Integrated Analysis of SNPs and Structural Variations via High-depth Whole-genome Sequencing Reveals the Genetic Architecture and Optimizes Genomic Prediction in Chickens

Abstract

The characterization of genetic architecture and the optimization of genomic prediction are pivotal for the genetic improvement of complex traits in poultry. In this study, we investigated the genetic basis of 15 growth and carcass traits in an F2 chicken population (n = 877) using high-depth whole-genome sequencing with an average coverage of 31.2 × . By implementing an ensemble strategy involving four independent callers, we identified 35,924 high-confidence structural variations (SVs), with deletions being the most prevalent type. Combining SNPs and SVs enhanced genomic heritability for 14 out of 15 traits compared to SNPs alone. SNP-based GWAS corroborated well-known genes, including the prominent QTL cluster on chromosome 1, the NCAPG-LCORL locus on chromosome 4, and IGF2BP1 on chromosome 27. Notably, SV-based analysis unveiled additional candidate genes, such as ZNF385D, MYH10, and MOB1B. To gain functional insights, eQTL-GWAS colocalization analysis integrating SNP-based GWAS signals with tissue-specific eQTL data identified significant colocalization signals for ITM2B in brain tissue, potentially implicating excitatory synaptic transmission, and TRIM13 in blood, potentially implicating inflammatory and immune regulation. To optimize genomic breeding value estimation through the effective utilization of multi-type markers, we developed GPDLBP, a hybrid deep learning framework that integrates locally connected networks to capture SV effects with the GBLUP model for SNP effects. Compared with the traditional SNP-only model, GPDLBP improved prediction accuracy for most traits, with gains exceeding 2% for BW21 (body weight at 21 days of age), BW49 (body weight at 49 days of age), EW (eviscerated weight), LMW (leg muscle weight), and AFW (abdominal fat weight); for example, prediction accuracy increased from 0.512 to 0.532 for BW49 and from 0.471 to 0.491 for EW. These findings show that SVs complement SNPs in both genetic dissection and genomic prediction of economically important traits in chickens. The integration of multiple variant types provides a practical strategy for accelerating precision breeding in high-depth sequencing-based poultry programs.