“Unraveling Signaling Pathway Crosstalk Using LLMs”

Living cells do not process information through isolated, linear pathways. Instead, they rely on an intricate network of signaling crosstalk to make critical fate decisions. Deciphering this hidden cellular conversation remains one of the greatest challenges in targeted oncology and regenerative medicine. Yet traditional mechanistic models struggle to capture its complexity, while the scale of modern single-cell datasets exceeds what can be analyzed manually. My lab asks a simple question: Can large language models, designed to understand human language, be repurposed to decode the language of cellular signaling? In this talk, I will describe our efforts to develop LLMs as specialized, locally deployed biological reasoning engines. By grounding these models in both the biomedical literature and high-dimensional single-cell data, we show how they can identify key regulatory nodes where signaling pathways converge. Finally, I will discuss how integrating biological knowledge with empirical data enables prediction of nonlinear cellular responses, providing a scalable, mechanistically informed framework for discovering novel therapeutic combinations.