Deep VASP-E: A Flexible Analysis of Electrostatic Isopotentials for Finding and Explaining Mechanisms that Control Binding Specificity
Visualization technique used in this project is Grad CAM++(https://arxiv.org/abs/1710.11063).
Models package is responsible in holding different models being used for different projects. Note, models do not do any hypertuning of hyperparameters. Keras has tuners available.
Package is responsible for preprocessing dataset and postprocessing model results (Grad CAM++).
Dataset structure for loading in training and test set is done with the following format:
Dataset_directory/protein001/protein001_example001.CNN
Currently, implementation requires code to specify the evaluation set following a leave one out approach.