Decoding SARS-CoV-2 Variants: An in-silico approach to RNA-Seq feature extraction using K-mers and N-grams

The world's enormous challenge is to combat the COVID-19 pandemic, which has spread by the SARS-CoV-2 virus over the past three years. Many waves are formed with diverse variations categorized as variants of concern and variants of interest by the World Health Organization (WHO). Moreover, the emergence of a new coronavirus and its susceptibility to various pandemic conditions throughout the world strongly imply that the virus is taking new shapes, resulting in divergent variations, and predicting the virus becomes more challenging.Objective: This article proposes an integrated intelligent prediction model to predict coronavirus variants of concern, such as alpha, beta, gamma, delta, and omicron, using RNA-Seq and machine learning techniques.Method: We proposed machine learning-based SARS-CoV-2 variants prediction model with 2028 RNA-Seq of alpha, beta, gamma, delta, and omicron were gathered to train prominent classifiers for the prediction of any unseen sequences of these variants. In this order, the feature vector was created from the motif of the sequence obtained through K-mer and N-gram methodologies to build prediction model.Result: The findings show that the developed prediction model based on Logistic Regression (LR) performs better by reaching an aggregated accuracy of 99% during training and testing compared to other models. We also validated the models using unseen variant sequences and found a significant prediction rate.Conclusion: Our experiment showed that the suggested SARS-CoV-2 variants prediction model is capable of predicting the SARS-CoV-2 variants of concern using a logistic regression trained model.