Machine Learning-Based Genome-Wide Association Study Reveals Genetic Loci Associated with Body Measurement Traits in Yili Horses.

Body measurement traits are key indicators for evaluating growth performance, production potential, and breeding value in Yili horses. However, studies investigating the association between body measurement traits and mutation loci in Yili horses remain limited. In this study, 255 adult Yili mares were used as the study population, including 152 speed-type and 103 meat-type individuals. Whole-genome resequencing was performed, and four phenotypic traits and body weight were measured. A mixed linear model (MLM)-based genome-wide association study (GWAS) was conducted using GEMMA (v 0.98.5), incorporating age, farm effects, and top three principal components as covariates. In parallel, a machine learning-based GWAS (ML-GWAS) framework integrating Lasso regression for feature selection and Random Forest (RF) with five-fold cross-validation was applied to improve the detection of complex genetic signals. Using both conventional GWAS methods and machine learning-based GWAS approaches, a total of 238 mutation loci significantly associated with body measurement traits were identified, and 277 candidate genes were annotated. These genes may play a role in several biological processes, including skeletal development, muscle formation, cell growth, energy metabolism, and protein synthesis. The findings suggest that genetic variations have already manifested among the studied groups. The results indicate that genetic differences have already emerged among different Yili horse populations at the genomic level. Furthermore, this study demonstrates that integrating machine learning with conventional GWAS effectively improves the detection efficiency of loci associated with complex traits, while also providing new molecular evidence for understanding the genetic mechanisms underlying differences in body measurement traits among Yili horse groups.