A Propagation-based Algorithm for Inferring Gene-Disease Associations

A fundamental challenge in human health is the identification of diseasecausing genes. Recently, several studies have tackled this challenge via a two-step approach: first, a linkage interval is inferred from population studies; second, a computational approach is used to prioritize genes within this interval. State-of-the-art methods for the latter task are based on the observation that genes causing the same or similar diseases tend to lie close to one another in a network of protein-protein or functional interactions. However, most of these approaches use only local network information in the inference process. Here we provide a global, network-based method for prioritizing disease genes. The method is based on formulating constraints on the prioritization function that relate to its smoothness over the network and usage of prior information. A propagation-based method is used to compute a function satisfying the constraints. We test our method on gene-disease association data in a cross-validation setting, and compare it to extant prioritization approaches. We show that our method provides the best overall performance, ranking the true causal gene first for 29% of the 1,369 diseases with a known gene in the OMIM knowledgebase.