must also be accounted for. Third, the very high complexity of analysing results and acculturating end-users with knowledge issued from academic research, particularly AI, is an obstacle to the appropriation of AI-tools by their users. This may lead to the preference for simpler and more easily accessible methods. However, the latter may not always be the most relevant or reliable. Citizen science projects, also known as community participation in human epidemiology, enable AH to co-design and co-construct the AI-tools of tomorrow with their end-users [ 95 ], to better meet their expectations and needs, and to increase their confidence in the predictions of sometimes obscure research models, especially when they are hard to read (e.g., lines of code). Similarly, these AI-tools could be developed together with public decision-makers, livestock farmers, agro-food industries and sectoral trade unions. Co-construction gives time to explain the science behind the tools and makes it more transparent and useful. This citizen participation, which is nowadays supported in many countries, guarantees decisions more in line with citizens’ expectations and corresponds to a general trend towards structured decision-making. AI must contribute to this democratisation of aid in public decision-making in AH.

6 Barriers to the development of research at the AI/AH interface

Research conducted at the interface between AI and AH requires strong interactions between biological disciplines (e.g., infectiology, immunology, clinical sciences, genetics, ecology, evolution, epidemiology, animal and veterinary sciences) and more theoretical disciplines (e.g., modelling, statistics, computer science), sometimes together with sociology and economics. Conducting research at this interface requires strengthening the few teams already positioned in Western Europe, but also bringing together teams working around the concepts of One Health, Ecohealth and Planetary Health to benefit from recent achievements in infectious disease ecology and modelling, plant health and environmental health [ 96 ]. This work must be based on a wide range of methodological skills (e.g., learning methods, data mining, information systems, knowledge representation, multi-agent systems, problem solving, metamodeling, optimisation, simulation architecture, model reduction, decision models). The need for research, training and support are crucial issues at national, European and international levels. Also, a facilitated and trusted connection is required between academics, technical institutes, and private partners, who are often the holders or collectors of data of interest to solve AH research questions through AI approaches. The construction of better inter-sectoral communication and coordination must be done at supra-institutional level, as this theme seems hyper-competitive and as some current divisions still go against information and data sharing.

An acculturation of researchers to AI, its methods and potential developments, but also its limitations, must be proposed to meet the challenges of 21 st century agriculture. Indeed, there are obstacles to conducting research on this scientific front. Establishing the new collaborations required between teams conducting methodological work and teams in the fields of application remains difficult given the low number of academic staff on these issues, their very high current mobilisation and their low availability to collaborate on new subjects, as well as the difficulty of understanding and mastering these methods. There is a need for watching and training on AI methods available or under development, new softwares/packages, and their applicability. To develop key collaborations and establish a strategic positioning, an interconnection can also be made via transversal teams which appears as a preferential path. Solutions must also be found to encourage method percolation in the community and the development of scientific and engineering skills.

Finally, AI methods, such as classification, machine learning, data mining, and the innovations in AH to which these methods can lead are rarely discussed in veterinary high school education, whereas these students represent the future professionals of AH [ 96 ]. Similarly, there is a quasi-absence of sentinel networks of veterinarians, even if it is developing, although AH questions can arise on a large and collective scale. The scientific community would also benefit from further increasing its skills and experience in the valuation, transfer and protection of intellectual property on these AI methods and associated outcomes.

7 Levers to create a fruitful AI/AH interface

7.1 data sharing and protection.

No innovation at the interface between AI and AH is possible without strong support for the organisation of data storage, management, analysis, calculation, and restitution. The major risk is that demands for AI developments inflate without being supported by available human resources. In addition, an expertise in law, jurisdiction and ethics is required with regard to the acquisition, holding, use and protection of data in AH. This question must be considered at least at the inter-institutional/national level, and could benefit from a similar thinking already engaged in human health. The issue is to be able to support any change with regard to data traceability to their ownership, whether being from public or private domains.

New data are rich and must be valued as much as possible, not by each owner separately, but through data sharing and the mobilisation of multi-disciplinary skills to analyse such heterogeneous and complex data. Hence, data interoperability skills are required and must be developed. Models for making federated data sustainable over decades are required [ 97 ]. In addition, further encouraging the publication of data papers as valuable research products can help to develop the necessary culture of sharing, documentation and metadata.

Finally, to be able to launch ambitious experiments with AI methods on real data, it is necessary to (1) remove unauthorised access to data by negotiating with owners at large scale; (2) analyse and understand the related effect on methodological developments; and (3) if necessary, extend such initiatives to other areas, at national scale, or even across European countries.

7.2 Attract the necessary skills

An undeniable barrier to conduct such research comes from human resources, in particular the current insufficient capacity of supervision by permanent scientists. Collaborations are a solution to attract new skills. However, initiating collaborations at the AH/AI interface becomes very complicated because the qualified teams are already overwhelmed. Skill development at this interface must be supported, the cross-fertilisation of disciplines being essential. A watch on methods must also be carried out, accompanied by explanations for application fields, to train researchers and engineers. Financial incentives for scientist internships in specialised laboratories would increase skill capitalisation in advanced methods, while facilitating future national or international collaborations. In a context of limited resources as observed in many countries nowadays (e.g., new opened positions in national institutions) and limited experts pool (e.g., skills), facilitating post-doctoral fellows and continuing education of researchers becomes crucial. Finally, to consolidate the pool of future researchers in AH, promoting basic AI education in initial training of AH researchers, engineers and veterinarians is paramount.

More specifically concerning current research in immunology, cell biology and infectiology, the contribution of AI has been more widely considered in human health, which could feed a similar reflection in AH as locks and advances are not very specific. Before embarking on the fronts of science (e.g., emerging epigenomics and metabolomics in AH), a few persons from these biological disciplines should acculturate into AI, or even acquire autonomy in the use of methods [ 98 ], which internationally tends to be the trend [ 63 ]. This can be done through the sharing of experiences and basic training on existing methods, their advantages and limitations compared to other methods coming from statistics and mathematical modelling.

7.3 Encourage the development of AH/AI projects

Projects at the AH/AI interface, like any interdisciplinary project, must mobilise teams from both groups of disciplines and allow everyone to progress in their own discipline. However, identifying the issues shared between the most relevant disciplines requires a good acculturation of the disciplines between them, as well as an otherness aimed at better understanding each other [ 95 ], which is not yet the case at the AH/AI interface.

In terms of funding, European project calls offer interesting opportunities, but a significant imbalance persists between the ability to generate data and analyse complex issues, and the availability of human resources and skills to address such issues through AI methods or other modern methods in statistics, mathematics and computer science. The major international foundations (e.g., Bill and Melinda Gates) can also be mobilised on emerging infectious diseases at the animal/human interface (e.g., characterisation of weak signals, phenologies, emergence precursors), with a more significant methodological value. However, risk-taking is rarely allowed by funding agencies, although it is crucial to initiate interdisciplinary work. Dedicated incentive funding would support projects in their initial phase and make larger projects emerge after consolidation of the necessary disciplinary interactions.

Finally, these projects are generally based on the use of significant computing resources. Thus, research institutes and private partners should contribute in a financial or material way to the shared development of digital infrastructures, data centres, supercomputing centres on a national scale, as well as support recognised open-source software platforms on which a large part of the research conducted is based (e.g., Python, R ).

7.4 Promoting innovation and public–private partnership

Encouraging public–private partnership would promote a leverage effect on public funding and would make it possible to place AI research and development on a long-term basis in AH. Mapping the highly changing landscape of companies in the AH/AI sector, whether international structures or start-ups, would provide a better understanding of the possible interactions. Similarly, mapping academic deliverables produced at this interface would increase their visibility and highlight their potential for valorisation or transfer. Finally, considering the production of documented algorithms as scientific deliverables, along with publications, would help support this more operational research. More broadly, it would be advisable to initiate a communication and education/acculturation policy around AI and its development in AH (e.g., links with the society, farmers, agricultural unions, public services).

8 Conclusion

The use of AI methods (e.g., machine learning, expert systems, analytical technologies) converges today with the collecting of massive and complex data, and allows these fields to develop rapidly. However, it is essential not to perceive massive data and AI as the same trend, because the accumulation of data does not always lead to an improvement in knowledge. Nevertheless, the more data are numerous and representative of working concepts and hypotheses, the more important results can be obtained from AI applications. The underlying ethical, deontological and legal aspects of data ownership, storage, management, sharing and interoperability also require that a reflection be undertaken nationally and internationally in AH to better manage these data of multi-sectoral origin and their various uses. Moreover, while the effort to acquire such data is impressive, the development of AI skills within the AH community remains limited in relation to the needs. Opportunities for collaborations with AI teams are limited because these teams are already in high demand. To ensure that AH researchers are well aware of the opportunities offered by AI, but also of the limits and constraints of AI approaches, a training effort must be provided and generalized. Finally, the current boom in AI now makes it possible to integrate the knowledge and points of view of the many players in the field of animal health and welfare further upstream. However, this requires that AI and its actors accept to deal with the specificity and complexity of AH, which is not a simple library of knowledge that can be digitised to search for sequences or informative signals.

Hosny A, Parmar C, Quackenbush J, Schwartz LH, Aerts HJWL (2018) Artificial intelligence in radiology. Nat Rev Cancer 18:500–510

Article   CAS   Google Scholar  

Karczewski KJ, Snyder MP (2018) Integrative omics for health and disease. Nat Rev Genet 19:299–310

Murphy KP (2012) Machine learning: a probabilistic perspective. In: Adaptive computation and machine learning series. MIT Press, USA

Zhang W, Chien J, Yong J, Kuang R (2017) Network-based machine learning and graph theory algorithms for precision oncology. NPJ Precis Oncol 1:25

Article   Google Scholar  

Saria S, Butte A, Sheikh A (2018) Better medicine through machine learning: what’s real, and what’s artificial? PLoS Med 15:e1002721. https://doi.org/10.1371/journal.pmed.1002721

Article   PubMed   PubMed Central   Google Scholar  

Bedi G, Carrillo F, Cecchi GA, Fernández Slezak D, Sigman M, Mota NB, Ribeiro S, Javitt DC, Copelli M, Corcoran CM (2015) Automated analysis of free speech predicts psychosis onset in high-risk youths. Schizophrenia 1:15030

Maclachlan MJ, Springborn MR, Fackler PL (2017) Learning about a moving target in resource management: optimal Bayesian disease control. Am J Agri Econ 99:140–162. https://doi.org/10.1093/ajae/aaw033

Lynn LA (2019) Artificial intelligence systems for complex decision-making in acute care medicine: a review. Patient Saf Surg 13:6. https://doi.org/10.1186/s13037-019-0188-2

Pinaire J, Azé J, Bringay S, Landais P (2017) Patient healthcare trajectory. An essential monitoring tool: a systematic review. Health Inf Sci Syst 5:1

Vrakas D, Vlahavas IPL (2008) Artificial intelligence for advanced problem solving techniques. Information Science Reference, Hershey, PA, pp. 369

Shakshuki E, Reid M (2015) Multi-agent system applications in healthcare: current technology and future roadmap. Proc Comput Sci 52:252–261. https://doi.org/10.1016/j.procs.2015.05.071

Roche B, Guégan JF, Bousquet F (2008) Multi-agent systems in epidemiology: a first step for computational biology in the study of vector-borne disease transmission. BMC Bioinform 9:435. https://doi.org/10.1186/1471-2105-9-435

Picault S, Huang Y-L, Sicard V, Ezanno P (2017) Enhancing Sustainability of Complex Epidemiological Models through a Generic Multilevel Agent-based Approach. In: Proceedings of the 26 th International Joint Conference on Artificial Intelligence (IJCAI). pp. 374–380, AAAI. https://doi.org/10.24963/ijcai.2017/53p

Russell S, Norvig P (2010) Artificial intelligence a modern approach. 3 rd edn. Upper Saddle River, New Jersey, pp. 1132

Google Scholar  

Ducrot C, Bed’Hom B, Béringue V, Coulon JB, Fourichon C, Guérin JL, Krebs S, Rainard P, Schwartz-Cornil I, Torny D, Vayssier-Taussat M, Zientara S, Zundel E, Pineau T (2011) Issues and special features of animal health research. Vet Res 42:96

Clark B, Stewart GB, Panzone LA, Kyriazakis I, Frewer LJ (2016) A systematic review of public attitudes, perceptions and behaviours towards production diseases associated with farm animal welfare. J Agric Environ Ethics 29:455–478. https://doi.org/10.1007/s10806-016-9615-x

Miguel E, Grosbois V, Caron A, Pople D, Roche B, Donnelly C (2020) A systemic approach to assess the potential and risks of wildlife culling for infectious disease control. Commun Biol 3:353. https://doi.org/10.1038/s42003-020-1032-z

Hur B, Hardefeldt LY, Verspoor K, Baldwin T, Gilkerson JR (2019) Using natural language processing and VetCompass to understand antimicrobial usage patterns in Australia. Aust Vet J 97:298–300. https://doi.org/10.1111/avj.12836

Article   CAS   PubMed   Google Scholar  

Behmann J, Hendriksen K, Mueller U, Buescher W, Pluemer L (2016) Support vector machine and duration-aware conditional random field for identification of spatio-temporal activity patterns by combined indoor positioning and heart rate sensors. Geoinformatica 20:693–714. https://doi.org/10.1007/s10707-016-0260-3

Suravajhala P, Kogelman LJA, Kadarmideen HN (2016) Multi-omic data integration and analysis using systems genomics approaches: methods and applications in animal production, health and welfare. Genet Sel Evol 48:38. https://doi.org/10.1186/s12711-016-0217-x

Article   CAS   PubMed   PubMed Central   Google Scholar  

Goldansaz SA, Guo AC, Sajed T, Steele MA, Plastow GS, Wishart DS (2017) Livestock metabolomics and the livestock metabolome: a systematic review. PLoS One 12:e0177675. https://doi.org/10.1371/journal.pone.0177675

Anvar SY, Tucker A, Vinciotti V, Venema A, van Ommen GJ, van der Maarel SM, Raz V, ’t Hoen PA (2011) Interspecies translation of disease networks increases robustness and predictive accuracy. PLoS Comput Biol 7:e1002258. https://doi.org/10.1371/annotation/fc0b4192-6427-4fb3-b347-c66651adf855

Morand S, Guégan J-F, Laurans Y (2020) From One Health to Ecohealth, mapping the incomplete integration of human, animal and environmental health. Iddri, Issue Brief No. 04/20

Ezenwa VO, Prieur-Richard A-H, Roche B, Bailly X, Becquart P, Garcia-Peña GE, Hosseini PR, Keesing F, Rizzoli A, Suzán GA, Vignuzzi M, Vittecoq M, Mills JN, Guégan J-F (2015) Interdisciplinarity and infectious diseases: an Ebola case study. PLoS Pathog 11:e1004992. https://doi.org/10.1371/journal.ppat.1004992

Van Boeckel TP, Takahashi S, Liao Q, Xing W, Lai S, Hsiao V, Liu F, Zheng Y, Chang Z, Yuan C, Metcalf CJE, Yu H, Grenfell BT (2016) Hand, foot, and mouth disease in China: critical community size and spatial vaccination strategies. Sci Rep 6:25248. https://doi.org/10.1038/srep25248

Holmstrom LK, Beckham TR (2017) Technologies for capturing and analysing animal health data in near real time. Rev Sci Tech 36:525–538

Neethirajan S (2017) Recent advances in wearable sensors for animal health management. Sens Biosensing Res 12:15–29

Perez AM, Zeng D, Tseng CJ, Chen H, Whedbee Z, Paton D, Thurmond MC (2009) A web-based system for near real-time surveillance and space-time cluster analysis of foot-and-mouth disease and other animal diseases. Prev Vet Med 91:39–45. https://doi.org/10.1016/j.prevetmed.2009.05.006

Article   PubMed   Google Scholar  

Wilkinson MD, Dumontier M, Aalbersberg IJ, Appleton G, Axton M, Baak A, Blomberg N, Boiten J-W, da Silva Santos LB, Bourne PE, Bouwman J, Brookes AJ, Clark T, Crosas M, Dillo I, Dumon O, Edmunds S, Evelo CT, Finkers R, Gonzalez-Beltran A, Gray AJG, Groth P, Goble C, Grethe JS, Heringa J, ’t Hoen PAC, Hooft R, Kuhn T, Kok R, Kok J, et al (2016) The FAIR guiding principles for scientific data management and stewardship. Sci Data 3:160018. https://doi.org/10.1038/sdata.2016.18

Binot A, Duboz R, Promburom P, Phimpraphai W, Cappelle J, Lajaunie C, Goutard FL, Pinyopummintr T, Figuié M, Roger FL (2015) A framework to promote collective action within the One Health community of practice: using participatory modelling to enable interdisciplinary, cross-sectoral and multi-level integration. One Health 1:44–48. https://doi.org/10.1016/j.onehlt.2015.09.001

Robert CP (2014) Bayesian computational tools. Annu Rev Stat Appl 1:153–177. https://doi.org/10.1146/annurev-statistics-022513-115543

Dunson DB (2001) Commentary: practical advantages of Bayesian analysis of epidemiologic data. Am J Epidemiol 153:1222–1226. https://doi.org/10.1093/aje/153.12.1222

Uusitalo L (2007) Advantages and challenges of Bayesian networks in environmental modelling. Ecol Model 203:312–318. https://doi.org/10.1016/j.ecolmodel.2006.11.033

Fokoué E (2019) On the ubiquity of the Bayesian paradigm in statistical machine learning and data science. Math Appl 8:189–209. https://doi.org/10.13164/ma.2019.12

Bailly X (2017) Hidden Markov phylogenetic models offer an interesting perspective to identify “high risk lineages” of environmental pathogens. Infect Genet Evol 55:45–47. https://doi.org/10.1016/j.meegid.2017.08.007

Babayan SA, Orton RJ, Streicker DG (2018) Predicting reservoir hosts and arthropod vectors from evolutionary signatures in RNA virus genomes. Science 362:577–580. https://doi.org/10.1126/science.aap9072

Wardeh M, Sharkey KJ, Baylis M (2020) Integration of shared-pathogen networks and machine learning reveals the key aspects of zoonoses and predicts mammalian reservoirs. Proc Biol Sci 287:20192882. https://doi.org/10.1098/rspb.2019.2882

Li J, Zhang S, Li B, Hu Y, Kang X-P, Wu X-Y, Huang M-T, Li Y-C, Zhao Z-P, Qin C-F, Jiang T (2020) Machine learning methods for predicting human-adaptive influenza A viruses based on viral nucleotide compositions. Mol Biol Evol 37:1224–1236. https://doi.org/10.1093/molbev/msz276

Peters DPC, McVey DS, Elias EH, Pelzel-McCluskey AM, Derner JD, Burruss ND, Schrader TS, Yao J, Pauszek SJ, Lombard J, Rodriguez LL (2020) Big data-model integration and AI for vector-borne disease prediction. Ecosphere 11:e03157. https://doi.org/10.1002/ecs2.3157

Lek S, Guégan J-F (2000) Artificial neuronal networks. In: Application to ecology and evolution. Springer, Berlin. https://doi.org/10.1016/j.it.2016.11.006

Go N, Touzeau S, Islam Z, Belloc C, Doeschl-Wilson A (2019) How to prevent viremia rebound? Evidence from a PRRSv data-supported model of immune response. BMC Syst Biol 13:15

Ferrer Savall J, Bidot C, Leblanc-Maridor M, Belloc C, Touzeau S (2016) Modelling Salmonella transmission among pigs from farm to slaughterhouse: interplay between management variability and epidemiological uncertainty. Intern J Food Microbiol 229:33–43. https://doi.org/10.1016/j.ijfoodmicro.2016.03.020

Widgren S, Engblom S, Bauer P, Frössling J, Emanuelson U, Lindberg A (2016) Data-driven network modelling of disease transmission using complete population movement data: spread of VTEC O157 in Swedish cattle. Vet Res 47:81

Qi L, Beaunée G, Arnoux S, Dutta BL, Joly A, Vergu E, Ezanno P (2019) Neighbourhood contacts and trade movements drive the regional spread of bovine viral diarrhoea virus (BVDV). Vet Res 50:30. https://doi.org/10.1186/s13567-019-0647-x

Buhnerkempe MG, Tildesley MJ, Lindström T, Grear DA, Portacci K, Miller RS, Lombard JE, Werkman M, Keeling MJ, Wennergren U, Webb CT (2014) The impact of movements and animal density on continental scale cattle disease outbreaks in the United States. PLoS One 9:e91724. https://doi.org/10.1371/journal.pone.0091724

Ezanno P, Andraud M, Beaunée G, Hoch T, Krebs S, Rault A, Touzeau S, Vergu E, Widgren S (2020) How mechanistic modelling supports decision 1 making for the control of enzootic infectious diseases. Epidemics 32:100398

Garira W (2018) A primer on multiscale modelling of infectious disease systems. Infect Dis Model 3:176–191. https://doi.org/10.1016/j.idm.2018.09.005

Traoré M, Zacharewicz G, Duboz R, Zeigler B (2018) Modeling and simulation framework for value-based healthcare systems. Simulation 95:481–497. https://doi.org/10.1177/0037549718776765

Childs LM, El Moustaid F, Gajewski Z, Kadelka S, Nikin-Beers R, Smith JW Jr, Walker M, Johnson LR (2019) Multi-scale models and data for infectious diseases: a systematic review. PeerJ Preprints 7:e27485v1. https://doi.org/10.7287/peerj.preprints.27485v1

Sandve GK, Nekrutenko A, Taylor J, Hovig E (2013) Ten simple rules for reproducible computational research. PLoS Comput Biol 9:e1003285. https://doi.org/10.1371/journal.pcbi.1003285

Leek JT, Peng RD (2015) Opinion: reproducible research can still be wrong: adopting a prevention approach. Proc Natl Acad Sci USA 112:1645–1646. https://doi.org/10.1073/pnas.1421412111

Widgren S, Bauer P, Eriksson R, Engblom S (2016) SimInf: an R package for data-driven stochastic disease spread simulations. ArXiv160501421 Q-Bio Stat. http://arxiv.org/abs/1605.01421

O’Hare A, Lycett SJ, Doherty TM, Salvador LC, Kao RR (2016) Broadwick: a framework for computational epidemiology. BMC Bioinform 17:65. https://doi.org/10.1186/s12859-016-0903-2

Bui TMA, Stinckwich S, Ziane M, Roche B, Ho TV (2015) KENDRICK: a domain specific language and platform for mathematical epidemiological modelling. In: proc. IEEE RIVF International Conference on Computing and Communication Technologies, Research, Innovation, and Vision for the Future. pp. 132–7. https://doi.org/10.1109/RIVF.2015.7049888

Mathieu P, Morvan G, Picault S (2018) Multi-level agent-based simulations: four design patterns. Simul Model Pract Theory 83:51–64. https://doi.org/10.1016/j.simpat.2017.12.015

Picault S, Huang Y-L, Sicard V, Arnoux S, Beaunée G, Ezanno P (2019) EMULSION: transparent and flexible multiscale stochastic models in human, animal and plant epidemiology. PLoS Comput Biol 15:e1007342. https://doi.org/10.1371/journal.pcbi.1007342

Sebag AS, Plancade S, Raulet-Tomkiewicz C, Barouki R, Vert J-P, Walter T (2015) Inferring an ontology of single cell motions from high-throughput microscopy data. In: Proc. IEEE International Symposium on Biomedical Imaging, Apr. 2015, New-York, USA, pp. 160–163. https://doi.org/10.1109/ISBI.2015.7163840

Tardivel P, Canlet C, Lefort G, Tremblay-Franco M, Debrauwer L, Concordet D, Servien R (2017) ASICS: an automatic method for identification and quantification of metabolites in complex 1D 1H NMR spectra. Metabolomics 13:109

Dórea FC, Muckle CA, Kelton D, McClure JT, McEwen BJ, McNab WB, Sanchez J, Revie CW (2013) Exploratory analysis of methods for automated classification of laboratory test orders into syndromic groups in veterinary medicine. PLoS One 8:e57334. https://doi.org/10.1371/journal.pone.0057334

Gandia P, Jaudet C, Chatelut E, Concordet D (2017) Population pharmacokinetics of tracers: a new tool for medical imaging? Clin Pharmacokinet 56:101–106

Chyba M, Coron J-M, Mileyko Y, Rezaei H (2016) Optimization of prion assemblies fragmentation. In: Proc. IEEE Conference on Decision and Control (CDC), Las Vegas, USA, 6

Nayeri S, Sargolzaei M, Tulpan D (2019) A review of traditional and machine learning methods applied to animal breeding. Anim Health Res Rev 20:31–46. https://doi.org/10.1017/S1466252319000148

Bassaganya-Riera J, Hontecillas R (2016) Introduction to computational immunology. In: Bassaganya-Riera J (ed) Computational immunology: models and tools. pp. 1–8

Liakos KG, Busato P, Moshou D, Pearson S, Bochtis D (2018) Machine learning in agriculture: a review. Sensors 18:2674. https://doi.org/10.3390/s18082674

Jones-Diette JS, Dean RS, Cobb M, Brennan ML (2019) Validation of text-mining and content analysis techniques using data collected from veterinary practice management software systems in the UK. Prev Vet Med 167:61–67. https://doi.org/10.1016/j.prevetmed.2019.02.015

Morota G, Ventura RV, Silva FF, Koyama M, Fernando SC (2018) Big data analytics and precision animal agriculture symposium: machine learning and data mining advance predictive big data analysis in precision animal agriculture. J Anim Sci 96:1540–1550. https://doi.org/10.1093/jas/sky014

Picault S, Ezanno P, Assié S (2019) Combining early hyperthermia detection with metaphylaxis for reducing antibiotics usage in newly received beef bulls at fattening operations: a simulation-based approach. In: Society of veterinary epidemiology and preventive medicine (SVEPM), pp. 13. Utrecht, The Netherland, 27-30/3/2019

Esener N, Green MJ, Emes RD, Jowett B, Davies PL, Bradley AJ, Dottorini T (2018) Discrimination of contagious and environmental strains of Streptococcus uberis in dairy herds by means of mass spectrometry and machine-learning. Sci Rep 8:17517. https://doi.org/10.1038/s41598-018-35867-6

Hepworth PJ, Nefedov AV, Muchnik IB, Morgan KL (2012) Broiler chickens can benefit from machine learning: support vector machine analysis of observational epidemiological data. J R Soc Interface 9:1934–1942. https://doi.org/10.1098/rsif.2011.0852

Marceau A, Madouasse A, Lehébel A, van Schaik G, Veldhuis A, Van der Stede Y, Fourichon C (2014) Can routinely recorded reproductive events be used as indicators of disease emergence in dairy cattle? An evaluation of 5 indicators during the emergence of bluetongue virus in France in 2007 and 2008. J Dairy Sci 97:6135–6150. https://doi.org/10.3168/jds.2013-7346

Fountain-Jones NM, Machado G, Carver S, Packer C, Recamonde-Mendoza M, Craft ME (2019) How to make more from exposure data? An integrated machine learning pipeline to predict pathogen exposure. J Anim Ecol 88:1447–1461. https://doi.org/10.1111/1365-2656.13076

Charras-Garrido M, Azizi L, Forbes F, Doyle S, Peyrard N, Abrial D (2013) On the difficulty to delimit disease risk hot spots. Int J Appl Earth Obs 22:99–105. https://doi.org/10.1016/j.jag.2012.04.005

Forbes F, Charras-Garrido M, Azizi L, Doyle S, Abrial D (2013) Spatial risk mapping for rare disease with hidden Markov fields and variational EM. Annals Appl Stat 7:1192–1216

Arsevska E, Valentin S, Rabatel J, de Goër de Hervé J, Falala S, Lancelot R, Roche M (2018) Web monitoring of emerging animal infectious diseases integrated in the French Animal Health Epidemic Intelligence System. PLoS One 13:0199960. https://doi.org/10.1371/journal.pone.0199960

Küker S, Faverjon C, Furrer L, Berezowski J, Posthaus H, Rinaldi F, Vial F (2018) The value of necropsy reports for animal health surveillance. BMC Vet Res 14:191. https://doi.org/10.1186/s12917-018-1505-1

Bollig N, Clarke L, Elsmo E, Craven M (2020) Machine learning for syndromic surveillance using veterinary necropsy reports. PLoS One 15:e0228105. https://doi.org/10.1371/journal.pone.0228105

Hoscheit P, Geeraert S, Beaunée G, Monod H, Gilligan CAG, Filipe J, Vergu E, Moslonka-Lefebvre M (2016) Dynamical network models for cattle trade: towards economy-based epidemic risk assessment. J Complex Netw 5:604–624. https://doi.org/10.1093/comnet/cnw026

Moslonka-Lefebvre M, Gilligan CA, Monod H, Belloc C, Ezanno P, Filipe JAN, Vergu E (2016) Market analyses of livestock trade networks to inform the prevention of joint economic and epidemiological risks. J R Soc Interface 13:20151099. https://doi.org/10.1098/rsif.2015.1099

Efron B (2020) Prediction, estimation, and attribution. J Am Stat Ass 115:636–655. https://doi.org/10.1080/01621459.2020.1762613

Ghahramani Z (2012) Probabilistic modelling, machine learning, and the information revolution. MIT Computer Science and Artificial Intelligence Lab, http://mlg.eng.cam.ac.uk/zoubin/talks/mit12csail.pdf , Accessed 17 Oct 2019

Hastie T, Tibshirani R, Friedman JH (2009) The elements of statistical learning: data mining, inference, and prediction. 2 nd edn. Springer Series in Statistics. Springer

Goodswen SJ, Kennedy PJ, Ellis JT (2017) On the application of reverse vaccinology to parasitic diseases: a perspective on feature selection and ranking of vaccine candidates. Int J Parasitol 47:779–790. https://doi.org/10.1016/j.ijpara.2017.08.004

Schneider G (2019) Mind and machine in drug design. Nat Mach Intell 1:128–130. https://doi.org/10.1038/s42256-019-0030-7

Beaunée G, Vergu E, Joly A, Ezanno P (2017) Controlling bovine paratuberculosis at a regional scale: towards a decision modeling tool. J Theor Biol 435:157–183. https://doi.org/10.1016/j.jtbi.2017.09.012

Viet A-F, Krebs S, Rat-Aspert O, Jeanpierre L, Belloc C, Ezanno P (2018) A modelling framework based on MDP to coordinate farmers’ disease control decisions at a regional scale. PLoS One 13:e0197612. https://doi.org/10.1371/journal.pone.0197612

Wang T, Hennessy DA (2015) Strategic interactions among private and public efforts when preventing and stamping out a highly infectious animal disease. Am J Agri Econ 97:435–451. https://doi.org/10.1093/ajae/aau119

Tago D, Hammitt JK, Thomas A, Raboisson D (2016) The impact of farmers’ strategic behavior on the spread of animal infectious diseases. PLoS One 11:e0157450. https://doi.org/10.1371/journal.pone.0157450

Probert WJM, Lakkur S, Fonnesbeck CJ, Shea K, Runge MC, Tildesley MJ, Ferrari MJ (2019) Context matters: using reinforcement learning to develop human-readable, state-dependent outbreak response policies. Phil Trans R Soc B 374:20180277. https://doi.org/10.1098/rstb.2018.0277

Liang R, Lu Y, Qu X, Su Q, Li C, Xia S, Liu Y, Zhang Q, Cao X, Chen Q, Niu B (2020) Prediction for global African swine fever outbreaks based on a combination of random forest algorithms and meteorological data. Transbound Emerg Dis 67:935–946. https://doi.org/10.1111/tbed.13424

Salje H, Tran Kiem C, Lefrancq N, Courtejoie N, Bosetti P, Paireau J, Andronico A, Hozé N, Richet J, Dubost C-L, Le Strat Y, Lessler J, Levy Bruhl D, Fontanet A, Opatowski L, Boelle P-Y, Cauchemez S (2020) Estimating the burden of SARS-CoV-2 in France. Science 369:208–211

Parlavantzas N, Pham LM, Morin C, Arnoux S, Beaunée G, Qi L, Gontier P, Ezanno P (2019) A service-based framework for building and executing epidemic simulation applications in the cloud. Concurr Comp Pract Exper 32:e5554. https://doi.org/10.1002/cpe.5554

Shah N, Malensek M, Shah H, Pallickara S, Pallickara SL (2019) Scalable network analytics for characterization of outbreak influence in voluminous epidemiology datasets. Concurr Comp Pract Exper 31:e4998. https://doi.org/10.1002/cpe.4998

Han BA, Majumdar S, Calmon FP, Glicksberg BS, Horesh R, Kumar A, Perer A, von Marschall EB, Wei D, Mojsilović A, Varshney KR (2019) Confronting data sparsity to identify potential sources of Zika virus spillover infection among primates. Epidemics 27:59–65. https://doi.org/10.1016/j.epidem.2019.01.005

Reddy S, Fox J, Purohit MP (2019) Artificial intelligence-enabled healthcare delivery. J R Soc Med 112:22–28

Duboz R, Echaubard P, Promburom P, Kilvington M, Ross H, Allen W, Ward J, Deffuant G, de Garine-Wichatitsky M, Binot A (2018) Systems thinking in practice: participatory modelling as a foundation for integrated approaches to health. Front Vet Sci 5:303. https://doi.org/10.3389/fvets.2018.00303

Van der Waal K, Morrison RB, Neuhauser C, Vilalta C, Perez AM (2017) Translating big data into smart data for veterinary epidemiology. Front Vet Sci 4:110. https://doi.org/10.3389/fvets.2017.00110

Reichman OJ, Jones MB, Schildhauer MP (2011) Challenges and opportunities of open data in ecology. Science 331:703–705. https://doi.org/10.1126/science.1197962

Schultze JL (2015) Teaching ‘big data’ analysis to young immunologists. Nat Immunol 16:902–905

Download references

Acknowledgements

This work has benefited from interactions with many French researchers (Additional file 1 B) interested in the AI/AH interface, for which we thank them here. We also thank Stéphane Abrioux, Didier Concordet and Human Rezaie for participating to the discussions.

PE is supported by the French Research Agency (project CADENCE: ANR-16-CE32-0007). XB is involved in the project “MOnitoring Outbreak events for Disease surveillance in a data science context” supported by the EU Framework Programme for Research and Innovation H2020 (H2020-SC1-BHC-2018–2019, Grant 874850). JFG is supported by both an “Investissement d’Avenir” managed by the French Research Agency (LABEX CEBA: ANR-10-LABX-25-01) and a US NSF-NIH Ecology of infectious diseases award (NSF#1911457), and is also supported by IRD, INRAE, and Université of Montpellier. The funding bodies had no role in the study design, data analysis and interpretation, and manuscript writing.

Author information

Authors and affiliations.

INRAE, Oniris, BIOEPAR, Nantes, France

Pauline Ezanno, Sébastien Picault & Gaël Beaunée

INRAE, EpiA, Theix, France

Xavier Bailly

ASTRE, Univ Montpellier, CIRAD, INRAE, Montpellier, France

Facundo Muñoz, Raphaël Duboz & Jean-François Guégan

Sorbonne Université, IRD, UMMISCO, Bondy, France

Raphaël Duboz

Université Paris-Saclay, INRAE, Jouy-en-Josas, MaIAGE, France

Hervé Monod

MIVEGEC, IRD, CNRS, Univ Montpellier, Montpellier, France

Jean-François Guégan

Comité National Français Sur Les Changements Globaux, Paris, France

You can also search for this author in PubMed   Google Scholar

Contributions

PE carried out the literature review and analysed the interviews. PE and JFG conducted the interviews, drafted and wrote the manuscript. SP, GB, FM, RD, HM provided complementary views and references in their respective disciplines. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Pauline Ezanno .

Ethics declarations

Competing interests.

The authors declare that they have no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1..

Systematic literature review, interviews, previous publication in French.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Ezanno, P., Picault, S., Beaunée, G. et al. Research perspectives on animal health in the era of artificial intelligence. Vet Res 52 , 40 (2021). https://doi.org/10.1186/s13567-021-00902-4

Download citation

Received : 16 June 2020

Accepted : 20 January 2021

Published : 06 March 2021

DOI : https://doi.org/10.1186/s13567-021-00902-4

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Animal disease
• Artificial intelligence
• Decision support tool

Veterinary Research

ISSN: 1297-9716

• General enquiries: [email protected]

Caterpillars Sense Hungry Wasps’ Electrical Field

Predators’ electricity gives caterpillars an early warning

Gennaro Tomma

Do Cats Really Hate Water?

Not all cats are hydrophobic

Meghan Bartels

$1 for Digital Access

Read all the stories you want.

Scientists Make ‘Cyborg Worms’ with a Brain Guided by AI

AI and tiny worms team up to get to treats

Matthew Hutson

A Dolphin That Has Been Biting People May Just Be Friendly

Dolphin ecologist Tadamichi Morisaka discusses common dolphin behaviors that could explain instances of the animals biting people in Japan

Anna Ikarashi, Nature magazine

Queen’s Brian May Is a Champion for Badgers and Science

Queen guitarist Brian May has spent a decade studying the science of bovine tuberculosis, which can be carried by badgers, and has identified a new method of spread

Elizabeth Gibney, Nature magazine

The End of the Lab Rat?

Replacing research animals with tools that better mimic human biology could improve medicine

Rachel Nuwer

Tardigrade Fossils Reveal When ‘Water Bears’ Became Indestructible

252 million years ago, tardigrades may have escaped extinction using this one weird trick

Mindy Weisberger

Invasive Joro Spiders Keep Their Cool Even when Stressed

Compared with other arachnids, invasive Joro spiders are calm, cool and collected under stress, a new study shows

Stephanie Pappas

Love the Ocean? Thank a Shark

Sharks provide multiple benefits for ocean ecosystems: their declining numbers threaten habitats for baby fish

Michael Heithaus, The Conversation US

Giant, Sparkly Clams Hide the Most Efficient Solar Panels Ever Found

Inside giant, iridescent clams are algal farms that could inspire highly efficient bioreactors

Honeybees Wing-Slap Ants That Try to Invade Their Hive

Japanese honeybees use their wings to slap back ants trying to invade their hive

Komodo Dragons’ Nightmare Iron-Tipped Teeth Are a Reptilian First

Reptile teeth have long been considered simple and cheap because the animals replace them regularly. That isn't so, Komodo dragons show

• Alumni and Giving
• International Programs
• Safety Training and Resources
• Undergraduate
• Transfer Students
• Online Learning
• Student Opportunities
• Faculty and Staff
• Graduate Students
• Faculty Emeriti
• Give Now 

Leading innovation for real-world problems

Cutting-Edge Research

There’s a reason top agencies like the National Institutes of Health and the National Science Foundation pour millions of research dollars into our department each year: Our faculty and students pioneer solutions to the problems facing the animal science industry.

From genetics and genomics to reproductive physiology, our students have opportunities to work with seasoned faculty and apply what they learn in the classroom to real-world issues.

research areas

Animal Well-Being

Research in animal well-being is typically incorporated within other focus areas such as genetics or physiology. Students pursuing this emphasis have investigated everything from the effect of breeding behavior on the reproductive efficiency of swine to the impact of housing conditions on stress-related hormones in captive primates.

Faculty in Animal Well-Being 

Biotechnology

Biotechnology is considered a support discipline and can be included in graduate programs in a variety of areas including physiology, nutrition, genetics/genomics and production management.

Faculty in Biotechnology 

Genetics and Genomics

Students can choose to major in Animal Science with a co-major in Genetics or Functional Genomics. Students pursuing research in genetics and genomics use a variety of animal models including beef cattle, dairy cattle, swine and mice in their thesis projects.

Faculty in Genetics and Genomics 

Our nutrition program focuses on a range of topics, from basic molecular nutrition approaches to studies with direct practical applications in the target species. Students can work with a variety of animal species, including cattle, swine, horses, sheep, goats, mice, companion animals and exotic animals.

Nutrition Research Faculty 

The physiology group conducts activities in two major areas: reproductive physiology and lactational physiology. Student research projects range from basic molecular studies to applied research projects in a variety of animal systems. Students use a variety of animal models including cattle, swine, mice, goats, horses, domestic cats and exotic cats in their research.

Physiology Research Faculty 

Production Agriculture

Students in production management investigate problems that involve applied research within their particular discipline (genetics, nutrition, physiology). Research can include assessment of the usefulness of early pregnancy detection by real-time ultrasonography in swine, the effectiveness of specific feed additives on growth and nutrient utilization of beef steers, or the effect of grazing-based dairy cattle management on the onset of puberty in heifers.

Production Agriculture Research Faculty 

• Skip to Guides Search
• Skip to breadcrumb
• Skip to main content
• Skip to footer
• Skip to chat link
• Report accessibility issues and get help
• Go to Penn Libraries Home
• Go to Franklin catalog

Talking to Science Deniers: Critical Writing Program Fall 2024: Researching the White Paper

• Getting started
• News and Opinion Sites
• Academic Sources
• Grey Literature
• Substantive News Sources
• What to Do When You Are Stuck
• Understanding a citation
• Examples of Quotation
• Examples of Paraphrase
• Chicago Manual of Style: Citing Images
• Researching the Op-Ed
• Researching Prospective Employers
• Resume Resources
• Cover Letter Resources

Research the White Paper

Researching the white paper:.

The process of researching and composing a white paper shares some similarities with the kind of research and writing one does for a high school or college research paper. What’s important for writers of white papers to grasp, however, is how much this genre differs from a research paper.  First, the author of a white paper already recognizes that there is a problem to be solved, a decision to be made, and the job of the author is to provide readers with substantive information to help them make some kind of decision--which may include a decision to do more research because major gaps remain. 

Thus, a white paper author would not “brainstorm” a topic. Instead, the white paper author would get busy figuring out how the problem is defined by those who are experiencing it as a problem. Typically that research begins in popular culture--social media, surveys, interviews, newspapers. Once the author has a handle on how the problem is being defined and experienced, its history and its impact, what people in the trenches believe might be the best or worst ways of addressing it, the author then will turn to academic scholarship as well as “grey” literature (more about that later).  Unlike a school research paper, the author does not set out to argue for or against a particular position, and then devote the majority of effort to finding sources to support the selected position.  Instead, the author sets out in good faith to do as much fact-finding as possible, and thus research is likely to present multiple, conflicting, and overlapping perspectives. When people research out of a genuine desire to understand and solve a problem, they listen to every source that may offer helpful information. They will thus have to do much more analysis, synthesis, and sorting of that information, which will often not fall neatly into a “pro” or “con” camp:  Solution A may, for example, solve one part of the problem but exacerbate another part of the problem. Solution C may sound like what everyone wants, but what if it’s built on a set of data that have been criticized by another reliable source?  And so it goes. 

For example, if you are trying to write a white paper on the opioid crisis, you may focus on the value of  providing free, sterilized needles--which do indeed reduce disease, and also provide an opportunity for the health care provider distributing them to offer addiction treatment to the user. However, the free needles are sometimes discarded on the ground, posing a danger to others; or they may be shared; or they may encourage more drug usage. All of those things can be true at once; a reader will want to know about all of these considerations in order to make an informed decision. That is the challenging job of the white paper author.     
 The research you do for your white paper will require that you identify a specific problem, seek popular culture sources to help define the problem, its history, its significance and impact for people affected by it.  You will then delve into academic and grey literature to learn about the way scholars and others with professional expertise answer these same questions. In this way, you will create creating a layered, complex portrait that provides readers with a substantive exploration useful for deliberating and decision-making. You will also likely need to find or create images, including tables, figures, illustrations or photographs, and you will document all of your sources. 

Nick Okrent

Connect to a Librarian Live Chat or "Ask a Question"

• Librarians staff live chat from 9-5 Monday through Friday . You can also text to chat: 215-543-7674
• You can submit a question 24 hours a day and we aim to respond within 24 hours 
• You can click the "Schedule Appointment" button above in librarian's profile box (to the left), to schedule a consultation with her in person or by video conference.  
• You can also make an appointment with a  Librarian by subject specialization . 
• Connect by email with a subject librarian

Find more easy contacts at our Quick Start Guide

• Next: Getting started >>
• Last Updated: Sep 16, 2024 8:46 AM
• URL: https://guides.library.upenn.edu/c.php?g=1424859