Automating AI Discovery for Biomedicine Through Knowledge Graphs And LLM Agents

The biomedical domain’s accelerating progress in understanding, early detection, and treatment of diseases has led to an exponentially growing and overwhelming body of literature. Researchers typically rely on this literature to find information relevant to their discoveries, using it to validate their findings. And despite the emergence of LLM systems, retrieving precise, relevant facts from the literature remains a significant challenge. The primary objective of this paper is to expedite biomedical discoveries by the integration of a biomedical knowledge graph with the scientific literature embedded in a Large Language Model. This work introduces a framework that employs a novel strategy to discover the most relevant pathway between any two biomedical entities in a knowledge graph, which is then leveraged by a multi-agent AI system to extract supporting facts from literature and design AI predictors for understanding the discovered pathway. Experiments with various biomedical entity pairs demonstrate the framework’s ability to identify highly relevant pathways and design plausible and complex AI predictors across diverse therapeutic areas. In addition, we developed Intelliscope, a web-based dashboard that makes this framework available to researchers worldwide. This first-of-its-kind platform could significantly accelerate scientific discoveries, potentially leading to breakthroughs in disease understanding, drug repurposing, and therapeutic development.