“Predicting ExWAS Findings from GWAS Data: a Shorter Path to Causal Genes”, Kevin Y. H. Liang, Yossi Farjoun, Vincenzo Forgetta, Yiheng Chen, Satoshi Yoshiji, Tianyuan Lu, J. Brent Richards2023-04-02 (rare mutations)⁠:

GWAS has identified thousands of loci associated with disease, yet the causal genes within these loci remain largely unknown. Identifying these causal genes would enable deeper understanding of the disease and assist in genetics-based drug development. Exome-wide association studies (ExWAS) are more expensive but can pinpoint causal genes offering high-yield drug targets, yet suffer from a high false-negative rate.

Several algorithms have been developed to prioritize genes at GWAS loci, such as the Effector Index (IE), Locus-2-Gene (L2G), Polygenic Prioritization score (PoPs), and Activity-by-Contact score (ABC) and it is not known if these algorithms can predict ExWAS findings from GWAS data. However, if this were the case, thousands of associated GWAS loci could potentially be resolved to causal genes.

Here, we quantified the performance of these algorithms by evaluating their ability to identify ExWAS statistically-significant genes for 9 traits. We found that Ei, 𝓁₂G, and PoPs can identify ExWAS statistically-significant genes with high areas under the precision recall curve (Ei: 0.52, 𝓁₂G: 0.37, PoPs: 0.18, ABC: 0.14). Furthermore, we found that for every unit increase in the normalized scores, there was an associated 1.3–4.6× increase in the odds of a gene reaching exome-wide statistical-significance (Ei: 4.6, 𝓁₂G: 2.5, PoPs: 2.1, ABC: 1.3).

Overall, we found that Ei, 𝓁₂G, and PoPs can anticipate ExWAS findings from widely available GWAS results. These techniques are therefore promising when well-powered ExWAS data are not readily available and can be used to anticipate ExWAS findings, allowing for prioritization of genes at GWAS loci.