Large Language Models in Drug Discovery: A Comprehensive Analysis of Drug-Target Interaction Prediction

Large Language Models have successfully caught the eyes of the pharmaceutical industry and provided aid in several tasks, such as de novo drug design, drug repurposing, molecular optimization, activity prediction, and Drug-Target Interaction (DTI) prediction. In this survey, we delved deep into these models, among their various applications, techniques, performance metrics, and the datasets they utilized. Additionally, we gave special attention to DTI predictive models due to the crucial rule of DTI prediction in the Drug Discovery (DD) process, which aims to enhance the efficiency of finding suitable drug-like candidates while lowering time and costs. Furthermore, we presented some of the successful models that have reached the market and showed some tangible success. While aiming to highlight the potential of LLMs and acknowledging their challenges that need to be overcome for broader adoption, we focused on some key areas, such as their clinical and financial aspects, generalizability, computational cost, interpretability, and hybrid models. Moreover, eXplainable Artificial Intelligence (XAI) was emphasized and we underscored its potential and need in the field of DD to further validate the trust-worthiness of such models.