RNN Tool-Kits for Predicting Molecular Properties from SMILESs

RNN Tool-Kits for Predicting Molecular Properties from SMILES

This software repository contains two types of deep recursive neural networks implemented in Tensorflow and Keras for molecular property predictions based on molecular graphs: inner- and outer recursive neural networks. Both types of networks can be trained to predict physical, chemical or biological properties of small molecules (e.g. aqueous solubility, melting point) starting from their SMILES strings. The Inner approach for molecules is described in [1], while the outer approach is described in [2].

[1] "Deep Architectures and Deep Learning in Chemoinformatics: The Prediction of Aqueous Solubility for Drug-Like Molecules" (A. Lusci et al., Journal of Chem. Inf. Modeling, 2013)
[2] "Convolutional Networks on Graphs for Learning Molecular Fingerprints" (D. Duvenaud et al., Advances in Neural Information Processing Systems 28, 2015)