BBLN: A bilateral-branch learning network for unknown protein-protein interaction prediction

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Predicting unknown protein-protein interactions (PPIs) is vital for biological analysis.
• Current methods struggle with limited data and the information monotonicity problem, hindering accurate prediction of novel PPIs.
• Leveraging multi-modal protein data (amino acid sequences, gene ontology) is essential for overcoming these limitations.

Purpose of the Study:

• To develop a novel method for enhancing the prediction of unknown protein-protein interactions (PPIs).
• To address the information monotonicity problem in existing PPI prediction models.
• To effectively integrate complementary and relationship information from diverse protein data modalities.

Main Methods:

• Proposed a bilateral-branch learning network to process multi-modal protein data.
• Utilized amino acid sequences and gene ontology (GO) information.
• Employed multi- and cross-modal learning strategies to capture complex relationships.

Main Results:

• The proposed bilateral-branch network significantly improved the prediction of both traditional and novel unknown PPIs.
• Experimental results on large-scale datasets demonstrated superior performance compared to state-of-the-art methods.
• The approach effectively enhanced complementary and relationship information extraction.

Conclusions:

• The bilateral-branch learning network offers a powerful solution for accurate unknown PPI prediction.
• Integrating multi-modal protein data through cross-modal learning is key to overcoming prediction challenges.
• This method advances the field of computational biology by providing a more robust PPI prediction tool.