Leveraging supervised learning for functionally informed fine-mapping of cis-eQTLs identifies an additional 20,913
Qingbo S Wang1,2,3, David R Kelley4, Jacob Ulirsch5,6,7
1Broad Institute of MIT and Harvard, Cambridge, MA, USA. qingbow@broadinstitute.org.
Nature Communications
|June 8, 2021
Summary
We developed a new method to predict how genetic variants affect gene expression, creating a valuable resource for understanding non-coding variants and their role in human diseases.
Area of Science:
- Genetics
- Genomics
- Bioinformatics
Background:
- Genome-Wide Association Studies (GWAS) identify numerous non-coding variants.
- Characterizing non-coding variant function is crucial but challenging.
- Existing high-throughput methods lack sufficient training data.
Purpose of the Study:
- To develop a high-throughput computational predictor for non-coding variant function.
- To create a large gold standard dataset for training and validation.
- To improve the identification of causal variants and associated genes.
Main Methods:
- Leveraged 14,807 putative causal expression quantitative trait loci (eQTLs) from statistical fine-mapping.
- Utilized 6121 features to train a predictor for gene expression modification.
- Developed the Expression Modifier Score (EMS) for variant prioritization.
Main Results:
- Validated EMS against existing scores for prioritizing functional variants.
- Used EMS to identify an additional 20,913 putatively causal eQTLs.
- Incorporated EMS into co-localization analysis to identify 310 new candidate genes for UK Biobank phenotypes.
Conclusions:
- EMS is a robust predictor of variant effects on gene expression.
- EMS enhances the discovery of causal variants and disease-associated genes.
- This work provides a valuable tool for functional genomics and precision medicine.
