Genetically predicted gene expression

cis-eQTLs can be used to predict gene expression levels from genotypes. These values estimate log-expression relative to the scenario of all reference alleles, rather than absolute expression levels, but can be used to:

• Stratify individuals by a gene's genetically predicted expression to test hypotheses about the gene's role in observed traits.
• Test whether observed gene expression differences in groups with divergent traits are only a downstream effect of exposures or whether they have a more direct genetic component. See Zhou et al. 2023

We provide predicted expression for 12,225 genes for all rats genotyped by our NIDA center (Round 10.5, N=20,484). These are predictions for average expression in the brain, using cis-eQTLs for whole brain hemisphere tissue, and they include only genes that have one or more significant eQTLs and positive correlation between predicted and observed expression in the training data.

Download predicted expression (91 MB)

The table below gives accuracy of the predictions in terms of correlation between predicted and observed expression levels. This is assessed using the same whole brain hemisphere dataset that was used for the cis-eQTL models, so it is the best-case accuracy. The r² values represent the cis-genetic proportion of expression variance. Genes with blank r/r² values did not have sufficient cis-regulatory variation to accurately predict expression levels.

Download TSV Loading…