DeepTrio: a ternary prediction system for protein-protein interaction using mask multiple parallel convolutional...

Area of Science:

• Computational Biology
• Bioinformatics
• Machine Learning

Background:

• Protein-protein interactions (PPI) are crucial for cellular functions, but predicting them accurately is challenging due to their relative nature.
• Existing PPI prediction models often lack unbiased architectures and struggle to differentiate between relative and intrinsic properties.

Purpose of the Study:

• To develop a novel, high-performance, sequence-based computational model for predicting protein-protein interactions.
• To enhance the generalization capabilities of PPI predictors and enable discrimination between relative and intrinsic properties.

Main Methods:

• A sequence-based approach named DeepTrio was developed.
• DeepTrio utilizes mask multiple parallel convolutional neural networks for PPI prediction.
• The model was evaluated using various quality metrics against state-of-the-art methods.

Main Results:

• DeepTrio demonstrated superior performance compared to existing state-of-the-art methods in PPI prediction.
• The model achieved high accuracy across multiple quality metrics.
• DeepTrio provides insights into the contribution of individual neurons to prediction outcomes.

Conclusions:

• DeepTrio represents a significant advancement in sequence-based PPI prediction.
• The developed model offers improved accuracy and interpretability for PPI prediction.
• The study provides an accessible online application and open-source code for the DeepTrio model.