deep learning models for the 3D binding affinity of proteins

Introduction

Due to the pandemic urgency, I would like to work on meaningful projects. So I worked on a Lawrence Livermore Lab project related to the COVID19 drug discovery. The project I worked on was to develop algorithms to explain the complicated deep learning model. I modified the most popular algorithms for computer vision problems like Gradient-weighted Class Activation Mapping, Concept Activation Vector, Automated Concept-based Explanation, and PointNets,code. I used them to explain the effect of hidden layers and their essential features for the 3D-CNN model to predict protein binding. This is the presentation I did in the end.