5/6 Functional network assignment can be predicted by cell-type abundances in post-mortem tissue. Motor, visual, and limbic networks, ventral attention/salient networks can be well-predicted. Default, frontoparietal/control, and dorsal attention networks are easier to be predicted as each other.

4/6: Large-scale networks possess distinct cellular profiles, hinting at computing motifs for specialized functions. For example, dorsal attention network is enriched by layer 2/3 and layer 5 ITs, whereas ventral attention/salient network is enriched by layer 6 neurons.

3/6: Multivariate cellular profiles track the functional gradient architecture of cortex. Permutational canonical correlation analysis, reveals linear combinations of cell type distributions that maximally correlate to each functional gradient.

2/6: Univariate cell type distributions align with gradient topographies: SNCG and SST interneurons and excitatory L5 IT are preferably localized at transmodal, and the L4 IT at the unimodal end. Cell types are imputed from gene expression in bulk tissue samples and single-nucleus data.

1/6: Here, we examine associations between the spatial abundance of cell-types with functional gradients and networks across the cortical sheet.