AI-Integrated Drug–Target Interaction Prediction for High-Affinity Kinase Inhibitors Using Molecular Modeling and Virtual Screening

Abstract

In the field of drug discovery, artificial intelligence (AI) has proven to be a gamechanger by enhancing the prediction of drug–target interaction (DTI) and accelerating discovery of potential therapeutic candidates. This research aimed to build an AI-based computational system to model and virtually screen drugs to predict drug–target interaction and discover high affinity kinase inhibitors. To perform the in-silico analysis, a secondary dataset of 1,000 drug–target interaction records was created, validated for the presence of both compound SMILES and protein target sequences, and binding affinity values were obtained from experiments. Following data preprocessing, molecular descriptors and protein sequence features were utilized to develop an AI-based regression model for binding affinity prediction. The compounds were then prioritized using a virtual screening workflow with the predicted affinity values. Molecular diversity of curated dataset consisted of 479 unique compounds and 326 protein targets in the range of 5.00 to 15.20. A significant percentage of the interactions had high binding affinity which enabled good model development and prioritization of candidates. The suggested framework was effective in pinpointing compounds with improved predicted affinity, underscoring the potential of AI-driven computational strategies in speeding up the process of identifying compounds that target kinases. In sum, the combination of AI and virtual screening with molecular modeling offers a robust method for rational lead identification and lays the groundwork for computational studies in the future for structure validation and experimental exploration.