Predição de linhagens em gado de leite para a redução da endogamia visando o aconselhamento de cruzamentos em programas de melhoramento genético animal

Abstract

Advancements in molecular genetics have enabled an efficient genotyping process in animal breeding programs, utilizing thousands of single nucleotide polymorphism (SNP) markers. These data expedite genetic selection, providing more accurate estimates of the genetic value of animals without their own phenotypic records. In Brazil, animal breeding programs aim to identify superior animals for genetic multiplication, ensuring economic and environmental sustainability in dairy farming. Genetic evaluations employ models and statistics to calculate genetic values. The need to address inbreeding arises in the context of genetic improvement, especially when selecting high-performance animals for crossbreeding and herd development. This study focuses on predicting lineages in dairy cattle using machine learning. Supervised analyses, based on pedigree, and unsupervised analyses, solely exploring molecular markers, were employed. The models exhibited high accuracy in classifying animals from the top five progeny bulls into lineages, validated in training and testing datasets. The unsupervised approach provided additional insights into the genetic composition of the population, indicating promising paths for future studies. The importance of genealogical annotation in lineage identification is emphasized. Supervised models showed promising improvements in identifying new animals with prior knowledge of the population. Additionally, the possibility of future implementation of these models as support for handling genetic diversity and enhancing animal selection in animal breeding programs was verified.