Group-Wise Hub Identification by Learning Common Graph Embeddings on Grassmannian Manifold

Area of Science:

• Neuroscience
• Network Science
• Computational Biology

Background:

• The human brain is a complex system with critical hub regions supporting major functions.
• Current methods for identifying common brain network hubs often lack reproducibility and ignore data geometry.

Purpose of the Study:

• To develop a novel group-wise hub identification method for human brain networks.
• To identify reliable hub nodes common across a population of individual brain networks.

Main Methods:

• Learning common graph embeddings representing local topological profiles.
• Utilizing Grassmannian manifold optimization with orthogonal embedding vectors.
• Developing a population-based common hub node map.

Main Results:

• The proposed method successfully identifies reliable hub nodes within individual networks.
• A population-based common hub node map was generated.
• Manifold learning approach demonstrated superior accuracy and replicability on synthetic and real network data.

Conclusions:

• The novel Grassmannian manifold optimization scheme provides a robust method for group-wise hub identification in brain networks.
• This approach enhances the reliability and reproducibility of findings in neuroscience.
• The method outperforms existing hub identification techniques.