Abstract

While numerous computational tools exist for predicting protein-protein interactions (PPIs) based on amino acid sequences, most are tailored to species-specific interactions and struggle to generalize across species boundaries. In particular, traditional homogeneous PPI prediction algorithms often fail to detect interactions between proteins from different organisms. To address this limitation, we developed a deep learning-based artificial intelligence model that encodes the frequency of consecutive amino acids within protein sequences. Our approach specifically targets the prediction of human-virus protein interactions by leveraging protein annotations and sequence patterns. The proposed representation technique is both simple and effective, offering several advantages: it enhances model performance, enables consistent feature vector generation, and supports application to a wide variety of protein types. Simulation results demonstrate that our method outperforms existing approaches, achieving a prediction accuracy of 98%, thereby highlighting its potential for advancing cross-species PPI prediction.

Authors

B. Sharmila, A.V. Santhosh Babu
Vivekanandha College of Engineering for Women, India

Keywords

Deep Learning, Protein Interaction, Prediction Antiviral-HPV Protein