Optimization of Genomic Breeding Value Estimation Model for Abdominal Fat Traits Based on Machine Learning.

Area of Science:

• Animal Genetics
• Quantitative Genetics
• Machine Learning in Animal Breeding

Background:

• Abdominal fat in chickens significantly impacts meat quality and feed efficiency.
• Breeding for reduced abdominal fat is crucial for economic viability in poultry production.
• Genomic selection (GS) offers precise and early selection for complex traits like abdominal fat.

Purpose of the Study:

• To develop an advanced genomic prediction framework for chicken abdominal fat.
• To identify and utilize key genetic markers for abdominal fat deposition.
• To enhance the accuracy of genomic estimated breeding values (GEBVs) prediction.

Main Methods:

• Combined genome-wide association studies (GWAS) and linkage disequilibrium (LD) for SNP identification.
• Employed a two-stage machine learning feature selection (Lasso and RFE).
• Developed and validated a Dynamic Adaptive Weighted Stacking Ensemble Learning Framework (DAWSELF) with Ridge as meta-learner.

Main Results:

• Identified relevant single-nucleotide polymorphisms (SNPs) for abdominal fat prediction.
• Linear and nonlinear models showed high accuracy as base learners.
• DAWSELF consistently outperformed individual models and traditional stacking in prediction accuracy across three populations.

Conclusions:

• DAWSELF provides an efficient framework for GEBV prediction in complex traits like chicken abdominal fat.
• The study offers a reusable SNP feature selection strategy for poultry breeding.
• This approach enhances breeding precision and improves chicken meat product quality.