Selinger, D. W.; Wall, T. R.; Stylianou, E.; Khalil, E. M.; Gaetz, J.; Levy, O.

The exponential increase in biomedical data offers unprecedented opportunities for drug discovery, yet overwhelms traditional data analysis methods, limiting the pace of new drug development. Here we introduce a framework for autonomous artificial intelligence (AI)-driven drug discovery that integrates knowledge graphs with large language models (LLMs). It is capable of planning and carrying out automated drug discovery programs while providing details of its research strategy, progress, and supporting data points, enabling a thorough assessment of its methods and findings. At the heart of this framework lies the focal graph - a novel construct that harnesses centrality algorithms to distill vast, noisy datasets into concise, transparent, data-driven hypotheses. By enabling high-throughput search and automated result interpretation, such a framework could be used to execute massive numbers of searches, identify patterns across complex, diverse datasets, and prioritize actionable hypotheses at a scale and speed unachievable by human researchers alone. We demonstrate that even small-scale applications of this approach can yield novel, transparent insights relevant to multiple stages of the drug discovery process and present a prototype system capable of autonomously planning and executing a multi-step target discovery workflow. The focal graph framework described here, and the automation it enables, represents a promising path forward: towards a deeper understanding of the mechanisms underlying disease and a true acceleration in the development of novel therapeutics.