Overlapping functional modules detection in PPI network with pair-wise constrained non-negative matrix...

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• High-throughput experiments generate vast protein-protein interaction (PPI) data.
• Identifying functional modules in PPI networks is crucial for understanding cellular mechanisms.
• Existing unsupervised community detection methods lack the ability to integrate prior knowledge of protein complexes.

Purpose of the Study:

• To develop a novel semi-supervised model for identifying overlapping functional modules in PPI networks.
• To leverage known protein complexes as a priori information to enhance module detection.
• To simultaneously model protein module memberships and inter-module correlations.

Main Methods:

• Proposed a novel semi-supervised model: pairwise constrains nonnegative matrix tri-factorisation (PCNMTF).
• PCNMTF utilizes protein module indicator and module correlation matrices derived from PPI networks.
• Employs non-negative matrix tri-factorisation to learn mixed module memberships and module correlations.

Main Results:

• PCNMTF successfully identifies overlapping functional modules by incorporating known protein complexes.
• Experiments on synthetic and real-world biological networks show improved precision.
• The model effectively captures both individual protein module associations and relationships between modules.

Conclusions:

• PCNMTF offers a more precise approach to functional module detection in PPI networks compared to state-of-the-art methods.
• The semi-supervised approach enhances biological interpretability by integrating prior biological knowledge.
• This method advances the understanding of cellular functions through improved network module analysis.