Increasing the goodness-of-fit of genomic prediction model with addition of maternal genomic relationship matrix

Abstract

Genomic prediction models use a genomic relationship matrix (GRM) to quantify relationships. However, the GRM relationships do not distinguish between the maternal and paternal origins, thereby ignoring parent-of-origin effects such as imprinting. Genomic imprinting is a phenomenon whereby gene expressions within progeny are varied based on the parental origin of haplotypes or alleles, generally due to epigenetic effects such as DNA methylation. Genomic imprinting has been reported for many economically important traits in livestock such as weight. In this study, we explored the effect of fitting a maternal and/or paternal genomic relationship matrix (GRM), in addition to a regular GRM, on the goodness-of-fit of a genomic prediction model for 600 day weight by measuring the log-likelihood of restricted maximum likelihood (REML). The results showed the log-likelihood of the model was improved significantly when using the combination of regular GRM and maternal GRM simultaneously suggesting a better model. This result could be due to maternal imprinting, however further research is required to differentiate the maternal effects from parent-of-origin-dependent effects.