Our latest paper, out today @NatureNeuro: "Hierarchy of transcriptomic specialization across human cortex captured by structural neuroimaging topography".
rdcu.be/36Dg
rdcu.be/36Dg
Big Q: How are cortical areas specialized across the human brain?
An important axis relates to sensory-association cortical hierarchy. Anatomical hierarchy is conventionally measured by invasive tract-tracing, so inaccessible for human cortex. What's a usable proxy measure?
An important axis relates to sensory-association cortical hierarchy. Anatomical hierarchy is conventionally measured by invasive tract-tracing, so inaccessible for human cortex. What's a usable proxy measure?
The noninvasive MRI-derived T1w/T2w map is a good candidate for a hierarchy proxy measure, as it exhibits high values in sensory areas and low values in association areas. There's good homology between human and macaque, for which anatomical hierarchy can be measured.
And indeed, in macaque, T1w/T2w values are strongly negatively correlated with anatomical hierarchy levels (derived from tract-tracing). So maybe we can use the T1w/T2w as a noninvasive proxy, applicable in humans.
Now back to the Big Q of cortical specialization. In humans, we can analyze brain-wide patterns of gene expression, from the @AllenInstitute Human Brain Atlas (AHBA). Microcircuit specialization, e.g. in cell types or synapses, is reflected in expression levels of related genes.
Consistent with ground-truth anatomical measurements in monkeys, we find strong correlations with the T1w/T2w map for genes related to various aspects of cortical microcircuit specialization: lamination, inhibitory interneuron cell types, and synaptic receptors.
What about genes in general? If we do PCA, PC1 is the spatial map which capture the most variation in gene expression. PC1 captures a lot of variance (cortical gene expression looks quasi-1D). Remarkably, PC1 is strongly correlated with the T1w/T2w map.
So to 'zeroth order', cortex is different than thalamus/striatum/cerebellum/etc. To 'first order', cortical areas differ from each other following a hierarchical sensory-association axis, captured by the MRI-derived T1w/T2w "myelin" map.
P.S. If you're interested in how we can use these cortical gradients in large-scale computational models of the human brain, take a look at the @muratdemirtas_ preprint: biorxiv.org/content/early/…
