Measurement of Disease Comorbidity Using Semantic Profiling of Disease Genes

TL;DR

This paper introduces a new method to measure disease comorbidity by analyzing similarities in gene sets, even when diseases share no common genes.

Contribution

The novel contribution is GS.CoMoD, a gene-set-based approach to detect comorbidity without requiring overlapping genes.

Findings

GS.CoMoD achieved higher scores for comorbid disease pairs compared to random pairs in simulations.

GS.CoMoD outperformed existing methods in detecting disease comorbidity.

Comorbidity can be inferred from similarities in gene-set enrichment p-values even without shared genes.

Abstract

The identification of overlapping disease genes between different diseases is the first step in the elucidation of the biological mechanism of disease comorbidity; however, in the absence of common genes, it is difficult to determine the mechanism of comorbidity even if clinical evidence of disease co-occurrence exists. In this research, a gene-set-based measurement of the comorbidity of diseases (GS.CoMoD) was proposed. The underlying assumption of GS.CoMoD is that if the p-value vectors obtained from the enrichment analyses of different disease gene lists indicate similarity, the diseases are possibly comorbid. Therefore, comorbidity can be detected even without overlapping genes. A simulation analysis showed that GS.CoMoD yielded higher scores for comorbid disease pairs vs. random disease pairs. Moreover, comparison analyses revealed that GS.CoMoD outperformed the pre-existing methods for detecting comorbidity.

Figures