Disease Treatment Key Genes Identification In Heterogeneous Interaction Networks

Systematically, network pharmacology based on homogeneous interaction network of encoding genes has effectively elucidated treatment effects, multiple actions, and associated side effects of drugs by quantifying the proximity between disease modules and drug modules. However, further exploration is required to understand the contributions of heterogeneous and multi-scale factors, including non-coding genes and GO Terms, to proximity. This study aims to integrate a heterogeneous interaction network(HIN) consisting of encoding genes, non-coding genes, and GO Terms. We find that incorporating non-coding genes and GO Terms significantly enhances the proximity between drug modules and the diseases modules they address. Additionally, we propose an index for quantifying crucial nodes based on network proximity, enabling the identification of treatment key coding genes, non-coding genes, and GO terms (T
key
) involved in drug effect processes. In a systematic validation of 712 drug-disease pairs encompassing 471 drugs and 35 diseases, we effectively identify heterogeneous and multi-scale T
key
factors that play crucial roles in elucidating the mechanisms underlying drug treatment for various diseases. As a case study, we present T
key
and elucidate the pivotal biological pathways involved in Theophylline treatment for Asthma, multiple Theophylline actions of Theophylline for Pulmouary Emphysema, and the pathways leading to the side effect of Hypotension induced by Theophylline. In conclusion, heterogeneous and multi-scale T
key
are imperative for mediating drug effects in interaction networks, thus treating diseases or causing side effects. Our research presents a feasible computational tool for understanding how drugs treat diseases in drug-disease relationship analysis.