Supervised Machine Learning Techniques for Breeding Value Prediction in Horses: An Example Using Gait Visual Scores.

Gait scores are widely used in the genetic evaluation of horses. However, the nature of such measurement may limit genetic progress since there is subjectivity in phenotypic information. This study aimed to assess the application of machine learning techniques in the prediction of breeding values for five visual gait scores in Campolina horses: dissociation, comfort, style, regularity, and development. The dataset contained over 5000 phenotypic records with 107,951 horses (14 generations) in the pedigree. A fixed model was used to estimate least-square solutions for fixed effects and adjusted phenotypes. Variance components and breeding values (EBV) were obtained via a multiple-trait model (MTM). Adjusted phenotypes and fixed effects solutions were used to train machine learning models (using the EBV from MTM as target variable): artificial neural network (ANN), random forest regression (RFR) and support vector regression (SVR). To validate the models, the linear regression method was used. Accuracy was comparable across all models (but it was slightly higher for ANN). The highest bias was observed for ANN, followed by MTM. Dispersion varied according to the trait; it was higher for ANN and the lowest for MTM. Machine learning is a feasible alternative to EBV prediction; however, this method will be slightly biased and over-dispersed for young animals.

Genetic analysis of feed efficiency and novel feeding behavior traits measured in group-housed broilers using electronic feeders.

This study aimed to estimate genetic parameters for feeding behavior (FB) traits and to assess their genetic relationship with performance traits in group-housed broilers. In total, 99,472,151 visits were recorded for 95,711 birds between 2017 and 2022 using electronic feeders. The visits were first clustered into 2,667,617 daily observations for ten FB traits: daily feed intake (DFI), daily number of visits (NVIS), time spent at the feeders (TSF), number of visited feeders (NVF), visiting activity interval (VAI), feeding rate (FR), daily number of meals (NMEAL), average intake per meal (INTMEAL), number of visits per meal (VISMEAL) and interval between meals (MEALIVL). All FB traits were then considered as the average per bird across the feeding test period. Three growth traits (body weight at the start - SBW and at the end of the feeding test - FBW, and weight gain over the test period - BWG), and 2 feed efficiency (FE) traits (Feed Conversion Rate - FCR and Residual Feed Intake - RFI) were also recorded. The (co)variance components were estimated using multitrait animal mixed models. For growth and FE, the heritability (h2) estimates were moderate, ranging from 0.20 ± 0.01 (BWG) to 0.32 ± 0.02 (RFI). Overall, the h2 estimates for FB traits were higher than for productive traits, ranging from 0.31 ± 0.01 (DFI) to 0.56 ± 0.02 (TSF). DFI presented high genetic correlations (0.53-0.86) with all performance traits. Conversely, the remaining FB traits presented null to moderate genetic correlations with these traits, ranging from -0.38 to 0.42 for growth traits and between -0.14 and 0.25 for FE traits. Genetic selection for favorable feeding behavior is expected to exhibit a fast genetic response. The results suggest that it is possible to consider different feeding strategies without compromising the genetic progress of FE. Conversely, breeding strategies prioritizing a higher bird activity might result in lighter broiler lines in the long term, given the negative genetic correlations between visit-related traits (NV, NVF, and NMEAL) and growth traits (SBW and FBW).

Benchmarking machine learning and parametric methods for genomic prediction of feed efficiency-related traits in Nellore cattle.

Genomic selection (GS) offers a promising opportunity for selecting more efficient animals to use consumed energy for maintenance and growth functions, impacting profitability and environmental sustainability. Here, we compared the prediction accuracy of multi-layer neural network (MLNN) and support vector regression (SVR) against single-trait (STGBLUP), multi-trait genomic best linear unbiased prediction (MTGBLUP), and Bayesian regression (BayesA, BayesB, BayesC, BRR, and BLasso) for feed efficiency (FE) traits. FE-related traits were measured in 1156 Nellore cattle from an experimental breeding program genotyped for ~ 300 K markers after quality control. Prediction accuracy (Acc) was evaluated using a forward validation splitting the dataset based on birth year, considering the phenotypes adjusted for the fixed effects and covariates as pseudo-phenotypes. The MLNN and SVR approaches were trained by randomly splitting the training population into fivefold to select the best hyperparameters. The results show that the machine learning methods (MLNN and SVR) and MTGBLUP outperformed STGBLUP and the Bayesian regression approaches, increasing the Acc by approximately 8.9%, 14.6%, and 13.7% using MLNN, SVR, and MTGBLUP, respectively. Acc for SVR and MTGBLUP were slightly different, ranging from 0.62 to 0.69 and 0.62 to 0.68, respectively, with empirically unbiased for both models (0.97 and 1.09). Our results indicated that SVR and MTGBLUBP approaches were more accurate in predicting FE-related traits than Bayesian regression and STGBLUP and seemed competitive for GS of complex phenotypes with various degrees of inheritance.

Classification Performance of Machine Learning Methods for Identifying Resistance, Resilience, and Susceptibility to Haemonchus contortus Infections in Sheep.

This study investigated the feasibility of using easy-to-measure phenotypic traits to predict sheep resistant, resilient, and susceptible to gastrointestinal nematodes, compared the classification performance of multinomial logistic regression (MLR), linear discriminant analysis (LDA), random forest (RF), and artificial neural network (ANN) methods, and evaluated the applicability of the best classification model on each farm. The database comprised 3654 records of 1250 Santa Inês sheep from 6 farms. The animals were classified into resistant (2605 records), resilient (939 records), and susceptible (110 records) according to fecal egg count and packed cell volume. A random oversampling method was performed to balance the dataset. The classification methods were fitted using the information of age class, the month of record, farm, sex, Famacha© degree, body weight, and body condition score as predictors, and the resistance, resilience, and susceptibility to gastrointestinal nematodes as the target classes to be predicted considering data from all farms randomly. An additional leave-one-farm-out cross-validation technique was used to assess prediction quality across farms. The MLR and LDA models presented good performances in predicting susceptible and resistant animals. The results suggest that the use of readily available records and easily measurable traits may provide useful information for supporting management decisions at the farm level.