SDNN-PPI: self-attention with deep neural network effect on protein-protein interaction prediction

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Protein-protein interactions (PPIs) are crucial for cellular functions, including transcriptional regulation and signal transduction.
• Traditional experimental methods for identifying PPIs are resource-intensive and time-consuming.
• Developing computational methods for PPI prediction is essential for accelerating biological research.

Purpose of the Study:

• To propose SDNN-PPI, a novel deep learning-based method for predicting protein-protein interactions.
• To leverage sequence-derived features and self-attention mechanisms for enhanced prediction accuracy.
• To provide an efficient and accurate computational tool for PPI analysis.

Main Methods:

• Utilized amino acid composition (AAC), conjoint triad (CT), and auto covariance (AC) for feature extraction from protein sequences.
• Employed a deep neural network (DNN) architecture enhanced with a self-attention mechanism.
• Validated the model using 5-fold cross-validation on intraspecific and interspecific datasets, and independent datasets.

Main Results:

• Achieved high prediction accuracies: 95.48% (Saccharomyces cerevisiae), 98.94% (human intraspecific).
• Demonstrated strong performance on interspecific datasets: 93.15% (human-Bacillus Anthracis), 88.33% (Human-Yersinia pestis).
• Attained 100% accuracy on independent datasets (Caenorhabditis elegans, Escherichia coli, Homo sapiens, Mus musculus), outperforming previous methods.

Conclusions:

• SDNN-PPI effectively predicts protein-protein interactions using sequence encoding and self-attention deep learning.
• The method shows robust performance across intraspecific, interspecific, and cross-species predictions.
• SDNN-PPI offers valuable insights into PPI mechanisms and potential applications in drug design and disease prevention.