SpliPath enhances disease gene discovery in case-control analyses of rare splice-altering genetic variants.

Abstract

We developed SpliPath as a generalizable framework to discover disease associations mediated by rare variants that induce experimentally supported mRNA splicing defects. Our approach integrates components of burden tests (BTs), traditional splicing quantitative trait locus (sQTL) analyses, and sequence-to-function AI models (SpliceAI and Pangolin). Central to the workings of SpliPath is our concept of collapsed rare variant splicing QTL (crsQTL). crsQTL groups rare variants that are predicted to alter splicing in the same way, specifically by linking them to shared splice junctions observed in independent (unpaired) RNA sequencing (RNA-seq) datasets. We demonstrate the utility of SpliPath through applications in amyotrophic lateral sclerosis (ALS). Through this, we showcase scenarios where SpliPath detects genetic associations that cannot be recovered by more simplistic combinations of BT and SpliceAI. We also nominate crsQTL for splice defects detected in large-scale analyses of ALS patient tissue.