lmogil / divpop

DivPop

This study used the MESA cohort individuals with genotype and expression data to generate cis-eQTLs and multi-ethnic predictors of gene expression and show that training set with ancestry similar to the test set is better at predicting gene expression in test populations, emphasizing the need for diverse population sampling in genomics.

MESA models are avaliable for download at PredictDB for use with PrediXcan and S-Predixcan.

• On the PredictDB page, click on contributed/. These models are filtered by zscore_pval < 0.05 and rho_avg > 0.1, see here for explanation of performance statistics.
• Unfiltered models are in the unfiltered_dbs/ directory above

cis-eQTL summary statistics can be found here.

For more details, see our BioRxiv preprint:

Genetic architecture of gene expression traits across diverse populations

Lauren S Mogil, Angela Andaleon, Alexa Badalamenti, Scott P Dickinson, Xiuqing Guo, Jerome I Rotter, W. Craig Johnson, Hae Kyung Im, Yongmei Liu, Heather E Wheeler