Test-Time Tuned Language Models Enable End-to-end De Novo Molecular Structure Generation from MS/MS Spectra
This addresses the challenge of identifying compounds absent from reference databases in fields like metabolomics and natural product discovery, representing a novel method for a known bottleneck.
The paper tackles the problem of identifying unknown compounds from tandem mass spectrometry data by introducing a test-time tuning framework that enables end-to-end de novo molecular structure generation, surpassing the state-of-the-art DiffMS by 100% on NPLIB1 and 20% on MassSpecGym benchmarks.
Tandem Mass Spectrometry enables the identification of unknown compounds in crucial fields such as metabolomics, natural product discovery and environmental analysis. However, current methods rely on database matching from previously observed molecules, or on multi-step pipelines that require intermediate fragment or fingerprint prediction. This makes finding the correct molecule highly challenging, particularly for compounds absent from reference databases. We introduce a framework that, by leveraging test-time tuning, enhances the learning of a pre-trained transformer model to address this gap, enabling end-to-end de novo molecular structure generation directly from the tandem mass spectra and molecular formulae, bypassing manual annotations and intermediate steps. We surpass the de-facto state-of-the-art approach DiffMS on two popular benchmarks NPLIB1 and MassSpecGym by 100% and 20%, respectively. Test-time tuning on experimental spectra allows the model to dynamically adapt to novel spectra, and the relative performance gain over conventional fine-tuning is of 62% on MassSpecGym. When predictions deviate from the ground truth, the generated molecular candidates remain structurally accurate, providing valuable guidance for human interpretation and more reliable identification.