Decoder part of the adversarial autoencoder for generating molecular descriptors (fingerprints) of compounds with anti-HIV properties (inhibitors of gp120 protein, exposed on the surface of HIV-1)
Using conda, create an environment and activate it:
conda env create -f environment.yml
activate aae
In order to use the model for generating fingerprints with a certain binding energy threshold, run app.py with the following arguments:
--energy: Energy threshold to generate with.--number_fps: Amount of fingerprints to generate.
python app.py --energy -9 --number_fps 100
As the training was based on compounds with binding energy property ranged from 0 to -14, model is not expected to generate coherent results outside those boundaries