RNN Tool-Kits for Predicting Molecular Properties from SMILES
A Python/Keras implementation of Neural Fingerprint  models supporting both Theano and Tensorflow as back end. These are a type of recurrent convolutional neural networks specialized for regression tasks on molecules (using SMILES as input). Models can be trained to predict properties like e.g., aqueous solubility, melting points or toxicity of molecules.
This is a Tensorflow implementation for models proposed in . These are recursive neural networks operating on molecular graphs (SMILES) of arbitrary size for chemical property prediction - this is a different class of models solving the same problems as Neural Fingerprints.
 "Convolutional Networks on Graphs for Learning Molecular Fingerprints" (D. Duvenaud et al. )
 "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)