The protein-protein interaction network alignment using recurrent neural network

Area of Science:

• Bioinformatics
• Computational Biology
• Network Science

Background:

• Biological network alignment is crucial for understanding cellular pathways, drug discovery, and disease recognition.
• The computational complexity of biological network alignment, often NP-hard, presents a significant challenge.
• Protein-protein interaction (PPI) networks are vital for various biological and medical applications.

Purpose of the Study:

• To develop a novel deep learning-based method for biological network alignment.
• To address the NP-hard nature of network alignment by reformulating it as a classification problem.
• To introduce RENA (Network Alignment using REcurrent neural network) for accurate node alignment in biological networks.

Main Methods:

• RENA transforms the network alignment problem into a classification task.
• The method involves three phases: extracting sequence and topological similarities, creating a classification dataset, and predicting node alignments using deep learning.
• Utilized a recurrent neural network architecture for the classification of node alignments.

Main Results:

• The RENA method demonstrated high efficiency in predicting protein-protein interaction network alignments.
• Achieved 100% accuracy in predicting node alignments within PPI networks.
• Outperformed traditional classification methods like Support Vector Machine, K-Nearest Neighbors, and Linear Discriminant Analysis.

Conclusions:

• RENA offers an effective and accurate deep learning solution for biological network alignment.
• The classification-based approach significantly improves the prediction accuracy of node alignments.
• This method holds promise for advancing bioinformatics research and applications, including drug discovery and disease understanding.