Comparison of Embedding Techniques used in Prediction of Drug Target Binding Affinity

Abstract

The process of identifying of drug-target (DT) interactions is an integral part of the drug discovery process. Drug discovery is a process which focuses on determining new compounds that can be used to cure and treat diseases. Embedding is the process of converting high-dimensional data to low-dimensional data by representing it in the form of a vector. Drug Target sequences need to be transformed into a matrix before they can be fed into a deep learning model. Since the performance of embedding techniques directly affects the quality of the deep learning models and thus the accuracy of the predicted values of drug target binding affinity, we compare the performance and effect of various embedding techniques used for target embedding on deep learning models. Here, the embedding techniques whose performance has been compared are PLUSRNN Embedder, ProtTransBertBFD Embedder, Beppler Embedder, CPCProt Embedder and SeqVec Embedder.