A Systematic Framework for Analyzing Drug Mechanisms in the Era of Biomedical Big Data

Abstract

Biomedical big data explosion has caused a paradigm shift in the approach to drug research that had been based on hypothesis until the development of data intensive systemic analysis. Conventional reductionist methods do not generally explain the polypharmacological character of drugs and complexity of interrelation of multifactorial diseases. The present paper suggests an integrative computational model that deciphers the drug mechanisms of actions based on building hierarchical tripartite drug-gene-disease networks. The framework uses a more or less the topographic measure of the therapeutic influence in form of a propagation matrix and network topology features, including the degree centrality and shortest path distance, to determine pivotal regulatory hubs, and mechanistic proximity. A case simulation study with simulated interactions illustrates how the framework has been able to prioritize drug to disease interactions as well as identifies necessary bridge genes that mediate systemic effects. Additionally, we cover the possibility of incorporating this model into AI-based drug discovery pipelines, in particular, in property prediction and generative design sub- modules. Although the existing restrictions are that it depends on static and linear modeling, the framework offers a scalable and explainable basis of systems pharmacology. The data in the future will combine dynamic multi-omics and non-linear deep learning architecture to make drug mechanism studies more interpretable and predictive.