LIDR-TB: large language model-integrated platform for traceable drug repurposing in tuberculosis

Abstract

Tuberculosis (TB) remains a global challenge exacerbated by the ongoing emergence of drug-resistant strains. Although drug repurposing offers a cost-effective strategy to expedite therapeutic discovery, its application is impeded by the lack of unified, disease-specific frameworks that integrate fragmented chemical and biological data from heterogeneous sources. Herein, we developed LIDR-TB ( http://lidrtb.sysbio.org.cn ), a TB-focused, LLM-integrated informatics platform for traceable drug repurposing research. The platform integrates three core components: a curated knowledge base that consolidates heterogeneous chemical and biological evidence into a graph-structured representation; an interactive network visualization module for exploring multi-layered molecular and biological relationships; and a RAG-based question-answering model that anchors LLM outputs within the curated knowledge base. These components form a unified workflow that supports cross-module verification, allowing generative responses to be directly traced back to raw experimental records. Benchmarking on expert-curated evaluation suites showed promising results in semantic understanding and answer fidelity. Notably, we observed a potential "semantic compensatory effect", wherein the LLM leverages latent semantic context to partially offset structural parsing misalignments in some cases. This observation may contribute to maintaining answer quality in complex biomedical queries. Scientific contribution To address the lack of a unified, domain-specific informatics framework for integrating heterogeneous chemical and biological data, LIDR-TB provides a tuberculosis-focused platform for knowledge exploration, visualization, and evidence-grounded retrieval in drug repurposing. It integrates a graph-structured knowledge base, an interactive network visualization module, and a RAG-based question-answering model, enabling multi-level verification of generated responses through direct linkage to underlying experimental evidence. This openly accessible platform advances chemical-biological data integration and cheminformatics-driven knowledge discovery, facilitating systematic evidence exploration for TB and related infectious diseases in drug repurposing studies.