Factor analysis applied to genome prediction for high-dimensional phenotypes in pigs.

Area of Science:

• Animal Genetics
• Quantitative Genetics
• Bioinformatics

Background:

• Genome-wide selection (GWS) aims to improve livestock breeding by utilizing genomic information.
• Selecting for multiple traits simultaneously presents challenges in GWS due to complex genetic correlations.
• Factor analysis (FA) is a statistical technique for identifying underlying latent variables.

Purpose of the Study:

• To propose and evaluate the use of factor analysis (FA) for creating latent variables representing multiple pig traits.
• To assess the utility of these factors in genome-wide selection (GWS) studies.
• To compare the effectiveness of factor-based selection with traditional single-trait selection.

Main Methods:

• Utilized data from 345 F2 pigs from crosses between Brazilian Piau and commercial pigs.
• Genotyped pigs for 237 SNPs and recorded 41 different traits.
• Applied factor analysis (FA) to identify latent factors, followed by genomic selection models (Bayes A, Bayes B, RR-BLUP, Bayesian LASSO).

Main Results:

• FA successfully extracted four interpretable factors: "weight", "fat", "loin", and "performance".
• Accuracy of GWS using these factors was comparable to using individual traits.
• Selection of top individuals based on factors showed satisfactory overlap with selection based on individual traits.

Conclusions:

• Factor analysis (FA) provides a viable approach for simultaneous multi-trait selection in GWS.
• This method simplifies the selection process while maintaining selection accuracy.
• FA facilitates a more integrated understanding of genetic architectures for complex traits in pigs.