Identifying HIV-induced subgraph patterns in brain networks with side information

Area of Science:

• Neuroscience
• Graph Theory
• Machine Learning

Background:

• Brain connectivity networks are crucial for neurological disorder identification.
• Current methods primarily rely on graph representations, often ignoring valuable supplemental data.
• Clinical, immunologic, serologic, and cognitive measures offer rich information for diagnosis.

Purpose of the Study:

• To develop a method for subgraph selection from brain networks guided by multi-view side information.
• To identify optimal subgraph patterns for enhanced graph classification in neurological disorder diagnosis.
• To leverage supplemental data to improve the accuracy of diagnostic models.

Main Methods:

• Proposed a feature evaluation criterion, gSide, to assess subgraph pattern usefulness based on side views.
• Developed a branch-and-bound algorithm, gMSV, for efficient subgraph pattern searching.
• Integrated subgraph mining with discriminative feature selection for optimal pattern discovery.

Main Results:

• Empirical studies demonstrated significant improvements in graph classification performance for neurological disorders.
• Selected subgraph patterns were found to be relevant for disease diagnosis.
• The multi-side-view-guided approach effectively boosted diagnostic accuracy.

Conclusions:

• Integrating multi-view side information with brain network analysis enhances neurological disorder identification.
• The gSide criterion and gMSV algorithm provide an effective framework for subgraph selection.
• This approach offers a promising avenue for improving diagnostic tools in clinical neuroscience.