Genetic Algorithm Mutations for Molecules with a Hybrid Language Model-Based GAN Architecture

LM-GAN Molecules

Using a language model on GAN architechture to generate molecules. Generated molecules from this LM-GAN are produced by following the process of randomly masking input molecules to this architechture and then using the LM-GAN model to make our predictions. Genetic algorithm is performed for optimizing generated population of molecules. To get more information of our langugage model using genetic algorithm please see our earlier work. 

Dependencies

• rdkit
• pytorch
• transformers

Example

An example script is provided here for using this LM-GAN model. Please create a results directory before running.

To perform optimization without LM-GAN model, please use the following flag in command '--generator_only':