Bat-inspired optimizer for prediction of anti-viral cure drug of SARS-CoV-2 based on recurrent neural network

Abstract

COVID-19 is a large family of viruses that causes diseases ranging from the common cold to more severe diseases such as SARS-CoV. There are currently several attempts to create an anti-viral drug to combat the virus. The antiviral medicines could be promising treatment choices for COVID-19. Therefore, a fast strategy for drugs application that can be utilized to the patient immediately is necessary. In this context, deep learning-based architectures can be considered for predicting drug-target interactions accurately. This is due to a large amount of complicated knowledge, such as hydrophobic interactions, ionic interactions, and bonding with hydrogen. In this paper, Recurrent Neural Network (RNN) is used to build drug-target interaction prediction model to predict drug-target interactions. Bat Algorithm (BA) is used in this paper to optimize the model parameters of RNN (RNN-BA) and then using the corona virus as a target. The drug with the best binding affinity will be a potential cure for the virus. The proposed model consists of different four phases; data preparation phase, hyper-parameters optimizing phase, learning phase and fine-tuning for specific ligand subsets. The used dataset in this paper to train and evaluate the proposed model is selected from a total of 677,044 SMILES. The experimental results of the proposed model showed high level of performance in comparison with the related approaches.