Chemistry as Language

========

ChemAsLang is a collection of resources which treats chemistry as language. It breaks sets of chemistry down into component parts (fragments), implements networks of fragments, and trains Word-to-Vector algorithms on fragments to track changes in chemistry over time.

Features

• Utilizes RDKit generate chemical fragments
• Uses Gensim to train Word-to-Vector models of fragments & chemical compounds

Usage

• To generate fragments: upload a set of chemicals (in SMILES or SMARTS format) to the Fragments/Data directory, edit Fragments/common_frags_parallel.py as necessary.
• To train W2V: edit Word2Vec/build_KEGG_gensim.py as necessary.

Progress

This is an ongoing project as part of my graduate research at Arizona State Univeristy. Future plans are to:

• Simplify the workflow for generating fragments from a list of user-specified chemical compounds (estimated date: Summer 2021)
• Train a W2V model as part of the fragment generation process (estimated date: Fall 2021)
• Release this code as a fully documented Python package (estimated data: Spring 2022)

Support

If you have questions, email me at: jfmalloy1@gmail.com

License

The project is licensed under the MIT license.