Direct molecular formula prediction from tandem mass spectra with a multi-task-trained encoder and deterministic stoichiometric correction
- 1MoloGPI
- 2Dalian Customs District Technology Center
- 3Dongbei University of Finance and Economics
Abstract
Molecular formula assignment is the first step of metabolite identification from tandem mass spectra, yet deep-learning methods have repeatedly underperformed classical enumeration, reaching usable accuracy only through candidate generation and learned reranking. Here we present MStoic, a mass spectrum encoder that predicts molecular formulas directly and finalizes them with a deterministic, zero-parameter stoichiometric correction that selects a mass-consistent formula against the precursor mass (20 ppm; falling back to the closest-mass candidate when none qualifies). Every peak is embedded by soft Gaussian binning of log(m/z), Fourier features, pairwise-difference embeddings and an intensity embedding, and the encoder is trained with chemically grounded auxiliary tasks, including fragment-level supervision from computational fragmentation trees. On the multi-source Spectraverse benchmark (48,622 held-out spectra) MStoic achieves 77.8% top-1 exact-match accuracy, and 79.6% on the CASMI-2022 challenge, matching or exceeding specialized deep-learning, candidate-ranking, and rule-based baselines on every benchmark evaluated. Applying the same correction to a competing model shows that mass-based correction is a portable physical prior that amplifies rather than substitutes model capability. Factorial ablation shows that the hybrid encoding is selectively beneficial for rare-element-bearing formulas with information-dense spectra. And imposing mass conservation as a soft training loss reduces direct accuracy more than seven-fold, indicating that physical constraints in discrete chemical spaces belong in hard post-processing rather than in the training objective.
Related
- Circular RNA aptamers as superb sub-stoichiometric chaperones of TDP-43 condensation
- AlvaQSAR: A Web-Based Platform for Molecular Prediction of Environmental and Toxicological Endpoints Aligned with OECD Principles
- Machine Learning for Toxicity Prediction in Low-Sample Molecular Classes
- Mathematical Invariant-Enabled Topological Neural Networks for Molecular and Materials Property Prediction
- Sub-stoichiometric degradation is dispensable for potent PROTACs as demonstrated by irreversible covalent BTK degraders
BibTeX
@misc{zhang2026direct,
title = {{Direct molecular formula prediction from tandem mass spectra with a multi-task-trained encoder and deterministic stoichiometric correction}},
author = {Shang Zhang and Zhenlin Dong and Ming Gao},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15011102/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15011102/v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - Direct molecular formula prediction from tandem mass spectra with a multi-task-trained encoder and deterministic stoichiometric correction
AU - Shang Zhang
AU - Zhenlin Dong
AU - Ming Gao
PY - 2026
DA - 2026/10/27
PB - ChemRxiv
DO - 10.26434/chemrxiv.15011102/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15011102/v1
AB - Molecular formula assignment is the first step of metabolite identification from tandem mass spectra, yet deep-learning methods have repeatedly underperformed classical enumeration, reaching usable accuracy only through candidate generation and learned reranking. Here we present MStoic, a mass spectrum encoder that predicts molecular formulas directly and finalizes them with a deterministic, zero-parameter stoichiometric correction that selects a mass-consistent formula against the precursor mass (20 ppm; falling back to the closest-mass candidate when none qualifies). Every peak is embedded by soft Gaussian binning of log(m/z), Fourier features, pairwise-difference embeddings and an intensity embedding, and the encoder is trained with chemically grounded auxiliary tasks, including fragment-level supervision from computational fragmentation trees. On the multi-source Spectraverse benchmark (48,622 held-out spectra) MStoic achieves 77.8% top-1 exact-match accuracy, and 79.6% on the CASMI-2022 challenge, matching or exceeding specialized deep-learning, candidate-ranking, and rule-based baselines on every benchmark evaluated. Applying the same correction to a competing model shows that mass-based correction is a portable physical prior that amplifies rather than substitutes model capability. Factorial ablation shows that the hybrid encoding is selectively beneficial for rare-element-bearing formulas with information-dense spectra. And imposing mass conservation as a soft training loss reduces direct accuracy more than seven-fold, indicating that physical constraints in discrete chemical spaces belong in hard post-processing rather than in the training objective.
ER -