Data for Structural bias in three-dimensional autoregressive generative machine learning of organic molecules

Molecular databases used for training the autoregressive generative ML model G-SchNet, and databases of molecules generated by these trained models. Fitted PCA models for structural and bonding descriptors of molecules and the resulting principal components of all molecules, used for spatial analysis in a latent chemical space, are included.Databases tagged with properties and principal components trained/evaluated using the GSchNetTools Python package, available on GitHub.Detailed file descriptions are provided in the attached README.