A Novel Protein Mapping Method for Predicting the Protein Interactions in COVID-19 Disease by Deep Learning

Area of Science:

• Computational Biology
• Bioinformatics
• Artificial Intelligence in Medicine

Background:

• The COVID-19 pandemic, caused by SARS-CoV-2, necessitates rapid identification of therapeutic targets.
• Understanding SARS-CoV-2 and human protein-protein interactions is vital for drug discovery and understanding disease mechanisms.
• Current experimental methods for identifying protein interactions are time-consuming and have limitations in coverage.

Purpose of the Study:

• To develop and evaluate a novel, efficient protein-mapping method for predicting SARS-CoV-2 and human protein interactions.
• To leverage artificial intelligence, specifically deep learning, for accelerated protein interaction prediction.
• To contribute a new computational tool to the field of virology and drug development.

Main Methods:

• Proposed a novel protein-mapping method utilizing the AVL tree data structure for efficient sequence mapping.
• Mapped SARS-CoV-2 and human protein sequences using the proposed AVL tree method and compared it with existing methods.
• Utilized bidirectional recurrent neural networks (BRNNs) for normalizing and classifying mapped protein sequences.
• Evaluated the performance using standard metrics: accuracy, F1-score, precision, recall, and AUC.

Main Results:

• The proposed AVL tree-based mapping method achieved high prediction accuracy (97.76%) for SARS-CoV-2 and human protein interactions.
• Achieved excellent precision (97.60%) and recall (98.33%), indicating reliable identification of true interactions.
• Demonstrated strong overall performance with an average AUC of 89% and an F1-score of 79.42%.

Conclusions:

• The novel AVL tree-based protein mapping method is a highly accurate and efficient tool for predicting virus-host protein interactions.
• This AI-driven approach significantly accelerates the identification of potential drug targets for COVID-19.
• The method offers a valuable computational alternative to experimental techniques, aiding in faster drug discovery and development.