In animal breeding, there has been an increasing interest in investigating the added value of intermediate omics traits such as transcriptomes, metabolites and methylation patterns in genomic predictions. Such data are available only for small number of animals. The “singlestep genomic prediction” machinery, which was first proposed to combine pedigree information of a large number of individuals, and genomic information of a fraction of the population, can be useful to handle incomplete omics data. Such an approach, when applied to incomplete omics data scenarios, imply a simple linear relationship from genotypes to different omics traits, which in reality may be very complex. Little is known about the accuracy of genetic evaluations when the omics traits are generated for the whole population. Here, we present two different approaches to handle incomplete omics data, and investigate their impact on genomic predictions, using simulations.