Stable Overlapping Replicator Dynamics for Brain Community Detection.

Area of Science:

• Neuroscience
• Network Science
• Computational Biology

Background:

• Understanding brain organization relies on identifying functional subnetworks.
• Existing methods often create distinct partitions, failing to capture overlapping brain regions.
• Brain subnetworks are known to exhibit overlap, necessitating new analytical approaches.

Purpose of the Study:

• To develop a novel technique for identifying overlapping subnetworks in weighted graphs.
• To incorporate statistical control for false node inclusion in subnetwork detection.
• To improve upon existing community detection methods for brain network analysis.

Main Methods:

• Proposed a technique based on replicator dynamics, incorporating graph augmentation for overlaps.
• Implemented a graph incrementation scheme to merge potentially split subnetworks.
• Integrated stability selection for statistical control over false node inclusion, termed Stable Overlapping Replicator Dynamics (SORD).

Main Results:

• SORD demonstrated significantly higher accuracy in subnetwork identification on synthetic data compared to state-of-the-art methods.
• Achieved higher test-retest reliability in network measures using Human Connectome Project data.
• Successfully identified neuroanatomically meaningful subnetworks and network hubs.

Conclusions:

• SORD provides an accurate and reliable method for identifying overlapping functional brain subnetworks.
• The technique offers improved statistical control and neuroanatomical interpretability.
• SORD advances the analysis of complex brain network structures.