Dataset of Duroc×Erhualian F2 pig population

Background:The development of multi-omics has increased the likelihood of further improving genomic prediction (GP) of complex traits. Gene expression data can directly reflect the genotype effect, and thus they are widely used for GP. Generally, the gene expression data are integrated into multiple random effect model as independent data layers or used to replace genotype data for genomic prediction. In this study, we integrated pedigree, genotype, and gene expression data into single-step method and investigated the effects of this integration on prediction accuracy.Results: The single-step method integrating genotype and gene expression data effectively improved genomic prediction accuracy of three complex traits in the Drosophila melanogaster genetic reference panel (DGRP) dataset. In addition, single-step method also improved the prediction accuracy of more than 90% of the 54 traits in Duroc×Erhualian F2pig population dataset. On average, the prediction accuracy of the single-step method integrating gene expression data was 27.0% and 9.5% higher than that of the pedigree-based best linear unbiased prediction (ABLUP) and genome-based best linear unbiased prediction (GBLUP), when the weighting factor (w) was set as 0, and it was 4.3% higher than that of the single-step best linear unbiased prediction (ssBLUP) under different wvalues.Conclusions:Overall, the analyses of two datasets confirmed that integration of gene expression data into single-step method could effectively improve genomic prediction accuracy. Our findings enrich the application of multi-omics data to genomic prediction and provide valuable reference for integrating multi-omics data into genetic evaluation model, which will contribute to genetic improvement.