ReactionAtlas: Ab origine exploration of chemical reaction networks with machine learning — supplemental dataset
Authors/Creators
-
1.
Technische Universität Berlin
-
2.
Berlin Institute for the Foundations of Learning and Data
-
3.
Google (United States)
-
4.
Korea University
-
1.
Technische Universität Berlin
Description
This dataset is the chemical reaction network (CRN) explored by the ReactionAtlas system described in the accompanying paper (see Related identifiers).
Contents:
12,542 unique compounds (191 molecular formulas, C0–C4, charge −3 to +2); 33,659 potential-energy-surface minima (avg. 2.68 conformers per compound), plus per-compound conformer ensembles generated with CREST/GFN-xTB; 24,172 intramolecular (conformer-to-conformer) transition states within the per-compound PES graphs; and 47,987 elementary-reaction transition states collapsing to 21,150 unique reactions (by reactant–product identity). Each reaction includes the full ML reaction path — the transition state, both IRC relaxation trajectories, and the connected reactant/product minima — with ML energies, forces, and Hessians. Reactive TSs are refined at the PBE0/def2-TZVPP level: single-point electronic energies on the reactant, TS, and product geometries (via PySCF) yield in-box and separated DFT barriers, with D3(BJ) and D4 dispersion corrections layered on top. Where the ML-predicted TS is flagged invalid, a corrected PBE0/def2-TZVPP saddle point is re-optimised with frequency analysis (via ORCA) and its energy difference and RMSD to the ML geometry are recorded; a PBE0 Hessian on the transition state is provided for nearly all reactive TSs.
Format
PostgreSQL 15 database dump (plain SQL, gzip-compressed; ≈26 GB compressed, ≈70 GB restored). Restore with:
createdb reaction_atlas
gunzip -c reaction-atlas-*.sql.gz | psql -v ON_ERROR_STOP=1 -d reaction_atlas
Geometry and Hessian columns are stored as serialized NumPy arrays in `bytea` columns (`numpy.save` for single arrays; `numpy.savez_compressed` for trajectories). A column-level schema reference and restore guide are at https://reactionatlas.bifold.berlin/downloads.
The full network can be browsed at https://reactionatlas.bifold.berlin. The exploration and database code is at https://github.com/mx-e/reaction-atlas.
Files
Files
(26.9 GB)
| Name | Size | |
|---|---|---|
|
md5:3ea2e02391a36f0ac1e8a35586f96e10
|
26.9 GB | Download |
Additional details
Identifiers
Related works
- Is documented by
- Software: https://github.com/mx-e/reaction-atlas (URL)
- Is supplement to
- Publication: arXiv:2606.30778 (arXiv)
- Is supplemented by
- Other: https://reactionatlas.bifold.berlin (URL)
Funding
- Swiss National Science Foundation
- Postdoc Mobility Fellowship 225476
- Ministerium für Wissenschaft, Forschung und Kultur
- BIFOLD award 01IS18037A, 01IS14013A-E, 01GQ1115, 01GQ0850, 01IS18025A, 031L0207D
- Institute for Information and Communications Technology Promotion
- AI Graduate School Program, Korea University 2019-0-00079, 2022-0-00984
Software
- Repository URL
- https://github.com/mx-e/reaction-atlas
- Programming language
- Python
- Development Status
- Active
References
- M. Eissler, T. Korjakow, S. Ganscha, O. T. Unke, K.-R. Müller, S. Gugler, "How simple can you go? An off-the-shelf transformer approach to molecular dynamics," J. Chem. Phys. 164, 094308 (2026). doi:10.1063/5.0295035
- S. Ganscha, O. T. Unke, D. Ahlin, H. Maennel, S. Kashubin, K.-R. Müller, "The QCML dataset: quantum chemistry reference data from 33.5M DFT and 14.7B semi-empirical calculations," Sci. Data 12, 406 (2025). doi:10.1038/s41597-025-04720-7
- K. Kahouli, S. S. P. Hessmann, K.-R. Müller, S. Nakajima, S. Gugler, N. W. A. Gebauer, "Molecular relaxation by reverse diffusion with time step prediction," Mach. Learn.: Sci. Technol. 5, 035038 (2024). doi:10.1088/2632-2153/ad652c
- C. Adamo, V. Barone, "Toward reliable density functional methods without adjustable parameters: the PBE0 model," J. Chem. Phys. 110, 6158-6170 (1999). doi:10.1063/1.478522
- F. Weigend, R. Ahlrichs, "Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: design and assessment of accuracy," Phys. Chem. Chem. Phys. 7, 3297-3305 (2005). doi:10.1039/b508541a
- F. Neese, "Software update: the ORCA program system - version 5.0," WIREs Comput. Mol. Sci. 12, e1606 (2022). doi:10.1002/wcms.1606
- Q. Sun et al., "Recent developments in the PySCF program package," J. Chem. Phys. 153, 024109 (2020). doi:10.1063/5.0006074
- P. Pracht, F. Bohle, S. Grimme, "Automated exploration of the low-energy chemical space with fast quantum chemical methods," Phys. Chem. Chem. Phys. 22, 7169-7192 (2020). doi:10.1039/C9CP06869D
- C. Bannwarth, S. Ehlert, S. Grimme, "GFN2-xTB: an accurate and broadly parametrized self-consistent tight-binding quantum chemical method with multipole electrostatics and density-dependent dispersion contributions," J. Chem. Theory Comput. 15, 1652-1671 (2019). doi:10.1021/acs.jctc.8b01176
- D. G. Goodwin, H. K. Moffat, R. L. Speth, "Cantera: an object-oriented software toolkit for chemical kinetics, thermodynamics, and transport processes, version 2.2.1" (2016). doi:10.5281/zenodo.45206
- S. Grimme, J. Antony, S. Ehrlich, H. Krieg, "A consistent and accurate ab initio parametrization of density functional dispersion correction (DFT-D) for the 94 elements H-Pu," J. Chem. Phys. 132, 154104 (2010). doi:10.1063/1.3382344
- S. Grimme, S. Ehrlich, L. Goerigk, "Effect of the damping function in dispersion corrected density functional theory," J. Comput. Chem. 32, 1456-1465 (2011). doi:10.1002/jcc.21759
- E. Caldeweyher, S. Ehlert, A. Hansen, H. Neugebauer, S. Spicher, C. Bannwarth, S. Grimme, "A generally applicable atomic-charge dependent London dispersion correction," J. Chem. Phys. 150, 154122 (2019). doi:10.1063/1.5090222