A Network Integration Approach for Drug-Target Interaction Prediction and Computational Drug Repositioning from Heterogeneous Information

Abstract

The emergence of large-scale genomic, chemical and pharmacological data provides new opportunities for drug discovery and repositioning. In this work, we develop a computational pipeline, called DTINet, to predict novel drug-target interactions from a constructed heterogeneous network, which integrates diverse drug-related information. DTINet focuses on learning a low-dimensional vector representation of features, which accurately explains the topological properties of individual nodes in the heterogeneous network, and then makes prediction based on these representations via a vector space projection scheme. DTINet achieves substantial performance improvement over other state-of-the-art methods for drug-target interaction prediction. Moreover, we experimentally validate the novel interactions between three drugs and the cyclooxygenase proteins predicted by DTINet, and demonstrate the new potential applications of these identified cyclooxygenase inhibitors in preventing inflammatory diseases. These results indicate that DTINet can provide a practically useful tool for integrating heterogeneous information to predict new drug-target interactions and repurpose existing drugs.

A data-driven drug repositioning framework discovered a potential therapeutic agent targeting COVID-19

“…The initial list of drug candidates targeting SARS-CoV-2 was first screened using a network-based knowledge mining algorithm modified from our previous work [68,69].…”

NetMix: A network-structured mixture model for reduced-bias estimation of altered subnetworks

“…A standard paradigm in computational biology is to use interaction networks as prior knowledge in the analysis of high-throughput 'omics data, with applications in protein function prediction [79,73,65,25,18], gene expression [32,91,16,48,27], germline variants [55,12,56,43,45], somatic variants in cancer [66,87,57,84,64,42], and other data [39,10,20,89,35,77,13,60]. One classic approach is to identify active, or altered, subnetworks of an interaction network that contain outlier measurements.…”

Leveraging Distributed Biomedical Knowledge Sources to Discover Novel Uses for Known Drugs

“…More recently, Luo et al [67] constructed a computational pipeline for predicting drug-target interactions called DTINet which similarly uses a heterogeneous set of networks of biologically significant entities containing some 12K nodes and 1.9M edges, built from known interactions between drugs, proteins, diseases and side-effects. Similar to the method presented in this manuscript, low-dimensional feature vectors were generated from this network via random walk with restart, which approximates the network diffusion.…”

“…While machine learning approaches to the drug repurposing problem have been proposed [16,17,20,21,22,56,67], these methods typically use only a few sources of information, and typically only a few characteristics of a given drug and disease, such as genomic information, chemical structure information, and/or drug interaction information [65]. More recently, integrative approaches that combine information from multiple data sources have been proposed which were shown to outperform non-integrative approaches [75,101].…”