eGoT: enhanced graph-of-thoughts for multi-hop knowledge retrieval and hypothesis generation in biomedicine

Abstract

Biomedical research increasingly requires synthesizing knowledge fragmented across thousands of publications to understand complex disease mechanisms and generate therapeutic hypotheses. In this work, we present eGoT, an end-to-end framework that addresses this challenge by constructing knowledge graphs directly from unstructured literature and enabling iterative multi-hop reasoning over the resulting graphs. Unlike prior approaches that require pre-existing curated knowledge bases or rely on text retrieval only, eGoT provides a complete pipeline from primary literature to knowledge graph cons