Support Vector Machine based prediction models for drug repurposing and designing novel drugs for colorectal cancer

Abstract

Colorectal cancer (CRC) has witnessed a concerning increase in incidence and poses a significant therapeutic challenge due to its poor prognosis. There is a pressing demand to identify novel drug therapies to combat CRC. In this study, we addressed this need by utilizing the library of CRC pharmacological profiles of anticancer drugs and developed QSAR models for prediction of alternative and promiscuous anti-cancer compounds for CRC treatment. Our QSAR models demonstrated their robustness by achieving high correlation of determination (R2) after 10-fold cross validation. For 12 CRC cell lines, R2 ranges from 0.609-0.827. Highest performance was achieved for SW1417 and GP5d cell lines with R2, 0.827 and 0.786, respectively. Further, we listed the most common descriptors in the drug profiles of the CRC cell lines and we also checked the correlation of the descriptors with the drug activity. The KRFP314 fingerprint was the predominantly occurring descriptor, with the KRFPC314 fingerprint following closely in prevalence within the drug profiles of the CRC cell lines. Beyond predictive modeling, we also successfully validated drug-to-oncogene relationship via in-silico methods and identified FDA-approved drugs which could be used as potential anti-CRC drugs, paving the way for subsequent in vitro or in vivo experiments to validate their efficacy. To provide easy accessibility and utility of our research findings, we have incorporated these models into a free-to-use user-friendly prediction webserver named, "ColoRecPred" hosted on project.iith.ac.in/cgntlab/colorecpred. This web-based tool can be used to screen for potential anti-cancer compounds for CRC.