Spot deconvolution: SPOTlight option

[lcolladotor]

https://bioconductor.org/packages/release/bioc/html/SPOTlight.html

• Likely drop rare snRNA-seq cell types
• Can start exploring how to run it by using the Tran et al, Neuron, 2021 data
• Create a bash script with sgejobs such that we can run it later (or series of scripts) once we have the n = 19 snRNA-seq data
• Potentially write code in a way that the input SCE object can be adapted to other ones, to make it easier to run with psychENCODE datasets in the future
• SPOTlight has several diagnostic plots that would be very useful to make
• It can be an iterative process if there's collinearity (aka high similarity) between two cell types: drop one and re-run

We'll want to run this with both the SPE from the n = 30 Visium samples and the n = 4 Visium IF samples

Could be useful to adapt some of Lukas's LC code from https://github.com/lmweber/locus-c/search?q=spotlight.

[lcolladotor]

Export results in a CSV file that has the key column + 1 column for every cell type. Similar to what https://github.com/LieberInstitute/spatialLIBD/blob/master/R/cluster_export.R does, so we can then read into the SPE objects using code similar to https://github.com/LieberInstitute/spatialLIBD/blob/master/R/cluster_import.R.

[lcolladotor]

We'll want to check if the results are concordant when you run it at the broad cell type level vs the specific one. We'll add up the specific results (like pseudo-bulk) to then compare vs the broad cell type results. This is something similar to what @lahuuki has done for deconvolution of bulk RNA-seq via snRNA-seq reference datasets.

[lcolladotor]

We are re-assigning this one to @Nick-Eagles