Mechanistic Reaction Prediction via Discrete Flow Matching on Graph-Structured Electron Occupation
2026-08-27 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors present MAELLE, a new method that models chemical reactions by tracking how electrons move during the reaction instead of directly predicting products or changing molecular graphs. They use a mathematical framework called a Continuous-time Markov Chain to represent electron rearrangements across different sites on molecules. This approach generates step-by-step pathways of electron flow that match real chemical mechanisms without needing detailed reaction step labels. MAELLE performs well compared to other models, especially on complex or unfamiliar reactions, and can also predict side products by considering full electron redistribution.
electron rearrangementContinuous-time Markov ChainOptimal Transportreaction predictionmolecular topologyelectron occupation vectorsflow matchingmechanistic pathwayside productsUSPTO-480K benchmark
Authors
Nguyen Xuan-Vu, Octavian Susanu, Daniel Armstrong, Philippe Schwaller
Abstract
Chemical reactions are fundamentally transformations in electron space, yet most machine learning approaches model them either through \textit{de novo} generation of product molecules or through heuristic graph edits that operate directly on molecular topology. We introduce MAELLE (\textbf{M}ech\textbf{A}nistic \textbf{E}dit f\textbf{L}ow-matching on e\textbf{L}ectron r\textbf{E}arrangements), which instead models reactions as discrete flow matching over electron occupation vectors. Concretely, we formulate the reactant-to-product mapping as a Continuous-time Markov Chain (CTMC) over the graph-structured integer-valued electron occupation space defined on all bonding, non-bonding, and hydrogen sites. To construct the intermediate edit trajectories, we generalize the discrete flow matching mixture path to discrete electron rearrangements using Optimal Transport, yielding a sequence of mechanistically interpretable edit moves without requiring elementary step annotations. MAELLE achieves competitive performance on the USPTO-480K benchmark compared with leading reaction prediction models. Beyond in-distribution accuracy, we evaluate robustness across two out-of-distribution settings - structural complexity and reaction type - and find that MAELLE maintains strong performance where existing methods degrade. Finally, because the learned flow operates over the full electron redistribution, MAELLE naturally recovers mechanistic trajectories that align with known chemistry and can predict side products of a reaction.