Skip to main content
. 2023 Jul 13;12:e83843. doi: 10.7554/eLife.83843

Figure 1. Organization of the cortical receptome.

(A) Analytic workflow of receptome generation and gradient decomposition. Node-level neurotransmitter receptor and transporter molecule (NTRM) fingerprints are derived from PET images of 19 different NTRM (in the top left, italic font denotes transporters). The fingerprints are then Spearman rank correlated to capture node-level similarity in chemoarchitectural composition, generating the receptome matrix. Next, to determine similarity between all rows of the receptome matrix, we used a normalized angle similarity kernel to generate an affinity matrix. Finally, we employ diffusion embedding, a nonlinear dimensionality reduction technique, to derive gradients of receptomic organization. (B) Receptome (RC) gradients projected on the cortical surface. Top: first receptome gradient (RC G1); middle: second receptome gradient (RC G2); bottom: third receptome gradient (RC G3). (C) Spearman rank correlations of cortical receptome gradients with individual NTRM densities. Top: first receptome gradient; middle: second receptome gradient; bottom: third receptome gradient. Saturated blue coloring corresponds to statistically significant correlations at p < 0.05.

Figure 1.

Figure 1—figure supplement 1. Cortical receptome gradients.

Figure 1—figure supplement 1.

(A) The first three chemoarchitectural similarity axes generated by principal component analysis (PCA). Spearman rank correlations to the respective gradients exceeds r > 0.99, indicating a high similarity between axes derived using linear and nonlinear dimensionality reduction methods. PCA-derived chemoarchitectural similarity axes replicate findings by Hansen et al., 2022. (B) Significance of receptome gradients compared to gradients derived from randomized neurotransmitter receptor and transporter molecule (NTRM) density profiles. The blue circles indicate the eigenvalues of the true components, the boxplots display the eigenvalues of components resulting from performing gradient decomposition of n = 1000 randomized receptomes. Up to the 11th component, the true components display significantly higher eigenvalues (p<0.05).
Figure 1—figure supplement 2. Robustness of receptome gradients.

Figure 1—figure supplement 2.

Robustness of receptome gradient decomposition across different parcellation granularities. Left: RC G1, RC G2 and RC G3 (top-to-bottom) projected on the cortical surface. Middle: variance explained by gradient decomposition. Right: receptome matrix. (A) 100 parcels, (B) 200 parcels, (C) 300 parcels, and (D) 400 parcels.