Leveraging Quantum LSTM for High-Accuracy Prediction of Viral Mutations

The rapid mutations of viruses such as COVID-19 pose significant challenges for vaccine development and effective disease management. In response to these challenges, this study introduces a novel quantum-enhanced LSTM (QLSTM) model designed to predict genetic mutations, specifically focusing on viral protein sequences. The QLSTM model leverages quantum computing techniques, including superposition and entanglement, to improve the model’s ability to handle the high-dimensional, nonlinear structure inherent in viral genomic datasets. These quantum enhancements allow the model to capture complex relationships in the data, improving its accuracy and performance in mutation detection compared to traditional methods. To improve model predictions, we use two key preprocessing techniques TF-IDF for efficient feature extraction and PCA for dimensionality reduction of genomic sequences. TF-IDF helps the model focus on the most informative nucleotide features, while PCA reduces the size of the data, making the model computationally efficient without sacrificing important information. The one-hot encoding technique is a standard technique in machine learning for encoding protein sequences into data that can be used in neural networks.The proposed QLSTM outperformed existing deep learning architectures such as the Attention-Augmented Convolutional Neural Network (AACNN), Stacked Recurrent Neural Network (Stacked RNN), Retention Network (RetNet), and Bidirectional Long Short Term Memory (BiLSTM). These results indicate that QLSTM not only provides high accuracy in mutation predictions in SARS-CoV-2 protein sequences but also provides deep insights into the functional implications of such mutations. The model identifies mutation hotspots that affect virus spread, immune evasion, and protein structure, providing key biological insights for future vaccine development and therapeutic strategies.