Drug repurposing of liver cancer based on multimodal neural network and virtual screening.

Abstract

Liver cancer is a malignant tumor with high incidence and mortality rates globally. Existing therapeutic methods have certain limitations in terms of efficacy, safety, and applicability, creating an urgent need for the development of new treatment strategies. Drug repurposing, which explores new indications for already approved drugs, offers significant advantages in reducing research and development costs and shortening development timelines. This study utilized a multimodal neural network model to integrate molecular fingerprints and molecular graphs to construct a drug-target interaction prediction system for liver cancer targets. A large-scale virtual screening of The US Food and Drug Administration-approved drugs was conducted, followed by molecular docking and molecular dynamics simulations to analyze binding stability. The screening identified 17 potential antiliver cancer drugs. Further verification of their in-vitro activity was performed through 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) and wound healing assays in MHCC97-H, HepG2, and Huh7 cell lines, revealing that Manidipine and Vildagliptin exhibited significant cytotoxicity and migration inhibition. These results demonstrate the feasibility of the proposed screening workflow and provide a set of candidate molecules for further mechanistic investigation in liver cancer drug repurposing.