Optimization of resistance to gastrointestinal nematodes in Santa Inês sheep: a genomic selection, machine learning and image analysis approach

• Autor(a) principal: Januário, Luara Afonso de Freitas
• Data de Publicação: 2023
• Tipo de documento: Tese
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/17/17135/tde-05062023-132524/
• topic: Análise de imagem; Gastrointestinal nematodes; Genomic selection; Image analysis; Machine learning; Machine learning; Nematódeos gastrointestinais; Ovis aries; Ovis aries; Seleção genômica

Gastrointestinal nematode infection represents a major threat to the health and productivity of sheep populations, and the Haemonchus contortus is the most pathogenic species. This study analyzed a population of Santa Ines sheep and it was composed of five chapters with the following objectives: Chapter 1) Literature review; Chapter 2) To evaluate the feasibility of using easily-measured phenotypic traits in order to predict the susceptibility of sheep to gastrointestinal nematodes through the use of machine learning methods; Chapter 3) To analyze ocular conjunctiva images to classify anemia levels based on Famacha© scores (FAM); Chapter 4) To examine the additive-genetic patterns of estimated breeding values (EBVs) for indicator traits of resistance to gastrointestinal nematodes; Chapter 5) To assess the accuracy of parametric models (GBLUP, BayesA, BayesB e Bayesian Lasso - BLASSO) and artificial neural networks (BRANN) in genomic predictions of indicator traits of resistance to gastrointestinal nematodes. In the Chapter 2, the animals were classified into resistant, resilient, and susceptible according to fecal egg count (FEC) and packed cell volume (PCV), and the methods were fitted using the information of age class, the month of record, farm, sex, FAM, body weight, and body condition score as predictors. In the Chapter 3, a random forest model (RF) was used to segment the images. After segmentation, the quantiles of color intensity (1, 10, 20, 30, 40, 50, 60, 70, 80, 90, and 99%) in each image channel (red, blue, and green) were determined and used as explanatory variables in the classification models, and the FAM 1 to 5 were the target classes to be predicted. In the Chapter 4, The EBVs for FAM, PCV, and FEC were estimated by Bayesian inference in a single-trait animal model. After, cluster analyses were done using the EBVs for FAM, PCV, and FEC in order to identify animals that are resistant, resilient, and susceptible to gastrointestinal nematodes. In the Chapter 5, the prediction accuracy and mean squared errors were obtained for PCV, FEC, and FAM using parametric models and artificial neural network. The results suggest that the use of easily measurable traits may provide useful information for supporting management decisions at the farm level. The image analysis results indicate that is possible to successfully predict Famacha© score, especially for scores 2 to 4, in sheep via image analysis and RF model using ocular conjunctiva images collected in farm conditions. The resistant cluster presented positive EBVs for PCV and negative for FAM and FEC, being consisted of the most desirable animals to be used as selection candidates in order to genetically improve resistance to gastrointestinal nematodes. In summary, parametric models are suitable for genome-enabled prediction of PCV, FEC and FAM in sheep, due to the small differences in accuracy found between them. Despite this, the use of the GBLUP model is recommended due to its lower computational costs and the possibility of incorporating non-genotyped animals into the analysis using single-step procedures.