Genolator enables protein function interpretation using a multimodal large language model fusing genomic and structural interpretation with natural language interaction.

Genolator, a multimodal large language model, achieves high accuracy in answering protein function queries, outperforming GPT 4.1 and smaller domain-specific models.

• Why it matters: Decoding the genome's complex language is crucial for understanding disease mechanisms and developing targeted treatments, but current models struggle to integrate genomic and structural data with natural language.
• What they did: The authors fine-tuned Genolator on over 365,000 question-answer pairs using Gene-Ontology terms, integrating DNA, amino acid sequences, and protein structures with natural language queries.
• The result: Genolator enables natural language interaction with protein data, improves biological discovery and clinical research, and demonstrates benefits of a multimodal approach through hidden state organization and attention head analysis.