How do I create the extdata/sig_matrix.txt equivalent of my own data?

[sayadennis]

Hi Giotto community!

I'm quite new to this field so I apologize for the basic question. How is the extdata/sig_matrix.txt file generated?

What I am trying to accomplish is to run runDWLSDeconv() on my giotto object. This function requires a sign_matrix= argument, which I'm not sure how to generate. I'm guessing this is something similar to sig_matrix.txt, because I tried to run the function with sign_matrix= as this, and the only error I got is that there are no overlapping gene names between my giotto object and the signature matrix. (Which is not too surprising, considering the data comes from totally different sources.)

I was wondering if the signature matrix can be generated with makeSignMatrixDWLS(), but I haven't been able to get this to work either. Specifically, I'm not sure what should go in the sign_gene= and cell_type_vector= variables in this function. Can we generate these variables from the Giotto existing object?

Thank you in advance for your time! Any input is very much appreciated. I'm also happy to provide any further information if anything is unclear.

[sayadennis]

I think I figured this out with help from someone else, so leaving a record here for future reference. The instructions are from here: https://github.com/rdong08/spatialDWLS_dataset/blob/main/codes/Heart_ST_tutorial_spatialDWLS.Rmd

scran_markers_subclusters = findMarkers_one_vs_all(gobject = visium_brain,
                                                   method = 'scran',
                                                   expression_values = 'normalized',
                                                   cluster_column = 'leiden_clus')
Sig_scran <- unique(scran_markers_subclusters$genes[which(scran_markers_subclusters$ranking <= 100)])

norm_exp <- 2^(visium_brain@norm_expr)-1
id <- visium_brain@cell_metadata$leiden_clus
ExprSubset <- norm_exp[Sig_scran,]
Sig_exp <- NULL
for (i in unique(id)){
  Sig_exp <- cbind(Sig_exp,(apply(ExprSubset,1,function(y) mean(y[which(id==i)]))))
}
colnames(Sig_exp) <- unique(id)

## Perform deconvolution
visium_brain <- runDWLSDeconv(visium_brain, sign_matrix = Sig_exp, n_cell = 20)