Network-informed gene ranking tackles genetic heterogeneity in exome-sequencing studies of monogenic disease

Genetic heterogeneity presents a significant challenge for the identification of monogenic disease genes. Whole exome sequencing generates a large number of candidate disease-causing variants and typical analyses rely on deleterious variants being observed in the same gene across several unrelated affected individuals. This is less likely to occur for genetically heterogeneous diseases, making more advanced analysis methods necessary. To address this need, we present HetRank, a flexible gene-ranking method which incorporates interaction network data. We first show that different genes underlying the same monogenic disease are frequently connected in protein interaction networks. This motivates the central premise of HetRank: that genes carrying potentially pathogenic variants and whose network neighbours do so in other affected individuals are strong candidates for follow-up study. By simulating 1,000 exome sequencing studies (20,000 exomes in total) we model varying degrees of genetic heterogeneity and show that HetRank consistently prioritises more disease-causing genes than existing analysis methods. We also demonstrate a proof-of-principle application of the method to prioritise genes causing Adams-Oliver syndrome, a genetically heterogeneous rare disease. An implementation of HetRank in R is available via the website http://sourceforge.net/p/hetrank/. This article is protected by copyright. All rights reserved.