Identification of Repurposable Drugs for Colorectal Cancer Using Drug-Network-Based Classification Models

Abstract

Colorectal cancer (CRC) is the second most lethal cancer with more than one million new cases diagnosed worldwide every year. To defuse the increasing CRC threat, more effective and less harmful treatments for CRC patients are urgently needed. Computational drug repurposing, which is an in silico based approach to uncover new indications of approved drugs, is a promising strategy to accelerate the time to market of drugs. However, there are not many computational drug repurposing methods for CRC. In this work, we proposed drug-network-based classification models to identify repurposable drugs for CRC. Initially, four drug networks, the chemical structure network (CSN), the target protein network (TPN), the drug pathway network (PWN), and the drug-drug interaction network (DIN), were formulated. Based on the drug features properly extracted from the networks, we created four multi-layer perceptron (MLP) models. By comparing the performance of the models, the DIN model outperformed the others with the highest accuracy and an F₁ score of 96.9%. After predicting the repurposability of over 1,200 non-CRC approved drugs using the DIN model, 306 drugs discovered as potentially repurposable drugs for CRC. In summary, the drug-network-based classification models can efficiently identify repurposable drug candidates for CRC, which would be applicable for efficient therapeutic treatment of CRC.