Graph Statistics in the Brain

Inherent variability within a single network or  populations of networks is an increasingly desirable phenomenon to characterize. Implications of this can be used to improve diagnosis or discovery of possible causes for neurological disorders, improving artificial intelligence by modeling it on how the brain operates, and understanding how network distinctions among brains can manifest into phenotypic differences. To approach this problem, a connectome (a connected graph – directed or undirected – that models connections) can be constructed to show relationships between brain regions. The GraSPy python package has been created to tackle the problem of how to analyze these constructed connectomes. In our project, we added to the functionality of GraSPy with new tools for classifying graphs for multiple outcomes, graph matching, covariate assisted spectral embedding, & network classification using signal-subgraphs. We venture to ask such questions as ‘Does our brain connectivity influence our phenotypes?’ and ‘Do phenotypes embed something in common in terms of brain connectomes?’ If the answer is yes, then the question arises to what extent they are similar. Can we measure the similarities between two connectomes?