Test-Time Tuned Language Models Enable End-to-end De Novo Molecular Structure Generation from MS/MS Spectra

Laura Mismetti, Marvin Alberts, Andreas Krause, Mara Graziani · 2025

Read Test-Time Tuned Language Models Enable End-to-end De Novo Molecular Structure Generation from MS/MS Spectra by Laura Mismetti, Marvin Alberts, Andreas Krause, Mara Graziani (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-23746.txt.

Description

Laura Mismetti and three co-authors build a system that reads a molecule's structure straight from its tandem mass spectrum. Mass spectrometry is a workhorse for identifying unknown small molecules in metabolomics, natural products, and environmental work. Yet the shattering process is probabilistic, chemical space is vast, and today's tools mostly match against databases of known spectra or chain several costly steps.

Their transformer framework takes the spectrum plus the molecular formula and generates candidate structures end to end, with no hand-built annotations between. Pretraining on simulated spectra supplies transfer learning, and a test-time tuning step adapts the model to fresh experimental data on the spot. On two benchmarks it beat conventional fine-tuning clearly, and even wrong answers stayed chemically close to the truth, scoring high on structural similarity. The authors pitch the tool as guidance for experts facing spectra nobody has seen before.

Catalog record

CREATORLaura Mismetti, Marvin Alberts, Andreas Krause, Mara Graziani
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-23746

Browse the full library · About the Tower of Babel