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

Machine learning combined with traditional genome-wide association studies was used to identify genetic loci linked to body measurement traits in Yili horses, revealing important genes related to growth and development. The study highlights the effectiveness of machine learning in detecting complex genetic signals influencing these traits.

Study Population and Traits Investigated

• The research focused on 255 adult Yili mares, divided into two groups: 152 speed-type and 103 meat-type horses.
• Four phenotypic body measurement traits along with body weight were measured as indicators of growth, production potential, and breeding value.

Genomic Data Collection and Analysis Methods

• Whole-genome resequencing was performed on all horses to obtain detailed genetic information.
• A mixed linear model (MLM)-based genome-wide association study (GWAS) was conducted using GEMMA software, accounting for age, farm effects, and population structure (top three principal components).
• In parallel, a novel machine learning-based GWAS (ML-GWAS) approach was employed that combined Lasso regression for feature selection and Random Forest (RF) with five-fold cross-validation to detect complex genetic associations more effectively.

Identification of Mutation Loci and Candidate Genes

• Using both conventional and machine learning GWAS approaches, 238 mutation loci significantly associated with body measurement traits were identified.
• A total of 277 candidate genes near these loci were annotated.
• These genes are implicated in biological processes crucial for body measurements, including:
  • Skeletal development
  • Muscle formation
  • Cell growth
  • Energy metabolism
  • Protein synthesis

Genetic Differences Among Yili Horse Groups

• The findings indicate that genetic variations related to body measurement traits have already emerged between the speed-type and meat-type Yili horse groups.
• Such genetic differentiation at the genomic level suggests ongoing selection or adaptation related to their distinct breeding goals.

Advantages of Integrating Machine Learning with Conventional GWAS

• The combined approach improved the efficiency and sensitivity in detecting loci associated with complex traits compared to using traditional GWAS alone.
• Machine learning techniques like Lasso regression and Random Forest help to better handle high-dimensional genomic data and uncover subtle genetic effects.
• This integration provides new molecular evidence to dissect the genetic architecture underlying differences in body measurement traits among Yili horses.

Overall Significance

• The study enhances understanding of the genetic basis of important body measurement traits in Yili horses, which can inform future breeding and management strategies.
• It validates the potential of machine learning-assisted GWAS frameworks to advance genomic research in livestock species with complex traits.
• Ultimately, the research supports precision breeding efforts aimed at improving growth, production, and performance characteristics in horse populations.