Accelerating natural product discovery with linked MS-genomics and language/transformer-based models

Abstract Integrated chem-bio characterization of microbial strain libraries can streamline natural product discovery by prioritizing candidate producers. Here, we employ language- and transformer-based models to extract actionable insights from linked mass spectrometry (MS)-genome datasets. Our framework enables ranking of microbial producers to prioritise high-potential candidates for targeted validation. Across three representative case studies, this approach prioritized producers of diverse natural products with 75–100% precision. These findings demonstrate the transformative potential of AI-enabled chem-bio characterization to significantly accelerate natural product discovery and enable access to microbial chemical diversity beyond reference knowledge.