MPCONF196GEN Benchmark

Molecular Conformer Generation Benchmark

A conformer generation benchmark created from the MPCONF196 benchmark energy dataset [1]. This evaluates a method's ability to generate conformers of macrocycles, a difficult task for molecular conformer generation.

We generated reference conformers with CREST iMTD-GC [2] using GFN2-xTB [3] and computed conformer energies using Egret-1 [4] neural network potential.

Usage

• SMILES.txt contains the number of rotateable bonds, canonical SMILES strings and identifier for each molecule.
• Each molecules/{molecule_id}/ directory contains a crest_conformers.xyz file with conformers and energies for each molecule.

Creating the Benchmark

All packages necessary for making this benchmark can be installed by creating a conda envrionment from env.yml.

We ran CREST on each data/{molecule_id}.xyz file, outputs and conformers can be found in molecules/{molecule_id}/crest.out and molecules/{molecule_id}/crest_confromers.xyz respectively. The Egret-1 checkpoint and scripts we used to calculate conformer energies and create SMILES.txt can be found in tools/. We generated SMILES for each molecule from the data/{molecule_id}.xyz file with rowansci.com, except for data/YIVNOG_I.xyz which we made by hand.

References

[1] Řezáč, J.; Bím, D.; Gutten, O.; Rulíšek, L. Toward Accurate Conformational Energies of Smaller Peptides and Medium-Sized Macrocycles: MPCONF196 Benchmark Energy Data Set. J. Chem. Theory Comput. 2018, 14 (3), 1254–1266. https://doi.org/10.1021/acs.jctc.7b01074

[2] Pracht, P.; Grimme, S.; Bannwarth, C.; Bohle, F.; Ehlert, S.; Feldmann, G.; Gorges, J.; Müller, M.; Neudecker, T.; Plett, C.; Spicher, S.; Steinbach, P.; Wesołowski, P. A.; Zeller, F. CREST—A Program for the Exploration of Low-Energy Molecular Chemical Space. J. Chem. Phys. 2024, 160 (11), 114110. https://doi.org/10.1063/5.0197592

[3] Bannwarth, C.; Ehlert, S.; Grimme, S. 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. 2019, 15 (3), 1652–1671. https://doi.org/10.1021/acs.jctc.8b01176

[4] Mann, E. L.; Wagen, C. C.; Vandezande, J. E.; Wagen, A. M.; Schneider, S. C. Egret-1: Pretrained Neural Network Potentials for Efficient and Accurate Bioorganic Simulation. arXiv:2504.20955, 2025. https://doi.org/10.48550/arXiv.2504.20955