(Retracted) DeepFNN-DTBA: prediction of drug-target binding affinity via feed-forward neural network on drug-protein sequences

Abstract

Pharmaceutical research essentially depends on drug-target interactions (DTIs). Standard methods of experimentation to uncover DTIs are costly and time-consuming, and thus artificial intelligence and machine learning have become popular. Multimodal imaging also provides significant amounts of anatomical, functional, and molecular information, accelerating drug discovery and development. Imaging technologies help understand disease mechanisms find new pharmacological targets and evaluate new drug candidates and how well they work. In this research, we developed a model based on deep learning (DL) that employs sequence information for targets and medicines to ascertain binding affinities of DTIs named feed-forward neural network (FNN)-DT binding affinity. Existing studies for the prediction of binding affinity of DTs either employ three-dimensional structures of protein–ligand complexes or two-dimensional characteristics of compounds. A novel technique used in this research: a dense network with dropouts were used to show the protein and drug sequences. These findings support the proposed DL-based approach for predicting binding affinity in DTIs, which use 1D representations of targets and medicines. In one of the standard datasets, the proposed FNNs outperformed the Kronecker regularized least squares, gradient boosting machines, deep drug target affinity algorithm, wide drug-target affinity, and similarity-based convolutional neural network model techniques with a 0.89 concordance index and 0.235 mean square error.