Mechanism-grounded fragmentation trees enable explicit chemical reasoning for electron ionization mass spectrum prediction
5 affiliations
- 1School of Computer Science
- 2South China Normal University
- 3State Key Laboratory of Heavy Oil Processing, Petroleum Molecular Engineering Center (PMEC)
- 4China University of Petroleum
- 5College of Science
Abstract
Developing predictive models that connect molecular structures with measured spectra while preserving interpretable links to ion formation and fragmentation chemistry remains a central challenge in predictive spectroscopy. Here we present FragTreeMS, a mechanism-grounded framework that decomposes electron ionization (EI) mass spectrum prediction into fragmentation tree generation and repair followed by pathway- conditioned intensity learning. FragTreeMS organizes source-mechanism-product relations into traceable fragmentation trees and uses candidate-guided self-review to identify and recover plausible omissions before intensity prediction. The repaired fragmentation tree provides the pathway from the molecular root to each retained product ion, and these pathways are directly used for intensity prediction. FragTreeMS improves product ion recovery and complete spectrum similarity relative to direct spectrum prediction and established EI-MS baselines. The inferred ion formation processes and associated putative structural fragments are explicitly represented and visualized, advancing mass spectrum prediction toward interpretable chemical reasoning.
Related
- General Time-Dependent Surface Flux for Multiple Ionization and Fragmentation of Molecules. I: General Theory
- SKIER: A Scientific Skills Framework for Evidence-Grounded AI Reasoning
- Fragment-Grounded Neural Simulation of Electron Ionization Mass Spectra at Library Scale
- How Chemical Evidence Shapes Mechanistic Reasoning and Prediction by Large Language Models: Fine-Grained Inference in Nucleoside-Derived Supramolecular Gelation
- MSIonization: A Machine Learning Tool for Ionization Mode Prediction of Small Molecules.
BibTeX
@misc{sun2026mechanismgrounded,
title = {{Mechanism-grounded fragmentation trees enable explicit chemical reasoning for electron ionization mass spectrum prediction}},
author = {Hongyan Sun and Shuofan Li and Quan Shi and Jianxun Wu and Chunming Xu and Shuangyin Li},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15010026/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010026/v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Mechanism-grounded fragmentation trees enable explicit chemical reasoning for electron ionization mass spectrum prediction
AU - Hongyan Sun
AU - Shuofan Li
AU - Quan Shi
AU - Jianxun Wu
AU - Chunming Xu
AU - Shuangyin Li
PY - 2026
DA - 2026/10/07
PB - ChemRxiv
DO - 10.26434/chemrxiv.15010026/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15010026/v1
AB - Developing predictive models that connect molecular structures with measured spectra while preserving interpretable links to ion formation and fragmentation chemistry remains a central challenge in predictive spectroscopy. Here we present FragTreeMS, a mechanism-grounded framework that decomposes electron ionization (EI) mass spectrum prediction into fragmentation tree generation and repair followed by pathway- conditioned intensity learning. FragTreeMS organizes source-mechanism-product relations into traceable fragmentation trees and uses candidate-guided self-review to identify and recover plausible omissions before intensity prediction. The repaired fragmentation tree provides the pathway from the molecular root to each retained product ion, and these pathways are directly used for intensity prediction. FragTreeMS improves product ion recovery and complete spectrum similarity relative to direct spectrum prediction and established EI-MS baselines. The inferred ion formation processes and associated putative structural fragments are explicitly represented and visualized, advancing mass spectrum prediction toward interpretable chemical reasoning.
ER -