Validating DTAPs with large language models: A novel approach to drug repurposing

In the face of escalating costs and lengthy timelines associated with traditional drug discovery, our study introduces a novel approach aimed at enhancing the drug repurposing process through the integration of computational models. Specifically, we explore the potential to enhance drug target affinity predictors (DTAPs)—tools that predict how well a drug binds to its target—by integrating them with advanced large language models (LLMs), such as GPT-4 and Llama-2-70b. We hypothesized that this synergy between DTAPs and LLMs would significantly improve the accuracy of identifying suitable drug-target interactions, a crucial step in repurposing existing drugs for new medical uses. Employing a rigorous comparative analysis, we tested the efficacy of traditional DTAPs against a specialized dataset focused on psychotropic drugs and their interactions with the sigma-1 receptor, an area ripe with repurposing opportunities. We then assessed how the integration of these DTAPs with LLMs affected prediction accuracy. The results showed a marked improvement in binary prediction accuracy, especially when DTAPs were combined with GPT-4. The implications of our findings are significant, suggesting that the fusion of DTAPs with LLMs could revolutionize the process of drug repurposing. This integrated approach offers a faster, more cost-effective pathway to drug development, streamlining the identification of new therapeutic applications for existing drugs. Our study not only validates the hypothesis of enhanced performance through integration of LLMs with DTAPs but also sets the stage for a new era in pharmacology, where the combination of advanced AI techniques can lead to breakthroughs in treatment discovery.