The increasing volume of genomic literature presents significant challenges in the efficient retrieval and synthesis of genomic variant-specific information, which is critical for clinical and research applications. To address these challenges, we introduce VarChat, a novel generative AI-based platform designed to optimize the retrieval and summarization of up-to-date literature directly citing genomic variants of interest. VarChat uses different literature databases within a hybrid RAG framework that integrates both sparse and dense retrieval approaches and synthesizes these fragmented literature findings into concise, scientifically rigorous summaries while providing direct access to high-confidence references. The platform also integrates a guided chat feature, enabling users to explore specific questions related to the functional and clinical implications of genomic variants. VarChat integrates literature preprocessing and advanced hybrid retrieval techniques to accelerate the genomic variant interpretation process, supporting the genomic community in improving clinical decision. Availability: VarChat is freely available at varchat.engenome.com.
Generative AI Meets Genomics: VarChat, a RAG-Based Approach for Literature-Driven Variant Summarization
De Paoli F.;Berardelli S.;Tudisco A.;Blindu A.;Parimbelli E.;Zucca S.
2025-01-01
Abstract
The increasing volume of genomic literature presents significant challenges in the efficient retrieval and synthesis of genomic variant-specific information, which is critical for clinical and research applications. To address these challenges, we introduce VarChat, a novel generative AI-based platform designed to optimize the retrieval and summarization of up-to-date literature directly citing genomic variants of interest. VarChat uses different literature databases within a hybrid RAG framework that integrates both sparse and dense retrieval approaches and synthesizes these fragmented literature findings into concise, scientifically rigorous summaries while providing direct access to high-confidence references. The platform also integrates a guided chat feature, enabling users to explore specific questions related to the functional and clinical implications of genomic variants. VarChat integrates literature preprocessing and advanced hybrid retrieval techniques to accelerate the genomic variant interpretation process, supporting the genomic community in improving clinical decision. Availability: VarChat is freely available at varchat.engenome.com.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


