Fig. 4: Surface-based transcriptomic decoding of the Benzodiazepine (BZ) binding sites.

a A high-resolution in vivo atlas of the human brain’s BZ binding site of GABA_ARs provided by ref. ²². b False positive rate (FPR) of the α-null (blue), Linear Mixed Effects (LME, red), and General Least Squares (GLS, yellow) decoding techniques across different adjusted p-value thresholds (p_adj). For LME- and GLS-decoding, p-values were adjusted for multiple comparisons using False Discovery Rate (FDR) adjustments. For α-null decoding, adjusted permutation p-values (p_perm.adj) were used. c Upper panel: Transcriptomic associations between BZ binding sites and mRNA expression signatures of individual GABA_AR subunits for different decoding approaches. α-null: spatial autocorrelation preserving null model; LME: Linear Mixed Effects model; GLS: General Least Squares model. Lower panel: predicted mRNA expression profile of significant subunits across the cortical surface. ^* adjusted p-value < 0.05 (two-tailed). d Hierarchical clustering of GABA_AR subunits based on their vertex-level transcriptomic associations. Subunits were allocated to two classes with a distinct pattern of expression. The first cluster (Cluster 1) contained subunit genes GABRA3, GABRA2, GABRA5, GABRB1, GABRE, GABRB3, and GABRG1 with high predicted expression levels in the limbic circuitry. The second cluster (Cluster 2) contained subunit genes GABRG3, GABRA4, GABRD, GABRB2, GABRA1, and GABRG2 with high predicted expression levels across the cortical surface. The left panel shows the predicted mean expression signatures across genes within each subunit cluster.