Accuracy of single-step genomic selection for egg production in yellow-feathered broilers using a random regression...

Egg production is a crucial economic trait in poultry. Conventional cumulative yield analyses (e.g., single-trait models, STM) overlook dynamic environmental and physiological influences, while multi-trait models (MTM) partition production into discrete stages but fail to capture continuous temporal patterns. Random regression models (RRM) enable longitudinal data analysis to dissect time-dependent genetic effects, offering flexibility in predicting breeding values at any production stage. This study compared the accuracy of single-step genomic selection (ssGBLUP) combined with RRM for egg production in yellow-feathered broilers. It utilized RRM, MTM, and STM based on 168,596 records from 18,221 chickens. Of these, 17,619 animals had genotypic data, and pedigree information was available for 78,449 animals. RRM analyzed biweekly production (12 time points from 20 to 43 weeks) via Legendre polynomials; MTM evaluated three laying phases (20-26, 27-34, 35-43 weeks); STM used cumulative yield (20-43 weeks). The optimal RRM orders were fixed regression (2nd), additive genetic effects (1st), and permanent environmental effects (4th). RRM-derived heritability (0.01-0.19) showed an increase-decrease-increase trend, contrasting with MTM phase heritabilities (0.128, 0.267, 0.300) and STM heritability (0.329). The integration of single-step genomic selection with RRM significantly outperformed MTM and STM in terms of accuracy and unbiasedness of estimated breeding values (EBV). Specifically, RRM improved accuracy by 3 % (pedigree) and 2.9 % (genomic) over MTM, and 7 % (pedigree) and 6.8 % (genomic) over STM; unbiasedness increased by 26 % (pedigree) and 25.6 % (genomic) over MTM, and 14.5 % (pedigree) and 12 % (genomic) over STM. Notably, combining RRM with single-step genomic methods enhanced EBV accuracy by up to 84 % and unbiasedness by 26 % compared to pedigree-based models. This study validates that the integration of RRM with ssGBLUP presents a promising avenue for enhancing genomic selection programs aimed at improving longitudinal production traits like egg number in poultry breeding.