SMILES-X is an automated pipeline that uses only molecular SMILES as input to predict physicochemical properties, such as solubility, hydration-free energy, and lipophilicity. This tool is specifically designed for small datasets (less than 1000 samples) and doesn’t require human-made descriptors. With it, users can design neural architectures through Bayesian optimization, predict molecular characteristics from a list of SMILES based on these models, and visualize elements or substructures to understand the predictions better.


Official site
Openness ★★★

MIT License

Core Developers

Guillaume Lambard (National Institute for Materials Science)

Ekaterina Gracheva (National Institute for Materials Science, University of Tsukuba)

  • python 3.7

It is highly recommended to use GPUs rather than CPUs.

  • CUDA=10.1
  • cuDNN=8.0.3

is required for GPU use.