Progress 06/01/23 to 05/31/24

Outputs
Target Audience:Our main target audience isveterinarians as well as stakeholders related with the pig industry and management of animal health such as diagnosticians, farmers, animal health officials, extension specialists and other policy makers working in animal health Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?We are significantly contributing to the training of the next generation of veterinary-data-scientists. During this reporting year, we have involved3 post-doctoralresearchers and 5PhD students with different backgrounds and technical skills from veterinary science, computer science, bioinformatics and biology. We are also developing new curriculumand exposing >100 students per year to these methods and results in the different graduate classes at UC Davis. Specifically, methods and results of this project have been included in the curriculum by Dr. Martinez-Lopez in the course MPM 207 "Applied Epidemiology Problem Solving" and MPM 212 "Concepts and Methods in Infectious Disease Surveillance andControl" at UC Davis. These graduate courses are offered annually to the Graduate Group in Epidemiology (GGE), theMaster of Preventive Veterinary Medicine (MPVM) as well as the Master of Public Health (MPH)and the Graduate Group of Public Health (GGPH) at UC Davis. We meet weekly with our students, and they work collaboratively in the different activities, which has a lot of benefits as they share their experience, programming, analytical skills and domain knowledge in their respective areas. They havegained good real-world experience dealing with dirty data and develop important analytical, technical, communication and communication skills. How have the results been disseminated to communities of interest?We werepresenting our results in relevant Conferences and working to generate high impact Journals (see Products). We also work closely with our industry partners, which include the largest swine veterinary diagnostic laboratory (ISU-VDL), the largest swine breeding company (PIC), the largest veterinary clinic in the US (Pipestone), and some of the top swine producers (Seaboard, Pipestone, Tosh Farms). These partners are helping to co-develop, test, validate and maximize Disease BioPortal use and adoption. We hold frequent meetings with them and receive continuous feedback to improve the capabilities and user experience of the platform. We continued conducting one-to-one meetings and ad-hoc training of the different tools of the Disease BioPortal platform withveterinarians and producers that are interested in using the platform. In this way we are already havingmaximum impact in the swine industry, improving animal health and helping expand pork markets through our research. We are now also working to expand our tools and capabilities to other livestock systems like poultry. Our work was featured in different media such as the AVMA News: Artificial intelligence poised to transform veterinary care. Available at:https://www.avma.org/news/artificial-intelligence-poised-transform-veterinary-care What do you plan to do during the next reporting period to accomplish the goals? We will continue expanding Disease BioPortal capabilities, mostly focusing onimprovingvisualization and analytical tools using AI-assisted technology, including the use of natural language processing. This will accelerate and facilitate the generation and interpretation of dashboards. We are also planning to improve the Disease BioPortal website and generate documentation and video-demos for each component/tool. We will continue participating in conferences, seminars, workshop to maximize the communication of our work and engage new users and finalize all pending publications.

Impacts
What was accomplished under these goals? Objective 1. We continued extending our data lake infrastructure and building new APIs to connect with more diagnostic laboratories and production management softwares. This allow us to integrate and manage new data sources and new data structures quicker. This hasfacilitated the automatic push of data from those data sources to the Disease BioPortal so producers and veterinary clinics can access all their data and generate diagnostic dashboards faster. Objective 2 & 3. We expanded our testing of diferent analytical methods and machine learning algorithms for their use in animal health. In this reporting period wefocused on bioinformatics and machine learning toolsfor better management of bacteria pathogens and antimicrobial resistance.Weincorporateda user-friendly multiple sequence alignment viewer to more rapidly identify sequences tha share patterns and identify those that could be problematic or lack quality and should be removed from the phylogenetic trees. We also expanded the whole genome sequencing (WGS) capabilities within the Disease BioPortal and incorporated a new pipeline for quicker and more comprenhensive analysis of bacterial genomes using Bactopia. As a result, we are able to identify antimicrobial resistance patterns both using phenotypic information (i.e. Minimum inhibitory concentrations) and genotypic data (i.e. whole genome sequences). Users are now able to upload their WGS data directly into the Disease BioPortal and, in less than 24 hours, identify resistance determinants, get the multilocus sequence typing (MLST), generate dashboards with phylogenetic trees and donut charts and quickly assess theAMR changes or trendsin a farm or a group of farms.As a result, veterinarians can use these dashboards to more timely conduct outbreak investigations and select the best therapeutics or vaccines for the circulating variants, contributing tobetter prevention and control of those pathogens as well as enhancing antimicrobial stewardship.

Publications

• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Mart�ne-L�pez B. Keynote speaker. Advancements in the Veterinary Services through Digitalisation (Data Management, Veterinary Information Systems, Big Data, Meta Language, Artificial Intelligence). 31st Conference of the Regional Commission for Europe of the World Organisation for Animal Health (WOAH). Uzbekistan, 30 Sept-4 October 2024
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Mart�ne-L�pez B. Keynote speaker. Epidemiology and AI in veterinary medicine: state of the art and future challenges. Symposium on Artificial Intelligence in Veterinary Medicine (SAVY). April 19-21, 2024
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Rupasinghe R, Clavijo, MJ, Mart�nez-L�pez B. Oral presentation. Antimicrobial resistance surveillance: bridging genomics, microbiology, and epidemiology. Vet Next Conference 2023: Integrating Animal Health and Modern Agriculture, Aug 28-Sep 8, 2023, Davis, CA, USA.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Clavijo MJ et al.. Diagnostics and epidemiology of S. suis infection in pigs. 5th International Workshop on Streptococcus suis. Bangkok, Thailand. June 6-7th 2023.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Robbins R, Harper L, Clavijo MJ. Utilizing next generation sequencing for S. suis vaccine development. In proceedings 2023 ISU James D. McKean Swine Disease Conference, July 28th 2023.
• Type: Conference Papers and Presentations Status: Published Year Published: 2024 Citation: Mugabi, R., Hu, X., Macedo, N. R., Sahin, O., Li, G., Harms, P., Martinez-Lopez, B., & Clavijo, M. J. (2024, February). Genotypic and phenotypic antimicrobial resistance of Streptococcus suis strains from North American swine clinical cases (Poster 28). 55th Annual Meeting of the American Association of Swine Veterinarians, Nashville, TN.

Progress 06/01/22 to 05/31/23

Outputs
Target Audience:Our main target audience is swine veterinarians as well as stakeholders related with the pig industry and pig management such as diagnosticians, farmers, animal health officials, extension specialists and other policy makers working in animal health Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?We are significantly contributing to the training of the next generation of vet-data-scientists, involving directly 3 post-doctoral researchers and 6 PhD students with different backgrounds and technical skills and exposing >100 students per year to these methods and results in the different graduateclasses at UC Davis. We meet weekly with our students, and they work collaboratively in the different activities, which has a lot of benefits as theysharetheirexperience,programming,analyticalskillsanddomainknowledgeintheirrespectiveareas.Theyhavegainedgood real-worldexperience dealing with dirty data and develop important analytical,technical,communicationand communicationskills. Methodsandoutputsofthisprojecthavebeenincludedinthecurriculumby Dr. Martinez-Lopez in the course MPM 207 "Applied Epidemiology ProblemSolving"andMPM212"ConceptsandMethodsinInfectiousDiseaseSurveillanceandControl" at UC Davis.ThesegraduatecoursesareofferedannuallytotheGraduateGroupinEpidemiology (GGE),theMasterofPreventiveVeterinaryMedicine (MPVM)aswellasthe Master and Graduate GroupofPublicHealth(MPH and GGPH)atUCDavis. How have the results been disseminated to communities of interest?We are presenting our results in relevant Conferences andhigh impact Journals (see publications and conference presentations). We also work closely with ourindustrypartnerships, whichinclude the largest swine veterinary diagnostic laboratory (ISU-VDL), the largest swine breeding company (PIC), the largest veterinary clinic in the US (Pipestone), and some of the top swine producers (Seaboard, Pipestone, Tosh Farms). These partners are helping to co-develop, test, validate and maximize Disease BioPortal use and adoption. We hold frequent meetings with them and receive continuous feedback to improve the capabilities and user experience of the platform. We are also conducting meetings andad-hoc training with other veterinarians and producersthat are interested in using the platform.In this way we hope to have maximum impact in the swine industry, improving animal health and helping expand pork markets through our research. What do you plan to do during the next reporting period to accomplish the goals? Continue the integration of new data sources to facilitate real-time analytics and user-friendly visualization of animal health problems Enhanced ability to manage data sources, sequence alignments and WGS uploads. Continue developing, testing and validating the analytical capabilities of the Disease BioPortal to provide value to end-users in the following areas: Risk assessment; Outbreak investigation; Contact-tracing; Early detection of disease; Outbreak prediction; Pathogen classification & diagnosis; Trend & pattern recognition; Prioritization & benchmarking; Real-time metrics of interest; Data and model sharing. Improve visualization and analytical tools using AI-assisted technology, including the use of natural language processing to aid in the generation and interpretation of the dashboards. Improve the Disease BioPortal website and generate documentation and video-demos for each component/tool Participate in conferences, seminars, workshop to maximize the communication of our work and engage new users.

Impacts
What was accomplished under these goals? Development of the Disease BioPortal data lake:We have dedicated substantial amount of time to improve the connection, standardization, integration and security of our data sources by developing a data lake which is used as a centralized repository for all our data. This data lake us to easily and quickly ingest and handle large amounts of structured, semistructured, and unstructured data as well as to speed up any data analyses and visualization. It is definitely a game changer in scalability and flexibility. Developed, beta tested and launched new user-friendly Disease BioPortal tools/capabilities to enhance precision epidemiology at farm or production system level:We have tested different analytical methods, including, machine learning algorithms, for their use in prediction problems in animal health (see publications and conference presentations). We selected the methods that were best performing to develop a new set of operational tools of bioinformatics, spatial-temporal analyses, and artificial intelligence (AI) into Disease BioPortal (https://bioportal.ucdavis.edu), which allows rapid processing and analysis of complex data and easy visualization and interpretation of results. All those capabilities are alreadyaccessibletoourDiseaseBioPortalPROusers and collaborators and will be soon available to the general Disease BioPortal users. Production systems, veterinary clinics, and diagnostic laboratories can easily add proprietary data, such as disease status, production data, animal movement, biosecurity, AMR, etc., using APIs or manual uploads. Disease/AMR prevalence is evaluated at multiple spatial and temporal scales by space-time analyses (e.g., time-series analysis, scan statistics, phylogeography). Genetic diversity and the evolutionary relationship of pathogens are estimated using maximum likelihood-based phylogenetics or core-genome phylogenomics. AMR patterns are assessed using phenotypic (e.g., Minimum inhibitory concentrations) and genotypic (e.g., whole genome sequences) data. The presumed resistance determinants (e.g., AMR genes) are determined using CARD and ResFinder databases. We demonstrate the value of this tool using de-identified datasets ofStreptococcus suisand Porcine Reproductive and Respiratory Syndrome (PRRS) from the United States to understand the spatio-temporal trends of these pathogens and their heterogeneity and resistance to antibiotics, which will be beneficial to develop effective preventive and control strategies. Genomic surveillance ofS. suisand PRRS offers insights into the pathogen evolution, population dynamics, and resistance mechanisms at different spatial and temporal scales. This AI-assisted tool allows users to generate interactive dashboards and facilitates the visualization, analysis, and interpretation of disease/AMR patterns of various swine pathogens promptly. It better informs users of prevention and control strategies appropriate for the target condition (e.g., selecting correct therapeutics/vaccines for circulating variants) and contributes to antimicrobial stewardship.

Publications

• Type: Journal Articles Status: Published Year Published: 2023 Citation: Aguilar-Vega C, Scoglio C, Clavijo MJ, Robbins R, Karriker L, Liu X, Mart�nez-L�pez B. A tool to enhance antimicrobial stewardship using similarity networks to identify antimicrobial resistance patterns across farms. Sci Rep. 2023 Feb 20;13(1):2931. doi: 10.1038/s41598-023-29980-4. PMID: 36804990; PMCID: PMC9941107.
• Type: Journal Articles Status: Published Year Published: 2023 Citation: Halev A, Mart�nez-L�pez B, Clavijo M, Gonzalez-Crespo C, Kim J, Huang C, Krantz S, Robbins R, Liu X. Infection prediction in swine populations with machine learning. Sci Rep. 2023 Oct 18;13(1):17738. doi: 10.1038/s41598-023-43472-5. PMID: 37853003; PMCID: PMC10584972.
• Type: Journal Articles Status: Published Year Published: 2023 Citation: D�az-Cao JM, Liu X, Kim J, Clavijo MJ, Mart�nez-L�pez B. Evaluation of the application of sequence data to the identification of outbreaks of disease using anomaly detection methods. Vet Res. 2023 Sep 8;54(1):75. doi: 10.1186/s13567-023-01197-3. PMID: 37684632; PMCID: PMC10492347.
• Type: Journal Articles Status: Published Year Published: 2023 Citation: Kim J, Rupasinghe R, Halev A, Huang C, Rezaei S, Clavijo MJ, Robbins RC, Mart�nez-L�pez B, Liu X. Predicting antimicrobial resistance of bacterial pathogens using time series analysis. Front Microbiol. 2023 May 11;14:1160224. doi: 10.3389/fmicb.2023.1160224. PMID: 37250043; PMCID: PMC10213968.
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Mart�nez-L�pez B, Clavijo MJ, Robbins R, Gonzalez-Crespo C, Gomez JP, Rupasinghe R, Liu X, Precision epidemiology in practice: applications to better prevent and control endemic diseases in the US swine industry. Veterinaria Italiana. 2023 Oct. Oral presentation. GEOVET 2023. Silvi Marina, Italy. Abstract available at: https://www.veterinariaitaliana.izs.it/index.php/GEOVET23/article/view/3308
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Rupasinghe R, Mugabi R, Clavijo MJ, Robbins R, Mart�nez-L�pez B. Development of an operational tool for genomic and phenotypic surveillance of antimicrobial resistance: applications for swine pathogens in the United States. Veterinaria Italiana. 2023 Oct. Oral presentation. GEOVET 2023. Silvi Marina, Italy. Abstract available at: https://www.veterinariaitaliana.izs.it/index.php/GEOVET23/article/view/3328
• Type: Conference Papers and Presentations Status: Published Year Published: 2023 Citation: Gomez-Vazquez JP, Clavijo MJ, Robbins R, Mart�nez-L�pez B. Spatio-temporal patterns of Pelvic Organ Prolapse in the swine industry in the Midwest region of the United States. Veterinaria Italiana. 2023 Oct. Oral presentation. GEOVET 2023. Silvi Marina, Italy. Abstract available at: https://www.veterinariaitaliana.izs.it/index.php/GEOVET23/article/view/3322

Progress 06/01/21 to 05/31/22

Outputs
Target Audience:Our main target audience is swine veterinarians as well asstakeholders related with the pig industry and pig management such as diagnosticians, farmers, animalhealth officials,extension specialists and other policy makers working in animal health. Changes/Problems:Due to COVID-19 most of our workshops or meetings were conducted online. This has limited our ability todirectly train and communicate with some of our end-users, which clearlyprefer communication/training in person. Similarly, the pandemicdelayed some of our tasks, mainly the collection of new data and development of new APIs, but we expect to catch up this second year with the scheduled plan. What opportunities for training and professional development has the project provided?We are significantly contributing to the training of the next generation of vet-data-scientists, involving directly 3 post-doctoral researchers and 6 PhD students with different backgrounds and technical skills and exposing >100 students per year to these methods and results in the different MPVM, MPH and GGE classes. We meet weekly with our students, and they work collaboratively in the different activities, which has a lot of benefits as theysharetheirexperience,programming,analyticalskillsanddomainknowledgeintheirrespectiveareas.Theyhavegainedgood real-worldexperience dealing with dirty data and develop important analytical,technical,communicationand communicationskills. Methodsandoutputsofthisprojecthavebeenincludedinthecurriculumby Dr. Martinez-Lopez in the course MPM 207 "Applied Epidemiology ProblemSolving"andMPM212"ConceptsandMethodsinInfectiousDiseaseSurveillanceandControl" andpresentedbyDr.ClavijointhecourseEPI290"Epidemiology Seminars" at UC Davis.ThesegraduatecoursesareofferedannuallytotheGraduateGroupinEpidemiology,theMasterofPreventiveVeterinaryMedicineaswellasthe MasterofPublicHealthatUCDavis. How have the results been disseminated to communities of interest?We work closely with ourindustrypartnerships, whichinclude the largest swine veterinary diagnostic laboratory (ISU-VDL), the largest swine breeding company (PIC), the largest veterinary clinic in the US (Pipestone), and some of the top swine producers (Seaboard, Pipestone, Tosh Farms). We have also extended our partnership to include other data providers and critical end users, such as, GlobalVetLINK (GVL), the US leader in aggregation of digital animal records and Metafarms, one of the largest providers of production management software. These partners are helping to co-develop, test, validate and maximize Disease BioPortal use and adoption. We hold bi-weekly meetings with them and receive continuous feedback from them to improve the capabilities and user experience of the platform. We are also conducting meetings andad-hoc training with other veterinarians and producersthat are interested in using the platform.In this way we hope to have maximum impact in the swine industry, improving animal health and helping expand pork markets through our research. What do you plan to do during the next reporting period to accomplish the goals? Continue the integration of new data sources and new data formats to facilitate real-time analytics and user-friendly visualization of animal health problems Enhanced ability to manage data sources, including sequence alignments and WGS uploads. Continue developing, testing and validating the analytical capabilities of the Disease BioPortal to provide value to end-users in the following areas: Risk assessment; Outbreak investigation; Contact-tracing; Early detection of disease; Outbreak prediction; Pathogen classification & diagnosis; Trend & pattern recognition; Prioritization & benchmarking; Real-time metrics of interest; Data sharing Expand the flexibility to visualize better the data and analytical results with new components and pre-defined dashboard templates. Improve the Disease BioPortal website and generate documentation and video-demos for each component/tool Participate in conferences, seminars, workshop to maximize the communication of our work and engage new users.

Impacts
What was accomplished under these goals? 1. Expanded data connection, standardization and integration of our Disease BioPortal platform using new APIs.Connectivity between key information management systems and other data sources available for the swine industry was greatly expanded. Specifically, theconnection between Disease BioPortal and key proprietary databases (i.e., Swine PathogenDatabase, the Iowa Diagnostic Laboratory Data Warehouse and Animal Health Monitoring and Evaluation System, GlobalVetLink LabHIMS and production system specific datasets)was refined and enhanced. The Disease BioPortal is now connected also to different opendatabases, including NCBI - GenBank, NOAH Weather Service and Twitter, allowing users to extract information from those open sources and add them into their dashboards. Theconnection to GenBank and NOAH Weather Service allows us to integrate pathogengenomics data and weather information, respectively, to complement the swine health specific proprietarydata.The integration of current or past weather data and weather forecasts with other animal health information, provides end users with the ability to conduct outbreak investigations (with past weather data) or predictive assessments of outbreak risks (with weather forecasts). Users can ask questions about disease and weather and their interaction, such as: Is the viral sequence of the pathogen I found on my farm different compared to pathogens already published or found somewhere else?; Could weather changes or extreme weather events contributed to the occurrence of outbreaks on my farm (e.g., cold weather linked with respiratory pathogens such as swine flu or PRRS)?. For Twitter, we developed text collection and classification methods that efficiently extract relevant information related to animal health from tweets, including geo-location. Specifically, we have developed a keyword dictionary to filter and collect mostly relevant tweets from Twitter's academic research API. We have also categorized collected tweets by hand and with document clustering methods.These tweets could help for example to early detect outbreaks (e.g. talking about disease or mortality in their animals) or better understand sentiments or comments about specific diseases (e.g. concerns, questions, lack of knowledge, etc. about African swine fever). 2. Evaluated 24 algorithms for anomaly detection and selection of the best ones to early detect disease outbreaks.Anomaly detection was used to generate an early warning system for the occurrence of outbreaks of disease. Anomaly detection methods can monitor real-time data from different sources, such as diagnostic results or mortality records, and detect outbreaks as they deviate from the normal pattern of data. However, the choice and correct implementation of a method is not easy because algorithms may perform (i.e. specificity and sensitivity) differently depending on time series characteristics and algorithm parameterization. A systematic evaluation of most methods available and selection of the best ones specifically for veterinary diagnostic data was conducted. The best algorithms identified are now being integrated into the Disease BioPortal so end users can use them to early detect outbreaks and take action before economic and sanitary consequences become too damaging. 3. Developed, beta tested and launched new user-friendly Disease BioPortaltools/capabilities to enhance precision epidemiology at farm or production system level.We have developed new components (i.e.,antimicrobialresistancecharts,WGSphyloviewer,bacteriaheatmaps,farmhealthstatus charts) to produce advanced diagnostic and AMR dashboards. All those capabilities are alreadyaccessibletotheDiseaseBioPortalusers throughhttps://bioportal.ucdavis.eduandaresummarizedbelow: Diagnostic dashboards:Thanks to the expanded data connection, standardizationandintegration,end-userscannowcreatecustomdiagnosticdashboardsbypathogen, by disease and/orbyfarm. This includes dashboards that allows: Access and visualization (by lesion code/etiology) of the histopathology reports information. Integration of pig movement data and production data with diagnostic information to facilitate outbreak investigations and high-risk contact tracing Early detection and evaluation of the impact of animal health problems associated with production changes Expansion of the phylogenetic suite of tools with the incorporation of the NCBI import tool within the Disease BioPortal and the possibility to upload NGS data (i.e., NGS data upload tool). AMR dashboards: The WGS is now integrated into the Swine Pathogen Databasefor detection and annotation of resistance determinants. Resistance Gene Identifiersoftware of Comprehensive Antibiotic Resistance Database (CARD) and Resfinder tools areconnectednowtotheDiseaseBioPortaltopredictresistomeforgenomicsequencing data. CARD focuses on providing high-quality reference data and molecular sequenceswithin a controlled vocabulary, and antibiotic resistance ontology to integrate withsoftware development efforts for resistome analysis and prediction. CARD's antibioticresistanceontologyhasgrownto6657ontologyterms,covering5078 AMR detection models and resistome predictions for more than 377 pathogens. In addition to the ontology model, CARD includes the rRNA gene variant model and the protein overexpression model. We used "perfect" and "strict" algorithms to identify perfect matches (100% identical) and previously unknown variants of known AMR genes in CARD. In Resfinder, the threshold was set to 90% to identify AMR genes that were identical to the best matching resistance genes in the Resfinder database, and the minimum length of coverage of each resistance gene was set to 60%. The AMR genes in each isolate determined from each method can be visualized as heatmaps and circle plots within the Disease BioPortal, which allows users to: Incorporate and visualize antimicrobial resistance (AMR) phenotypic information (MIC values and interpretation) Extract and visualize AMR genotypic profiles (i.e. resistance genes, virulence factors, MLST) based on whole genome sequences and creation of new visualization and analytical components (i.e. heatmaps, WGS trees and predicted resistomes with donut charts or circle plots). Thesenewvisualizationandanalyticaltoolshave beentested by our industry collaborators using their ownproprietary data. The new capabilities were also evaluated with de-identified datasets during practical exercises in graduate courses for the students of the Master of Preventative Veterinary Medicine (MPVM), Graduate Group in Epidemiology(GGE)and, Master of Public Health (MPH) programs at UC Davis.The feedback obtained from these interactions has allowed us to improve the look-and-feel of these capabilities and make them more useful and easier to understand by our target audience.

Publications

• Type: Journal Articles Status: Awaiting Publication Year Published: 2022 Citation: Diaz Cao JM, Liu X, Kim J, Clavijo MJ, Mart�nez-L�pez B. Evaluation of anomaly detection methods to detect disease outbreaks of disease from laboratory result data. Frontiers in Veterinary Sciences