The DeepChem Project¶
The DeepChem project aims to democratize deep learning for science.
What is DeepChem?¶
The DeepChem project aims to build high quality tools to democratize the use of deep learning in the sciences. The origin of DeepChem focused on applications of deep learning to chemistry, but the project has slowly evolved past its roots to broader applications of deep learning to the sciences.
The core DeepChem Repo serves as a monorepo that organizes the DeepChem suite of scientific tools. As the project matures, smaller more focused tool will be surfaced in more targeted repos. DeepChem is primarily developed in Python, but we are experimenting with adding support for other languages.
What are some of the things you can use DeepChem to do? Here’s a few examples:
- Predict the solubility of small drug-like molecules
- Predict binding affinity for small molecule to protein targets
- Predict physical properties of simple materials
- Analyze protein structures and extract useful descriptors
- Count the number of cells in a microscopy image
- More coming soon…
We should clarify one thing up front though. DeepChem is a machine learning library, so it gives you the tools to solve each of the applications mentioned above yourself. DeepChem may or may not have prebaked models which can solve these problems out of the box.
Over time, we hope to grow the set of scientific applications DeepChem can address. This means we need lots of help! If you’re a scientist who’s interested in open source, please pitch on building DeepChem.
If you’d like to install DeepChem locally, we recommend using
conda and installing RDKit with deepchem.
RDKit is a soft requirement package, but many useful methods like
molnet depend on it.
pip install tensorflow-gpu==1.14 conda install -y -c conda-forge rdkit deepchem
For CPU only support instead run
pip install tensorflow==1.14 conda install -y -c conda-forge rdkit deepchem
Then open your python and try running.
DeepChem developer calls are open to the public! To listen in, please email X.Y@gmail.com, where X=bharath and Y=ramsundar to introduce yourself and ask for an invite.
Licensing and Commercial Uses¶
DeepChem is licensed under the MIT License. We actively support commercial users. Note that any novel molecules, materials, or other discoveries powered by DeepChem belong entirely to the user and not to DeepChem developers.
That said, we would very much appreciate a citation if you find our tools useful. You can cite DeepChem with the following reference.
Support the DeepChem project by starring us on on GitHub. Join our forums at https://forum.deepchem.io to participate in discussions about research, development or any general questions. If you’d like to talk to real human beings involved in the project, say hi on our Gitter chatroom.
- Data Loaders
- Data Classes
- Model Classes
- Keras Models
- Hyperparameter Tuning
- Contributing a new dataset to MoleculeNet
- BACE Dataset
- BBBC Datasets
- BBBP Datasets
- Cell Counting Datasets
- Chembl Datasets
- Chembl25 Datasets
- Clearance Datasets
- Clintox Datasets
- Delaney Datasets
- Factors Datasets
- HIV Datasets
- HOPV Datasets
- HPPB Datasets
- KAGGLE Datasets
- Kinase Datasets
- Lipo Datasets
- Materials Datasets
- MUV Datasets
- NCI Datasets
- PCBA Datasets
- PDBBIND Datasets
- PPB Datasets
- QM7 Datasets
- QM8 Datasets
- QM9 Datasets
- SAMPL Datasets
- SIDER Datasets
- Thermosol Datasets
- Tox21 Datasets
- Toxcast Datasets
- USPTO Datasets
- UV Datasets
- Reinforcement Learning
- Coding Conventions