Utilizing machine learning and hemagglutinin sequences to identify likely hosts of influenza H3Nx viruses.

Influenza is a disease that represents both a public health and agricultural risk with pandemic potential. Among the subtypes of influenza A virus, H3 influenza virus can infect many avian and mammalian species and is therefore a virus of interest to human and veterinary public health. The primary goal of this study was to train and validate classifiers for the identification of the most likely host species using the hemagglutinin gene segment of H3 viruses. A five-step process was implemented, which included training four machine learning classifiers, testing the classifiers on the validation dataset, and further exploration of the best-performing model on three additional datasets. The gradient boosting machine classifier showed the highest host-classification accuracy with a 98.0 % (95 % CI [97.01, 98.73]) correct classification rate on an independent validation dataset. The classifications were further analyzed using the predicted probability score which highlighted sequences of particular interest. These sequences were both correctly and incorrectly classified sequences that showed considerable predicted probability for multiple hosts. This showed the potential of using these classifiers for rapid sequence classification and highlighting sequences of interest. Additionally, the classifiers were tested on a separate swine dataset composed of H3N2 sequences from 1998 to 2003 from the United States of America, and a separate canine dataset composed of canine H3N2 sequences of avian origin. These two datasets were utilized to look at the applications of predicted probability and host convergence over time. Lastly, the classifiers were used on an independent dataset of environmental sequences to explore the host identification of environmental sequences. The results of these classifiers show the potential for machine learning to be used as a host identification technique for viruses of unknown origin on a species-specific level.

Prevalence and distribution of carbapenem-resistant Enterobacterales in companion animals: A nationwide study in the United States using commercial laboratory data.

BACKGROUND: Carbapenem-resistant Enterobacterales (CRE) are a concern in both human and animal medicine globally. Despite extensive research in humans, limited data exist on CRE in companion animals, with a lack of nationwide prevalence estimates. HYPOTHESIS/OBJECTIVES: To assess the occurrence and trends of CRE in cats and dogs across the United States by analyzing 4 years of commercial antimicrobial susceptibility testing (AST) data. ANIMALS: Between 2019 and 2022, 477 426 ASTs were conducted on Enterobacterales isolates against imipenem. Isolates were derived from 379 598 dogs and 97 828 cats. Animal origin was not disclosed. METHODS: In this retrospective study, antimicrobial susceptibility test data from IDEXX Laboratories were analyzed. Analysis included resistance estimations to imipenem stratified by sampling site, an assessment of resistance patterns over time and location, and the application of space-time cluster analysis to identify potential outbreaks. Antibiograms were produced for carbapenem-resistant isolates. RESULTS: Susceptibility to imipenem was high, at 98.86%. Temporal analysis indicated stability in susceptibility, with an unexplained reduction in susceptible isolates in June 2019. Spatial analysis identified 2 high-risk clusters along the Western Coast (relative risk [RR]: 23.26; P < .001) and in Texas (RR: 10.72; P < .001) in that month. Three other clusters were found, in Missouri (RR: 39.55; P = .038), Florida (RR: 4.53; P < .001), and New York (RR: 9.20; P < .001). CONCLUSIONS AND CLINICAL IMPORTANCE: CRE are present at a low prevalence in dogs and cats across the United States. Variations in prevalence across patient-level and environmental factors highlight the need for tailored stewardship programs.

Examining the effects of voluntary avoidance behaviour and policy-mediated behaviour change on the dynamics of SARS-CoV-2: A mathematical model.

Background: Throughout the SARS-CoV-2 pandemic, policymakers have had to navigate between recommending voluntary behaviour change and policy-driven behaviour change to mitigate the impact of the virus. While individuals will voluntarily engage in self-protective behaviour when there is an increasing infectious disease risk, the extent to which this occurs and its impact on an epidemic is not known. Methods: This paper describes a deterministic disease transmission model exploring the impact of individual avoidance behaviour and policy-mediated avoidance behaviour on epidemic outcomes during the second wave of SARS-CoV-2 infections in Ontario, Canada (September 1, 2020 to February 28, 2021). The model incorporates an information feedback function based on empirically derived behaviour data describing the degree to which avoidance behaviour changed in response to the number of new daily cases COVID-19. Results: Voluntary avoidance behaviour alone was estimated to reduce the final attack rate by 23.1%, the total number of hospitalizations by 26.2%, and cumulative deaths by 27.5% over 6 months compared to a counterfactual scenario in which there were no interventions or avoidance behaviour. A provincial shutdown order issued on December 26, 2020 was estimated to reduce the final attack rate by 66.7%, the total number of hospitalizations by 66.8%, and the total number of deaths by 67.2% compared to the counterfactual scenario. Conclusion: Given the dynamics of SARS-CoV-2 in a pre-vaccine era, individual avoidance behaviour in the absence of government action would have resulted in a moderate reduction in disease however, it would not have been sufficient to entirely mitigate transmission and the associated risk to the population in Ontario. Government action during the second wave of the COVID-19 pandemic in Ontario reduced infections, protected hospital capacity, and saved lives.

Forecasting seasonal influenza activity in Canada-Comparing seasonal Auto-Regressive integrated moving average and artificial neural network approaches for public health preparedness.

INTRODUCTION: Public health preparedness is based on timely and accurate information. Time series forecasting using disease surveillance data is an important aspect of preparedness. This study compared two approaches of time series forecasting: seasonal auto-regressive integrated moving average (SARIMA) modelling and the artificial neural network (ANN) algorithm. The goal was to model weekly seasonal influenza activity in Canada using SARIMA and compares its predictive accuracy, based on root mean square prediction error (RMSE) and mean absolute prediction error (MAE), to that of an ANN. METHODS: An initial SARIMA model was fit using automated model selection by minimizing the Akaike information criterion (AIC). Further inspection of the autocorrelation function and partial autocorrelation function led to 'manual' model improvements. ANNs were trained iteratively, using an automated process to minimize the RMSE and MAE. RESULTS: A total of 378, 462 cases of influenza was reported in Canada from the 2010-2011 influenza season to the end of the 2019-2020 influenza season, with an average yearly incidence risk of 20.02 per 100,000 population. Automated SARIMA modelling was the better method in terms of forecasting accuracy (per RMSE and MAE). However, the ANN correctly predicted the peak week of disease incidence while the other models did not. CONCLUSION: Both the ANN and SARIMA models have shown to be capable tools in forecasting seasonal influenza activity in Canada. It was shown that applying both in tandem is beneficial, SARIMA better forecasted overall incidence while ANN correctly predicted the peak week.

Development and validation of a farm- and province-level swine flow simulation model using discrete events and Ontario swine farm and provincial input data.

Infectious disease events can cause disruptions in service-based and agricultural industries. The list of possible events is long and varies from the incursion or emergence of a reportable animal pathogen to the recently documented interruptions caused by the COVID-19 pandemic. There is a need to develop models that can determine the impact of pathogens and mitigation measures on populations that are not directly affected by the pathogen in the case of a reportable disease, particularly when the health and welfare of these populations could be affected due to resulting disruptions in trade and supply chains. The primary objective of this study was to develop a discrete-event simulation (DES) model of swine production, including pork processing, for scenarios without major disruptions, which could be scaled from the level of an individual farm to the entire province of Ontario, Canada. The secondary objective was to validate the developed simulation against observed farm- and province-level statistics. A weekly discrete-event simulation consisting of 3 connected areas (a sow farm, a pig farm, and abattoirs) was developed using AnyLogic modelling software. Using Mann-Whitney tests, model outputs representative of the standard industry statistics were compared to data from 6 individual farms separately, as well as to provincial data from Ontario. A scalable discrete-event simulation of the swine production system for typical scenarios was accomplished. The model outputs were consistent with individual farm and industry statistics. As such, the model can be used to simulate swine production at distinct levels and could be further modified to represent swine marketing in other provinces or internationally.

Les maladies infectieuses peuvent provoquer des perturbations dans les industries de services et agricoles. La liste des événements possibles est longue et varie de l'arrivée ou de l'émergence d'un agent pathogène animal à déclaration obligatoire aux interruptions récemment documentées causées par la pandémie de COVID-19. Il est nécessaire d'élaborer des modèles permettant de déterminer l'impact des agents pathogènes et des mesures d'atténuation sur les populations qui ne sont pas directement affectées par l'agent pathogène dans le cas d'une maladie à déclaration obligatoire, en particulier lorsque la santé et le bien-être de ces populations pourraient être affectés en raison des conséquences dues aux perturbations du commerce et des chaînes d'approvisionnement. L'objectif principal de cette étude était de développer un modèle de simulation à événements discrets (DES) de la production porcine, y compris la transformation du porc, pour des scénarios sans perturbations majeures, qui pourraient être étendus du niveau d'une ferme individuelle à l'ensemble de la province de l'Ontario, Canada. L'objectif secondaire était de valider la simulation développée par rapport aux statistiques observées au niveau de la ferme et de la province. Une simulation à événements discrets hebdomadaire composée de 3 zones connectées (un élevage de truies, un élevage de porcs et des abattoirs) a été développée à l'aide du logiciel de modélisation AnyLogic. À l'aide des tests de Mann-Whitney, les résultats du modèle représentatifs des statistiques standards de l'industrie ont été comparés aux données de 6 fermes individuelles séparément, ainsi qu'aux données provinciales de l'Ontario. Une simulation à événements discrets évolutive du système de production porcine pour des scénarios typiques a été réalisée. Les résultats du modèle étaient cohérents avec les statistiques individuelles des exploitations et des industries. Ainsi, le modèle peut être utilisé pour simuler la production porcine à des niveaux distincts et pourrait être modifié davantage pour représenter la commercialisation du porc dans d'autres provinces ou à l'échelle internationale.(Traduit par Docteur Serge Messier).

Does Vaccine-Induced Maternally-Derived Immunity Protect Swine Offspring against Influenza a Viruses? A Systematic Review and Meta-Analysis of Challenge Trials from 1990 to May 2021.

It is unclear if piglets benefit from vaccination of sows against influenza. For the first time, methods of evidence-based medicine were applied to answer the question: "Does vaccine-induced maternally-derived immunity (MDI) protect swine offspring against influenza A viruses?". Challenge trials were reviewed that were published from 1990 to April 2021 and measured at least one of six outcomes in MDI-positive versus MDI-negative offspring (hemagglutination inhibition (HI) titers, virus titers, time to begin and time to stop shedding, risk of infection, average daily gain (ADG), and coughing) (n = 15). Screening and extraction of study characteristics was conducted in duplicate by two reviewers, with data extraction and assessment for risk of bias performed by one. Homology was defined by the antigenic match of vaccine and challenge virus hemagglutinin epitopes. Results: Homologous, but not heterologous MDI, reduced virus titers in piglets. There was no difference, calculated as relative risks (RR), in infection incidence risk over the entire study period; however, infection hazard (instantaneous risk) was decreased in pigs with MDI (log HR = -0.64, 95% CI: -1.13, -0.15). Overall, pigs with MDI took about a ½ day longer to begin shedding virus post-challenge (MD = 0.51, 95% CI: 0.03, 0.99) but the hazard of infected pigs ceasing to shed was not different (log HR = 0.32, 95% CI: -0.29, 0.93). HI titers were synthesized qualitatively and although data on ADG and coughing was extracted, details were insufficient for conducting meta-analyses. Conclusion: Homology of vaccine strains with challenge viruses is an important consideration when assessing vaccine effectiveness. Herd viral dynamics are complex and may include concurrent or sequential exposures in the field. The practical significance of reduced weaned pig virus titers is, therefore, not known and evidence from challenge trials is insufficient to make inferences on the effects of MDI on incidence risk, time to begin or to cease shedding virus, coughing, and ADG. The applicability of evidence from single-strain challenge trials to field practices is limited. Despite the synthesis of six outcomes, challenge trial evidence does not support or refute vaccination of sows against influenza to protect piglets. Additional research is needed; controlled trials with multi-strain concurrent or sequential heterologous challenges have not been conducted, and sequential homologous exposure trials were rare. Consensus is also warranted on (1) the selection of core outcomes, (2) the sizing of trial populations to be reflective of field populations, (3) the reporting of antigenic characterization of vaccines, challenge viruses, and sow exposure history, and (4) on the collection of non-aggregated individual pig data.

Development of interactive dashboards for monitoring endemic animal pathogens in Ontario, Canada: Ontario interactive animal pathogen dashboards.

The advancement of web-based technologies makes it possible to build user interfaces or web pages that present and summarize complex data in easy-to-read graphical formats that emphasize key information. Taking advantage of this technologic progress, we addressed the need for real-time visualizations of trends for major pathogens in the largest livestock industries in Ontario: poultry, swine, and cattle. These visualizations were built using test data from the laboratory information management system of the Animal Health Laboratory at the University of Guelph, a large veterinary diagnostic laboratory in Ontario. The data were processed using R software and used to construct interactive and dynamic visualizations using Tableau Desktop v.2021.4 (Tableau Software). We designed 12 dashboards: in chickens-influenza A virus, fowl adenovirus, infectious bronchitis virus, and infectious laryngotracheitis virus; in turkeys-influenza A virus; in swine, influenza A virus, rotavirus, and porcine reproductive and respiratory syndrome virus; in cattle-bovine viral diarrhea virus, Mycoplasma bovis, Salmonella Dublin in individual samples, and Salmonella Dublin in bulk tank milk samples. Data for each pathogen are presented in 2 dashboards. One shows the data of the last 10 y (general view) and the other the data of the last 3 y, but in more detail (comprehensive view). Information on gaining access to all dashboards is available at https://iapd.lsd.uoguelph.ca/. The visualizations provide near-real-time access to aggregated assay results for selected pathogens for veterinarians, animal health regulatory agencies, researchers, and other users who are interested in livestock pathogen surveillance.

Epidemiology of companion animal AMR in the United States of America: filling a gap in the one health approach.

Introduction: Antimicrobial resistance (AMR) is a global health concern that affects all aspects of the One Health Triad, including human, animal, and environmental health. Companion animals, such as cats and dogs, may contribute to the spread of AMR through their close contact with humans and the frequent prescription of antimicrobials. However, research on AMR in companion animals is limited, and there are few surveillance measures in place to monitor the spread of resistant pathogens in the United States. Methods: This study aims to explore the practicality of using data from commercial laboratory antimicrobial susceptibility testing (AST) services for epidemiological analyses of AMR in companion animals in the United States. Results: The study analyzed 25,147,300 individual AST results from cats and dogs submitted to a large commercial diagnostic laboratory in the United States between 2019 and 2021, and found that resistance to certain antimicrobials was common in both E. coli and S. pseudintermedius strains. Conclusion: There has been a paucity of information regarding AMR in companion animals in comparison to human, environmental and other animal species. Commercial AST datasets may prove beneficial in providing more representation to companion animals within the One Health framework for AMR.

Classification of porcine reproductive and respiratory syndrome clinical impact in Ontario sow herds using machine learning approaches.

Since the early 1990s, porcine reproductive and respiratory syndrome (PRRS) virus outbreaks have been reported across various parts of North America, Europe, and Asia. The incursion of PRRS virus (PRRSV) in swine herds could result in various clinical manifestations, resulting in a substantial impact on the incidence of respiratory morbidity, reproductive loss, and mortality. Veterinary experts, among others, regularly analyze the PRRSV open reading frame-5 (ORF-5) for prognostic purposes to assess the risk of severe clinical outcomes. In this study, we explored if predictive modeling techniques could be used to identify the severity of typical clinical signs observed during PRRS outbreaks in sow herds. Our study aimed to evaluate four baseline machine learning (ML) algorithms: logistic regression (LR) with ridge and lasso regularization techniques, random forest (RF), k-nearest neighbor (KNN), and support vector machine (SVM), for the clinical impact classification of ORF-5 sequences and demographic data into high impact and low impact categories. First, baseline classifiers were evaluated using different input representations of ORF-5 nucleotides, amino acid sequences, and demographic data using a 10-fold cross-validation technique. Then, we designed a consensus voting ensemble approach to aggregate the different types of input representations for genetic and demographic data for classifying clinical impact. In this study, we observed that: (a) for abortion and pre-weaning mortality (PWM), different classifiers gained improvement over baseline accuracy, which showed the plausible presence of both genotypic-phenotypic and demographic-phenotypic relationships, (b) for sow mortality (SM), no baseline classifier successfully established such linkages using either genetic or demographic input data, (c) baseline classifiers showed good performance with a moderate variance of the performance metrics, due to high-class overlap and the small dataset size used for training, and (d) the use of consensus voting ensemble techniques helped to make the predictions more robust and stabilized the performance evaluation metrics, but overall accuracy did not substantially improve the diagnostic metrics over baseline classifiers.

Classification of group A rotavirus VP7 and VP4 genotypes using random forest.

Introduction: Group A rotaviruses are major pathogens in causing severe diarrhea in young children and neonates of many different species of animals worldwide and group A rotavirus sequence data are becoming increasingly available over time. Different methods exist that allow for rotavirus genotyping, but machine learning methods have yet to be explored. Usage of machine learning algorithms such as random forest alongside alignment-based methodology may allow for both efficient and accurate classification of circulating rotavirus genotypes through the dual classification system. Methods: Random forest models were trained on positional features obtained from pairwise and multiple sequence alignment and cross-validated using methods of repeated 10-fold cross-validation thrice and leave one- out cross validation. Models were then validated on unseen data from the testing datasets to observe real-world performance. Results: All models were found to perform strongly in classification of VP7 and VP4 genotypes with high overall accuracy and kappa values during model training (0.975-0.992, 0.970-0.989) and during model testing (0.972-0.996, 0.969-0.996), respectively. Models trained on multiple sequence alignment generally had slightly higher overall accuracy and kappa values than models trained on pairwise sequence alignment method. In contrast, pairwise sequence alignment models were found to be generally faster than multiple sequence alignment models in computational speed when models do not need to be retrained. Models that used repeated 10-fold cross-validation thrice were also found to be much faster in model computational speed than models that used leave-one-out cross validation, with no noticeable difference in overall accuracy and kappa values between the cross-validation methods. Discussion: Overall, random forest models showed strong performance in the classification of both group A rotavirus VP7 and VP4 genotypes. Application of these models as classifiers will allow for rapid and accurate classification of the increasing amounts of rotavirus sequence data that are becoming available.