RxnNet: An AI Framework for Reaction Mechanism Discovery─A Case Study in Mass Spectrometry
Understanding complex multistep reaction mechanisms remains a major challenge, particularly in the interpretation of tandem mass spectrometry (MS/MS) data, where fragmentation pathways often involve rearrangements and competing reaction channels. Here, we extend the application of RxnNet, a mechanistically driven framework for automated reaction discovery, to gas-phase fragmentation processes. RxnNet constructs reaction graphs from generalized reaction templates, explores multistep pathways, and generates transition state (TS) guesses using near-attack conformations and geodesic interpolation, followed by quantum chemical evaluation. We apply RxnNet to the fragmentation of retinal protonated Schiff base (RPSB) and its derivatives. The method reproduces known experimental pathways, captures multiple kinetically accessible mechanisms, and rationalizes differences in fragmentation patterns between related systems. In addition, RxnNet identifies a previously unreported pathway for the formation of the 112 m/z fragment in 13-des RPSB. These results demonstrate that RxnNet provides a systematic and physically grounded approach for uncovering complex fragmentation mechanisms, with broad applicability to MS/MS analysis.
Authors
- Yoni Toker (ORCID: https://orcid.org/0000-0003-1809-7659)
- Dan Thomas Major (ORCID: https://orcid.org/0000-0002-9231-0676)
- Jishnu Narayanan S J (ORCID: https://orcid.org/0000-0002-5950-5877)
- Shani Zev (ORCID: https://orcid.org/0000-0002-3731-5968)
- Michal E. Roth (ORCID: https://orcid.org/0000-0003-0910-9865)
Institutions
- Bar-Ilan University (IL)
Publication Details
- Journal
- Journal of Chemical Theory and Computation
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1021/acs.jctc.6c01229
- Primary Topic
- Mass Spectrometry Techniques and Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00