To prioritize likely causal genes for each of 241 genomewide-significant CAD association signals, eight locus-based or similarity-based predictors of causal genes were considered:
- Genes harboring pathogenic (or likely pathogenic) variants responsible for monogenic disorders of cardiovascular relevance according to ClinVar
- Genes encoding proteins of established causal relevance to CAD per Mendelian randomization studies, or that are targets for established cardiovascular drugs;
- The two genes with the highest Polygenic Priority Score (PoPS) within 500kb of the sentinel variant;
- Genes with eQTLs in CAD-relevant tissues from STARNET or GTEx
- Genes that display cardiovascular-relevant phenotypes in knock-out mice from the International Mouse Phenotyping Consortium or the Mouse Genome Informatics database
- Genes containing protein-altering (i.e. missense or predicted loss-of-function) variants that are in strong LD (r2≥0.8) with the CAD sentinel variant;
- Genes containing rare coding variants that have been associated with CAD risk in previous whole-exome sequencing or array-based studies;
- The nearest gene to the CAD sentinel variant.
See the figure below for a summary. Genes were prioritized as "likely causal" for each sentinel association using a consensus-based approach, selecting the gene with the highest, unweighted sum (0-8) of evidence across all eight predictors. Most likely causal genes were assigned for 206 associations where there were at least two concordant predictors for a gene, prioritizing 185 distinct genes. 94 of these genes had three or more concordant predictors, so were considered "strongly prioritized".