Datasets for understanding the importance of conformation in property prediction models

Descriptor and conformer data sets for molecular property and reaction selectivity prediction tasks. The PQC data set was created based on a part of the PubChemQC PM6 dataset (J. Chem. Inf. Model. 2020, 60, 12, 5891–5899), which contains two- and three-dimensional descriptors and conformers. The APTC data sets are based on the data sets for asymmetric phase transfer catalysts with enantio-selectivity (https://github.com/Laboratoire-de-Chemoinformatique/3D-MIL-QSSR/tree/main/datasets). The melting point data set was created from the Jean-Claude Bradley Double Plus Good (Highly Curated and Validated) Melting Points Dataset (https://doi.org/10.6084/m9.figshare.1031638.v1).

They contained descriptors and conformers to train and validate machine learning models.

Detailed explanations on how to use these datasets are found in the Github repository: https://github.com/YuHamakawa/Conformation-Importance-ML-Models.