A Guided Network Propagation Approach to Identify Disease Genes that Combines Prior and New Information

Abstract

A major challenge in biomedical data science is to identify the causal genes underlying complex genetic diseases. Despite the massive influx of genome sequencing data, identifying disease-relevant genes remains difficult as individuals with the same disease may share very few, if any, genetic variants. Protein-protein interaction networks provide a means to tackle this heterogeneity, as genes causing the same disease tend to be proximal within networks. Previously, network propagation approaches have spread "signal" across the network from either known disease genes or genes that are newly putatively implicated in the disease (e.g., found to be mutated in exome studies or linked via genome-wide association studies). Here we introduce a general framework that considers both sources of data within a network context. Specifically, we use prior knowledge of disease-associated genes to guide random walks initiated from genes that are newly identified as perhaps disease-relevant. In large-scale testing across 24 cancer types, we demonstrate that our approach for integrating both prior and new information not only better identifies cancer driver genes than using either source of information alone but also readily outperforms other state-of-the-art network-based approaches. To demonstrate the versatility of our approach, we also apply it to genome-wide association data to identify genes functionally relevant for several complex diseases. Overall, our work suggests that guided network propagation approaches that utilize both prior and new data are a powerful means to identify disease genes.

Linking cells across single-cell modalities by synergistic matching of neighborhood structure

“…As shown in (Qi et al ., 2008), the amount of activation at equilibrium can be efficiently computed as , where b is the elementary unit vector with 1 for the nodes introducing the flow and 0 for the rest. We note that this diffusion kernel has been successfully utilized in variety of computational problems ranging from protein function prediction (Tsuda and Noble, 2004) to cancer gene identification (Hristov et al ., 2020). We use S U ( u i , u j ) as a measure of the similarity between cells u i and u j .…”

DriverSubNet: A Novel Algorithm for Identifying Cancer Driver Genes by Subnetwork Enrichment Analysis

“…To solve this situation, many algorithms have been developed to rank the candidate genes (Hou and Ma, 2014;Dinstag and Shamir, 2019;Hristov et al, 2020). For instance, HotNet2 identifies rare mutations across pathways and protein-protein interaction (PPI) networks using the heat-diffusion theory (Leiserson et al, 2015).…”

Prediction of cancer driver genes through network-based moment propagation of mutation scores

“…Both types of methods underlie the common paradigm that genes influencing the same phenotype interact within a network. Especially network propagation methods have shown success in identifying novel cancer driver genes ( Hristov et al , 2020 ; Leiserson et al , 2015 ; Reyna et al , 2018 ; Ruffalo et al , 2015 ; Vandin et al , 2011 , 2012 ). However, network propagation methods exploit by construction the flow of information between genes along paths, and the longer the paths are, the more information gets diluted.…”

“…Although the aforementioned methods showed great success in many biomedical applications ( Cowen et al , 2017 ), including the discovery of novel cancer genes, they approach the task of gene identification from an unsupervised perspective. However, there exists knowledge on well-established cancer genes (e.g Sondka et al , 2018 ), an important layer of additional information that has, to the best of our knowledge, only been leveraged in few methods for the prediction of cancer driver genes, namely Bayesian modeling ( Sanchez-Garcia et al , 2014 ) and unsupervised network propagation ( Hristov et al , 2020 ). In most cases, well-established cancer genes are only used to validate the importance and correctness of findings from new methods as a post-processing step.…”