Poster
Predicting Bovine Trichomoniasis in Florida Beef Cattle Using Supervised Machine Learning: A 10-year retrospective study
| Date/Time: |
9/11/2025
|
| Author: |
Ameer Megahed |
| Clinic: |
University of Florida |
| City, State, ZIP: |
Gainesville, FL 32601 |
A. Megahed, BVSc, MS, PhD, MPVM
1
;
R. Bommineni, BVSc, MS, PhD
2
;
M. Short, DVM
2
;
J.J. Bittar , DVM, MS, PhD
1
;
1Department of Large Animal Clinical Sciences, College of Veterinary Medicine, University of Florida, Gainesville, FL, 32610
2Division of Animal Industry, Florida Department of Agriculture and Consumer Services
Introduction:
The limited availability of current data on the prevalence and risk factors of Bovine Trichomoniasis (BT) across the United States presents significant challenges for accurately evaluating production losses and implementing effective prevention and control measures. Supervised machine learning (SML) algorithms offer a promising approach for identifying risk factors associated with infectious diseases like BT. In this study, we evaluated and compared six distinct SML models to predict BT in the beef cattle populations within Florida.
Materials and methods:
A dataset comprising 5,129 preputial smegma samples collected from beef bulls in Florida was used for this study. These samples were submitted to the Bronson Animal Disease Diagnostic Laboratory for T. foetus testing using quantitative PCR (qPCR) between 2012 and 2022. Six supervised machine learning (SML) algorithms were employed for analysis, including logistic regression (LR), decision tree (DT), gradient boosting (GB), random forest (RF), neural network (NN), and support vector machine (SVM).
Results:
A total of 377 (7.4, 95% CI 6.6 to 8.1) preputial smegma samples were positive for T. foetus. Among the evaluated models, the RF algorithm demonstrated the highest predictive accuracy for T. foetus infection in beef cattle, as evidenced by the highest Kolmogorov-Smirnov (KS) statistic of 0.60, an area under the receiver operating characteristic curve (AUROC) of 0.86, and the lowest misclassification rate of 0.06. Conversely, the NN model exhibited the poorest performance. Feature importance analysis within the RF model indicated that testing Beefmaster bulls aged ≥5 years prior to the breeding season provided the most reliable prediction of T. foetus infection, offering valuable insights for targeted disease control strategies.
Significance:
We concluded that RF and other SML algorithms exhibit strong predictive potential for identifying BT in beef cattle in Florida. Age (≥5 years), breed (Beefmaster), and season (pre-breeding period) appear to be key risk factors for BT in bulls within Florida herds. These results provide a framework for developing precision-based predictive tools to enhance BT screening, establish risk-focused early warning systems, and implement targeted surveillance strategies, particularly for high-risk cohorts, to reduce transmission and economic impacts in endemic regions.
