sign_matrix for runSpatialDeconv

[Jennifer-Foltz]

Good afternoon,

Thank you for the nice package. We are excited to try runSpatialDeconv with DWLS; however, I am unable to find document on what the format should be for the sign_matrix. What should the sign_matrix look like?

Thank you!

Jennifer

[ndelrossi7]

Hi @JenniferWUSTL! Have you used the function makeSignMatrixDWLS? The input is the Giotto object, signature genes, and cell types. This will create the correct format for the signature matrix. Let me know if this works or if you have any other questions!

[Jennifer-Foltz]

Hi @ndelrossi7 ! Thank you so much for your quick response! I have not used that function. Our reference dataset is a Seurat object, is there an alternative method you would suggest? Can we generate the signmatrix from raw data from our reference (e.g. export the rawdata from Seurat as a matrix)?

[ndelrossi7]

Hi @JenniferWUSTL - yes you can make your own matrix! The format is the columns are cell_types and the rows are the genes. I've attached a screen shot of an example here:

Let me know if you have any other questions!

[Jennifer-Foltz]

Hi @ndelrossi7 ! Great! Just one more question- are the numbers the average gene expression in each cell type? If so, is this logged or unlogged?

Thank you!

[ndelrossi7]

Hi @JenniferWUSTL! It's unlogged average gene expression for each cell type/spot. Hope this helps!

[Jennifer-Foltz]

@ndelrossi7 , thank you, that helps! I'm able to get the function to run now without error; however, I cannot locate the result output. I tried to find it using pDataDT() and fDataDT(), this was the code I used to run the deconvolution.

object <- runSpatialDeconv(gobject = object, deconv_method = "DWLS", expression_values = "normalized", logbase = 2, cluster_column = "leiden_clus_0.6", sign_matrix = sign_matrix, n_cell = 10, cutoff = 2).

Can you point me to how to access the deconvolution results? Thank you!!

[RubD]

hi @JenniferWUSTL I might be able to help you find the results. At the moment the easiest way to get to your results is like this: object@spatial_enrichment[["DWLS"]]. Let us know if this doesn't work.

[Jennifer-Foltz]

@RubD this works! Thank you so much! Just to confirm, the numbers in this df correspond to the proportion of a cell type in each spot?

[RubD]

@JenniferWUSTL I think they should. Do they add up to 1 (or 100) in each row (=spot)? @ndelrossi7 can probably say with more certainty.

[ndelrossi7]

@JenniferWUSTL yes that's correct! The values should sum up to 1.

[Jennifer-Foltz]

@RubD @ndelrossi7 thank you! They do sum to 1 :) . What is the significance of negative values for some cell types?

[ndelrossi7]

@JenniferWUSTL The negative values should be adjusted to 0 - we're looking into making this adjustment. Thanks for bringing it to our attention!

[Jennifer-Foltz]

You're welcome :).

Thank you both for the help! You've been amazing in getting this up and running, and really appreciate the prompt support!