MIRRA: Fast, High-Fidelity Molecular Generation with Chemistry-Informed Discrete Flow Matching
Abstract
Fast, high-fidelity molecular generation requires accurate chemical choices and efficient sampling. Chemical building blocks reduce graph size, but make each node a choice among tens of thousands of related identities. We introduce MIRRA (Molecular Identity Representation, Refresh and Assembly), a discrete flow-matching generator that learns from chemical knowledge and evaluates identity probabilities only when needed, while keeping every identity in its vocabulary available. Chemical features share information across related blocks, and learned identity corrections distinguish them in molecular context. By deciding which positions will refresh before computing their probabilities, the sampler avoids unused predictions while preserving each sampling transition. Against the strongest baselines in each task, MIRRA reduces chemical and graph distribution errors on ZINC250k by 29-31% and 53-60%, respectively, with generation up to 8.6 times faster. Property targeting yields an estimated 7.8-20.8 times more distinct novel hits per second. On a held-out polymer test set, MIRRA achieves 100% periodic validity with known chain topology, versus 57.0% for a polymer variational autoencoder. These results demonstrate fast, high-fidelity generation with the full chemical vocabulary available throughout sampling.
Related
- Molecular representation shapes the balance between target fidelity and exploration in flow based polymer generation
- A Molecular Generation–Validation Workflow Combining Reversible Junction Tree Reinforcement Learning and Semi-empirical Quantum Chemistry: A Case Study on Molecular Solar Thermal Fuels
- Chemistry-Informed Transfer and Interaction-Resolved Learning Towards Data-Efficient Materials Exploration
- Natural Product Chemistry in Sickle Cell Disease Therapeutics: Phytochemical Mechanisms, Pharmacokinetics, and Computational Docking Insights for Next-Generation Anti-Sickling Agents: A Comprehensive Review
- LinkerGen: Guided Autoregressive Generation of Organic Linkers for Reticular Chemistry
BibTeX
@misc{sun2026mirra,
title = {{MIRRA: Fast, High-Fidelity Molecular Generation with Chemistry-Informed Discrete Flow Matching}},
author = {Ge Sun and Juan J. de Pablo},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.17244},
eprint = {2610.17244},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.17244},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - MIRRA: Fast, High-Fidelity Molecular Generation with Chemistry-Informed Discrete Flow Matching
AU - Ge Sun
AU - Juan J. de Pablo
PY - 2026
DA - 2026/10/12
PB - arXiv
DO - 10.48550/arxiv.2610.17244
UR - https://arxiv.org/abs/2610.17244
AB - Fast, high-fidelity molecular generation requires accurate chemical choices and efficient sampling. Chemical building blocks reduce graph size, but make each node a choice among tens of thousands of related identities. We introduce MIRRA (Molecular Identity Representation, Refresh and Assembly), a discrete flow-matching generator that learns from chemical knowledge and evaluates identity probabilities only when needed, while keeping every identity in its vocabulary available. Chemical features share information across related blocks, and learned identity corrections distinguish them in molecular context. By deciding which positions will refresh before computing their probabilities, the sampler avoids unused predictions while preserving each sampling transition. Against the strongest baselines in each task, MIRRA reduces chemical and graph distribution errors on ZINC250k by 29-31% and 53-60%, respectively, with generation up to 8.6 times faster. Property targeting yields an estimated 7.8-20.8 times more distinct novel hits per second. On a held-out polymer test set, MIRRA achieves 100% periodic validity with known chain topology, versus 57.0% for a polymer variational autoencoder. These results demonstrate fast, high-fidelity generation with the full chemical vocabulary available throughout sampling.
ER -