Mapping topological abnormalities in cortical similarity networks to schizophrenia-associated gene expression

Disrupted structural connectivity is recognized as a key pathophysiological feature of schizophrenia (SCZ). However, the relationship between cortical similarity network alterations and gene expression remains poorly understood.

We applied the Morphometric INverse Divergence framework to T1-weighted MRI from 1,216 participants. Cortical similarity networks were constructed, and global and nodal metrics were computed. Case-control comparisons were performed using linear models. Partial least squares (PLS) regression identified genes associated with spatial patterns of network alterations using Allen Human Brain Atlas data.

Among global metrics, only the rich-club coefficient differed between groups, with a negligible effect size. Nodal metrics showed reduced eigenvector centrality and k-coreness centrality in left temporal/insular (somatomotor), lateral occipital (visual), anterior cingulate (salience/ventral attention), and posterior cingulate (default mode) regions. Participation coefficient was widely reduced in SCZ. For each nodal metric, we identified PLS2-positive and PLS2-negative gene sets. Across degree, eigenvector centrality, and k-coreness centrality, PLS2-negative genes were enriched for metal ion transport, linked to manic and nonorganic psychosis, and upregulated in adolescence and early adulthood. PLS2-positive genes were enriched for neuron projection development and learning or memory, but not psychotic disorders.

These findings highlight synaptic and neurodevelopmental mechanisms underlying structural dysconnectivity in SCZ.