# Network-based identification of disease genes in expression data: the GeneSurrounder method

## TL;DR

GeneSurrounder is a novel network-based method that integrates gene expression data with pathway information to identify disease-driving genes more accurately than traditional approaches.

## Contribution

It introduces a two-step procedure combining local gene influence and differential expression to pinpoint mechanistically relevant disease genes.

## Key findings

- Identifies biologically relevant disease genes in ovarian cancer datasets.
- Detects genes missed by single-gene association tests.
- Provides more consistent results across multiple studies.

## Abstract

The advent of high--throughput transcription profiling technologies has enabled identification of genes and pathways associated with disease, providing new avenues for precision medicine. A key challenge is to analyze this data in the context of the regulatory networks and pathways that control cellular processes, while still obtaining insights that can be used to design new diagnostic and therapeutic interventions. While classical differential expression analysis provides specific and hence targetable gene-level insights, it does not include any systems-level information. On the other hand, pathway analyses integrate systems-level information with expression data, but are often limited in their ability to indicate specific molecular targets. We introduce GeneSurrounder, an analysis method that takes into account the complex structure of interaction networks to identify specific genes that disrupt pathway activity in a disease-specific manner. GeneSurrounder integrates transcriptomic data and pathway network information in a novel two-step procedure to detect genes that (i) appear to influence the expression of other genes local to it in the network and (ii) are part of a subnetwork of differentially expressed genes. Combined, this evidence can be used to pinpoint specific genes that have a mechanistic role in the phenotype of interest. Applying GeneSurrounder to three distinct ovarian cancer studies using a global KEGG network, we show that our method is able to identify biologically relevant genes and genes missed by single-gene association tests, integrate pathway and expression data, and yield more consistent results across multiple studies of the same phenotype than competing methods.