Molecule Design using Monte Carlo Tree Search with Neural Rollout. ChemTS can design novel molecules with desired properties(such as, HOMO-LUMO gap, energy, logp..). Combining with rDock, ChemTS can design molecules active to target proteins. The ChemTS paper is available at https://arxiv.org/abs/1710.00616 . Also, we introduced the distributed parallel ChemTS that can accerlate molecular discovery. And the distributed parallel ChemTS is available at https://github.com/tsudalab/DP-ChemTS.
For usage, please refer the following instructions. Currently, the package hasn't been finished very well... If you want to implement your own simulator, please check add_node_type.py. The full package will be updated later.
There are two versions of chemts for logP optimization. The search tree of the old version added all possible smiles symbols as children nodes. While the new version chemts only expands children nodes with high probabilities predicted by RNN.
This package is distributed under the MIT License.