Chainer Chemistry is a deep learning framework (based on Chainer) with applications in Biology and Chemistry. It supports various state-of-the-art models (especially GCNN - Graph Convolutional Neural Network) for chemical property prediction.

For more information, please refer to the documentation. Also, a quick introduction to deep learning for molecules and Chainer Chemistry is available here.

Chainer Chemistry depends on the following packages:

• chainer
• pandas
• scikit-learn
• tqdm

These are automatically added to the system when installing the library via the pip command (see Installation). However, the following needs to be installed manually:

• rdkit (release 2017.09.3.0)

Please refer to the RDKit documentation for more information regarding the installation steps.

Note that only the following versions of Chainer Chemistry's dependencies are currently supported:

[Footnote]

*1: We used FunctionNode in this PR, which is introduced after chainer v3. See this issue for details.

*2: Saliency modules only work with chainer v5.

Chainer Chemistry can be installed using the pip command, as follows:

pip install chainer-chemistry

If you would like to use the latest sources, please checkout the master branch and install with the following commands:

git clone https://github.com/pfnet-research/chainer-chemistry.git
pip install -e chainer-chemistry

Sample code is provided with this repository. This includes, but is not limited to, the following:

• Training a new model on a given dataset
• Performing inference on a given dataset, using a pretrained model
• Evaluating and reporting performance metrics of different models on a given dataset

Please refer to the examples directory for more information.

The following graph convolutional neural networks are currently supported:

• NFP: Neural Fingerprint [2, 3]
• GGNN: Gated Graph Neural Network [4, 3]
• WeaveNet [5, 3]
• SchNet [6]
• RSGCN: Renormalized Spectral Graph Convolutional Network [10]
  * The name is not from the original paper - see PR #89 for the naming convention.
• RelGCN: Relational Graph Convolutional Network [14]
• GAT: Graph Attention Networks [15]
• GIN: Graph Isomorphism Networks [17]
• MPNN: Message Passing Neural Networks [3]
• Set2Set [19]

We test supporting the brand-new Graph Warp Module (GWM) [18]-attached models for:

• NFP ('nfp_gwm')
• GGNN ('ggnn_gwm')
• RSGCN ('rsgcn_gwm')
• GIN ('gin_gwm')

The following datasets are currently supported:

• QM9 [7, 8]
• Tox21 [9]
• MoleculeNet [11]
• ZINC (only 250k dataset) [12, 13]
• User (own) dataset

If you use Chainer Chemistry in your research, feel free to submit a pull request and add the name of your project to this list:

• BayesGrad: Explaining Predictions of Graph Convolutional Networks (paper, code)
• Graph Warp Module: an Auxiliary Module for Boosting the Power of Graph Neural Networks (paper, code)

Chainer Chemistry:

• Documentation
• Research Blog

Other Chainer frameworks:

• Chainer: A Flexible Framework of Neural Networks for Deep Learning
• ChainerRL: Deep Reinforcement Learning Library Built on Top of Chainer
• ChainerCV: A Library for Deep Learning in Computer Vision
• ChainerMN: Scalable Distributed Deep Learning with Chainer
• ChainerUI: User Interface for Chainer

This project is released under the MIT License. Please refer to the this page for more information.

Please note that Chainer Chemistry is still in experimental development. We continuously strive to improve its functionality and performance, but at this stage we cannot guarantee the reproducibility of any results published in papers. Use the library at your own risk.

[1] Seiya Tokui, Kenta Oono, Shohei Hido, and Justin Clayton. Chainer: a next-generation open source framework for deep learning. In Proceedings of Workshop on Machine Learning Systems (LearningSys) in Advances in Neural Information Processing System (NIPS) 28, 2015.

[2] David K Duvenaud, Dougal Maclaurin, Jorge Iparraguirre, Rafael Bombarell, Timothy Hirzel, Alan Aspuru-Guzik, and Ryan P Adams. Convolutional networks on graphs for learning molecular fingerprints. In C. Cortes, N. D. Lawrence, D. D. Lee, M. Sugiyama, and R. Garnett, editors, Advances in Neural Information Processing Systems (NIPS) 28, pages 2224–2232. Curran Asso- ciates, Inc., 2015.

[3] Justin Gilmer, Samuel S Schoenholz, Patrick F Riley, Oriol Vinyals, and George E Dahl. Neural message passing for quantum chemistry. arXiv preprint arXiv:1704.01212, 2017.

[4] Yujia Li, Daniel Tarlow, Marc Brockschmidt, and Richard Zemel. Gated graph sequence neural networks. arXiv preprint arXiv:1511.05493, 2015.

[5] Steven Kearnes, Kevin McCloskey, Marc Berndl, Vijay Pande, and Patrick Riley. Molecular graph convolutions: moving beyond fingerprints. Journal of computer-aided molecular design, 30(8):595–608, 2016.

[6] Kristof Schütt, Pieter-Jan Kindermans, Huziel Enoc Sauceda Felix, Stefan Chmiela, Alexandre Tkatchenko, and Klaus-Rober Müller. Schnet: A continuous-filter convolutional neural network for modeling quantum interactions. In I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors, Advances in Neural Information Processing Systems (NIPS) 30, pages 992–1002. Curran Associates, Inc., 2017.

[7] Lars Ruddigkeit, Ruud Van Deursen, Lorenz C Blum, and Jean-Louis Reymond. Enumeration of 166 billion organic small molecules in the chemical universe database gdb-17. Journal of chemical information and modeling, 52(11):2864–2875, 2012.

[8] Raghunathan Ramakrishnan, Pavlo O Dral, Matthias Rupp, and O Anatole Von Lilienfeld. Quantum chemistry structures and properties of 134 kilo molecules. Scientific data, 1:140022, 2014.

[9] Ruili Huang, Menghang Xia, Dac-Trung Nguyen, Tongan Zhao, Srilatha Sakamuru, Jinghua Zhao, Sampada A Shahane, Anna Rossoshek, and Anton Simeonov. Tox21challenge to build predictive models of nuclear receptor and stress response pathways as mediated by exposure to environmental chemicals and drugs. Frontiers in Environmental Science, 3:85, 2016.

[10] Kipf, Thomas N. and Welling, Max. Semi-Supervised Classification with Graph Convolutional Networks. International Conference on Learning Representations (ICLR), 2017.

[11] Zhenqin Wu, Bharath Ramsundar, Evan N. Feinberg, Joseph Gomes, Caleb Geniesse, Aneesh S. Pappu, Karl Leswing, Vijay Pande, MoleculeNet: A Benchmark for Molecular Machine Learning, arXiv preprint, arXiv: 1703.00564, 2017.

[12] J. J. Irwin, T. Sterling, M. M. Mysinger, E. S. Bolstad, and R. G. Coleman. Zinc: a free tool to discover chemistry for biology. Journal of chemical information and modeling, 52(7):1757–1768, 2012.

[13] Preprocessed csv file downloaded from https://raw.githubusercontent.com/aspuru-guzik-group/chemical_vae/master/models/zinc_properties/250k_rndm_zinc_drugs_clean_3.csv

[14] Michael Schlichtkrull, Thomas N. Kipf, Peter Bloem, Rianne van den Berg, Ivan Titov, Max Welling. Modeling Relational Data with Graph Convolutional Networks. Extended Semantic Web Conference (ESWC), 2018.

[15] Veličković, P., Cucurull, G., Casanova, A., Romero, A., Liò, P., & Bengio, Y. (2017). Graph Attention Networks. arXiv preprint arXiv:1710.10903. [16] Dan Busbridge, Dane Sherburn, Pietro Cavallo and Nils Y. Hammerla. (2019). Relational Graph Attention Networks. https://openreview.net/forum?id=Bklzkh0qFm

[17] Keyulu Xu, Weihua Hu, Jure Leskovec, Stefanie Jegelka, ``How Powerful are Graph Neural Networks?'', arXiv:1810.00826 [cs.LG], 2018 (to appear at ICLR19).

[18] K. Ishiguro, S. Maeda, and M. Koyama, ``Graph Warp Module: an Auxiliary Module for Boosting the Power of Graph Neural Networks'', arXiv:1902.01020 [cs.LG], 2019.

[19] Oriol Vinyals, Samy Bengio, Manjunath Kudlur. Order Matters: Sequence to sequence for sets. arXiv preprint arXiv:1511.06391, 2015.