Robust Identification of Partial-Correlation Based Networks with Applications to Cortical Thickness Data

Area of Science:

• Neuroscience
• Computational Biology
• Graph Theory

Background:

• Understanding brain development and organization relies on analyzing correlations between structural and functional brain measures.
• Partial correlations help reduce ambiguity by isolating relationships between brain regions independent of other interacting regions.

Purpose of the Study:

• To introduce and investigate PC*, a novel graph pruning algorithm for identifying partial correlation networks.
• To compare PC* with direct calculation of partial correlations from the inverse of the sample correlation matrix.

Main Methods:

• Developed and applied the PC* graph pruning algorithm.
• Utilized partial correlations to identify significant relationships between brain regions.
• Analyzed covariation in cortical thickness within regions of interest (ROIs) on a parcellated cortex.

Main Results:

• The PC* algorithm was shown to be significantly more robust than direct calculation methods.
• PC* effectively identifies subnetworks obscured in standard correlation data.
• Demonstrated the utility of PC* in studying covariation patterns in cortical thickness.

Conclusions:

• PC* provides a robust method for uncovering partial correlation networks in brain data.
• The algorithm aids in revealing complex inter-regional relationships and subnetworks.
• PC* is a valuable tool for investigating brain organization and development through structural covariation analysis.