DeepPGDB: A Novel Paradigm for AI-Guided Interactive Plant Genomic Database

Le résumé fourni par la source

DeepPGDB ( https://www.deeppgdb.chat ) is the first AI-driven plant genomics database designed to lower technical barriers in multi-omics research by enabling natural language-based data access and analysis. Integrating over 50 high-quality plant genomes, DeepPGDB combines fine-tuned large language models (LLMs) with prompt engineering to interpret user queries, generate standardized bioinformatics commands, and retrieve or visualize genomic data. Key functionalities include sequence retrieval, BLAST alignment, gene localization, expression profiling, and population genetics analysis, all presented via an intuitive conversational interface. A summarization module further enhances biological reasoning, inferring insights such as haplotype differentiation or protein properties. Benchmarking revealed that Deepseek-r1:14b optimized for short pre-prompts delivers high accuracy and speed. By bridging computational and biological expertise, DeepPGDB democratizes genomic research, fostering interdisciplinary collaboration and accelerating discoveries in agriculture and biotechnology.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.