ABSTRACT
Since the improvement of a new drug is a complex, costly, and a quite overlong process, how to decrease prices and speed up new drug discovery has turned into a compelling and immediate problem in the industry. So, the importance of computer-aided drug discovery has been steadily increasing in recent times. Notwithstanding the increasing number of accomplished forward-looking practices, deep learning models need to be explainable as mathematical models are often difficult to interpret by the human mind. In this study, first, basic information about current explainable artificial intelligence (XAI) methods such as class activation mapping (CAM), local interpretable model-agnostic explanations (LIME), SHAP, Fairlearn, and Whitenoise is given. Afterward, XAI approaches used in the context of drug discovery such as feature attribution, instance-based, graph-convolution based, self- explaining, and uncertainty prediction are mentioned and predictions for future studies are made.
