XTB2-MolData : Dataset of 12 Million Molecules

This dataset is an open chemistry database containing optimized molecular geometries and electronic properties calculated by the GFN2-xTB method (C. Bannwarth et al.) for 12.6 million organic molecules contained C, H, O, and N atoms.

The initial geometries, before optimization by GFN2-xTB method, are taken from PubChem PM6 (Shimazaki et al.) database.

We also include our python code to manage a large molecule database. This code includes scripts to generate input files for Gaussian software, to read Gaussian output files, to create a small reduced dataset based on clustering algorithm, and many scripts to analyze the molecular properties included in the database.

This code can be also taken from github:  https://github.com/Castaneche/MolDataFW.