Executive Summary : | There is an urgent need to develop a novel system that predicts SNPs, mutations, diseases from viral genome sequences and drug molecules to cure the viral diseases including COVID-19 of Indian population. Understanding the virus’s genetic sequence is essential in order to act quickly to control the virus when outbreak, such as COVID-19, occurs. SNPs play a role in the pathophysiology of many human diseases as well as human biological variation. Identifying such SNPs are useful for genetic studies, personalized medicine, forensic science, and evolutionary studies. Discovering mutated patterns are helpful in diagnosing genetic disorders, prognosis, drug development and gene therapy. One of the major challenges in drug discovery is the enormous search space for novel molecules. Traditional drug discovery approaches are time consuming and expensive, so there is a compelling need to create new methodologies that minimize the search space for drug candidates. In pandemics, finding drugs quickly is critical to minimize life-threatening situations. This instilled a sense of urgency in the development of effective drug molecules through the exploration of drug discovery techniques as well as the development of novel drug-target interaction models to analyze drug effectiveness. The proposed hybrid system consists of three modules: Viral Disease Prediction including Mutations and SNPs, Novel Drug Molecules Generation and Refinement, Drug Target Interaction for identifying suitable drug candidates. Viral disease prediction is performed using convolutional neural networks and mutated patterns are identified using explainable AI. SNPs are detected by mapping using viral reference sequence. The translation rate of the viral sequences is predicted using codon usage bias and slow codon and di-codons frequency. Generative adversarial network models are trained to generate novel drug molecules and fine-tuned with existing commercial anti-viral drugs. Drug target interaction models using GCNNs are developed for finding affinities of the fine-tuned drug molecules and finally identifying suitable drug candidates. The detected mutations are validated using existing motif database tools. The novel generated molecules are evaluated using three quality measures namely, validity, uniqueness, and novelty. The drug candidates are validated using binding affinity scores. For the scenario like Indian population the proposed system can act as a fast and efficient tool for genetic studies, diagnosing genetic disorders, prognosis, drug development, personalized medicine, gene therapy, forensic science, and evolutionary studies. |