Clinical prediction of pathogenic variants in non-coding regions of the human genome.

Abstract

Whole genome sequencing has become a wide-spread diagnostic tool for rare diseases patients. This broadens analyses to non-coding regions of the genome showing strong evidence of clinical significance for human Mendelian diseases. Notwithstanding its importance, current in-silico prediction tools are restricted to coding sequences which limits its applicability. Additionally, lack of power in discriminating variants of uncertain significance (VUS) limits its clinical utility. Here we present PANCO, a genome-wide pathogenicity prediction tool aiming at reclassification of VUS with a rigorous imputation workflow adapted for non-coding variants. PANCO integrates functional, evolutionary and population frequency information to capture emerging biological signals correctly reclassifying VUS. Importantly, PANCO shows remarkable power in an external validation set, on VUS (AUROC=0.99 and AUROC=0.89, respectively).