Multi-source biological knowledge-guided hypergraph spatiotemporal subnetwork embedding for protein complex...

Area of Science:

• Systems biology
• Bioinformatics
• Computational biology

Background:

• Protein complexes are crucial for cellular functions and disease mechanisms.
• Existing protein-protein interaction (PPI) network analysis methods are limited by static data and pairwise relationship assumptions.
• Dynamic and higher-order interactions in biological systems are not fully captured by current models.

Purpose of the Study:

• To propose HGST, a multi-source biological knowledge-guided hypergraph spatiotemporal subnetwork embedding method.
• To overcome limitations of static PPI networks and model non-pairwise interactions.
• To identify biologically significant protein complexes more effectively.

Main Methods:

• Constructing spatiotemporal PPI subnetworks incorporating protein dynamics and multi-source knowledge.
• Transforming subnetworks into hypergraphs to model higher-order interactions.
• Integrating amino acid sequence and gene ontology features for multi-dimensional representation.
• Identifying protein complexes using a core-attachment strategy on reweighted subnetworks.

Main Results:

• HGST demonstrated competitive performance across four real PPI datasets.
• The method successfully identified protein complexes with high biological significance.
• Biological analyses validated the effectiveness of HGST in complex identification.

Conclusions:

• HGST offers an advanced approach for identifying protein complexes by leveraging spatiotemporal dynamics and higher-order interactions.
• The integration of multi-source biological knowledge and multi-dimensional features enhances the accuracy and biological relevance of identified complexes.
• This method provides a valuable tool for understanding protein functions and disease mechanisms.