Can NEST approach be applied to spatial correlation?

[K0ki116]

Thank you for your contribution to NEST!

My question is, can the NEST approach be applied to the analysis of the relationship between brain measurements and annotation maps (e.g., the neuroreceptor maps available in JuSpacce or Neuromaps)?

In my humble opinion, if we can analyze the network-level enrichment of the spatial relationship with the annotation maps (e.g., which neuroreceptors have similar spatial patterns to the subject atrophy maps especially in certain networks), this can be an interesting analysis.

Or, can this approach also be used to see if spatial correlation is more prominent in one group than the other?

Thank you in advance!

Koki

[smweinst]

Thanks for your interest in using NEST! Do you also have a (non-imaging) phenotype/covariate that you are quantifying an association with at each image location? In case it's helpful to reference, Figure 2 in the paper illustrate the general framework of NEST, including the input data needed for the method to be applicable. Feel free to reach out if you have any other questions! -Sarah

[K0ki116]

I appreciate your quick reply!

Thank you for sharing your figure. To be honest, I partly misunderstood the framework of NEST. Sorry for your confusion

However, I still wonder whether the [y1 ...... yj] needs to be constant across the vertices. I feel that in situations when T(v) can be calculated, whether [y1 ... yj] is constant across the vertices does not matter in the framework and thus NEST can be applied to the analysis of the relationship between a set of images of each participant. (Please tell me if I'm wrong... I'm new to bioinformatics)

For example, can we set T(v) as the correlation between brain atrophy and neuroreceptor density in each vertex v and perform NEST to see if the atrophy-receptor relationship is enriched in a particular network?

Koki

[smweinst]

No problem!

So far, we've only implemented NEST for testing associations with a phenotype that does not vary across vertices (e.g., participant age), but I think if you have two maps (with the same dimension) per person, and construct a spatial map of vertex-level associations (e.g., get a correlation at each vertex between map 1 and map 2 across participants), then you should be able to apply NEST. In the context of Figure 2, the only difference would be how you quantify T(v) in A and B, but everything else should be the same.

We also have a couple of other methods which might be relevant for testing between-map associations (SPICE, CLEAN-R) using subject-level data. These methods don't test enrichment in the way we've defined it in the NEST paper, but they test similarity between two subject-level brain maps 'globally' within a network (SPICE) or at the vertex level (CLEAN-R) of associations between subject-level brain maps.

Hope this helps, and happy to answer any additional questions!

-Sarah

[K0ki116]

Thank you for your clear answer!