Mapping Cortical Gradient Collapse Under DMT influence

To mathematically map the global alterations in brain organization induced by N,N-dimethyltryptamine (DMT), researchers deploy principal cortical gradients—a continuous, low-dimensional coordinate space that represents the macroscale functional hierarchy of the cerebral cortex 1, 2.

1. The Computational Methodology for Extracting Cortical Gradients

To translate raw functional magnetic resonance imaging (fMRI) blood-oxygen-level-dependent (BOLD) signals into a mathematical gradient, researchers implement a multi-step pipeline using specialized toolboxes such as BrainSpace in MATLAB 3:

Dimensionality Reduction and Vertex Downsampling: The cortical surfaces are first downsampled from saverage 5 space (which contains 20,484 vertices) to 10,000 vertices to maximize computational efficiency 3.

Pairwise Correlation Profiling: For each subject, a 10,000 × 10,000 functional connectivity (FC) matrix is computed by calculating the pairwise Pearson’s correlation between the BOLD time series of all vertices 3.

Sparsity Thresholding: This z-transformed correlation matrix is thresholded row-wise at 90% sparsity 3. This step retains only the strongest, most topologically significant connections and discards background noise 3.

Cosine Similarity Mapping: Researchers compute cosine similarity on the thresholded z-matrix 3. This generates a similarity matrix that captures the exact similarity in whole-brain connectivity patterns between any two given vertices 3.

Manifold Learning via Diffusion Map Embedding: The similarity matrix is passed into a nonlinear manifold learning algorithm known as diffusion map embedding, which belongs to the family of graph Laplacians 3. This algorithm is governed by a single parameter, \(\alpha\) (set to 0.5), which balances the influence of both global and local relationships on the manifold 3. The output is a set of low-dimensional "gradients" of continuous variation across the cortex, where spatial distance in this embedding space corresponds directly to the similarity of functional connectivity profiles 3.

Procrustes Alignment and Group Contrasts: To enable statistical comparisons across subjects, a Procrustes rotation is performed to align individual-subject gradients to an all-subjects group average template, ensuring that the gradient axes are matched 3. Group contrasts and behavioral associations are subsequently conducted using surface-based linear models in the SurfStat toolbox 3.

2. The Architecture of the Principal Sensory-Association Axis

The principal axis (Gradient 1) generated by this diffusion embedding algorithm represents the fundamental sensorimotor-to-association cortex gradient of the human brain 2, 4. This axis organizes brain regions along a continuous spectrum:

The Unimodal Pole: Situated at the lower end of the gradient, this pole contains lower-order unimodal systems, such as the somatomotor (SM) and primary visual (VIS) cortices, which are dedicated to direct sensory processing 2, 4.

The Transmodal Pole (TOP): Situated at the upper end of the gradient, this pole contains highly integrated, transmodal association networks—specifically the Default Mode Network (DMN) and Frontoparietal (FP) network 2, 4, 5. The transmodal association pole (TOP) is evolutionarily and ontogenetically recent, expanded in humans, and is responsible for abstract semantic representations, language, and self-referential cognitive processing 4, 6-8.

3. Hierarchical Collapse and Compression Under DMT

When healthy volunteers are administered a high dose of intravenous DMT, researchers observe a profound whole-brain functional compression and collapse of this principal cortical gradient 2, 9, 10.

Axis Compression Towards Zero: Under the acute influence of DMT, high-order TOP regions display significantly reduced gradient scores, while low-level unimodal regions display increased gradient scores 2. This simultaneous shift of both poles toward the center of the gradient accounts for a dramatic overall compression of the sensory-association axis 2.

Desegregation of the Hierarchical Architecture: This compression mathematically represents a loss of macroscale hierarchical differentiation 2. The spatial boundaries that ordinarily segregate lower-order sensory networks from higher-order transmodal systems collapse 2, 9. As a result, the TOP becomes less segregated from—and more integrated with—the rest of the cortex 4, 11.

Network-Wise Contrasts: Radial and vertex contrasts confirm this structural leveling 2. Under DMT, gradient scores increase significantly within unimodal networks (visual and somatomotor) as well as the dorsal attention network (DAN), while scores decrease significantly within high-level transmodal networks (frontoparietal, default mode, and limbic) 2.

4. Neurobiological and Phenomenological Correlates

The principal cortical gradient acts as a physical bridge linking molecular pharmacology, systems-level neurodynamics, and subjective experience:

Co-localization with Serotonin \(5\text{-}HT_{2A}\) Receptors: Dynamic global functional connectivity (GFC) maps under DMT strongly correlate with independent positron emission tomography (PET) density maps of the \(5\text{-}HT_{2A}\) receptor 1, 12. Because these receptors are most densely expressed on layer V pyramidal neurons in the TOP, DMT selectively dysregulates the highest levels of the brain's predictive hierarchy 6, 8, 12, 13.

Decompression of the Mind-at-Large: Ordinarily, the brain's highest-level networks act as "compressive models," utilizing heavily-weighted top-down prior beliefs to constrain, filter, and explain away bottom-up prediction errors 14-16. By collapsing the principal gradient, DMT impairs this top-down compressive function 9, 16.

The Anarchic Brain State: Electrophysiologically, this gradient collapse is mirrored by a localized collapse of parietal alpha waves (8–12 Hz), which normally act as an inhibitory gate 10, 17, 18. Deprived of alpha gating and DMN modular integrity, the brain transitions into an "anarchic," highly entropic, and hyperconnected state 8, 10, 17, 19. Subjectively, this functional compression of the principal gradient maps directly onto the dissolution of baseline space-time orientation, the loss of bodily boundaries, and ego dissolution 1, 9, 10.

🔬 We could evaluate how this functional gradient compression correlates with the spatial binding potential of other receptor systems, such as the \(5\text{-}HT_{1A}\) receptor, across these same network vertices 6, 20.