From comorbidities to gene expression fingerprints and back

Abstract

Epidemiological evidence shows that some diseases tend to co-occur more than expected by chance and that patientspecific trends are observed. However, the molecular processes underlying these phenomena remain unclear. Here we exploit the accumulating RNA-seq data on human diseases to calculate disease similarities at the transcriptomic level. We build a disease similarity network that significantly captures almost half of the medically known comorbidities, substantially outperforming previously published methods and providing biological explanations for such co-occurrences. Additionally, we group patients from a given disease with a similar expression profile into meta-patients and calculate their molecular similarities with the analyzed diseases, highlighting the need to study disease comorbidities within a personalized medicine scope. Finally, we provide a web application in which the networks and their underlying molecular mechanisms can be easily inspected.