Failed to identify T cells in Visium data

[vwtlin]

Hi there,

I used DWLS to predict the composition and failed to identify T cells in the Visium data. The single-cell dataset I used comprises enough cell types that could exist in the Visium samples. I did check the expression of some standard marker genes of T cells in the Visium data, and I could see the expression. I would like to ask for your comments on how I could improve this.

Cheers,
Weitao

[RubD]

I would suggest to start with signatures for a few key cell types (e.g. T-cell, B-cell, etc) and then evaluate your results. Using too many cell types often times lead to suboptimal results for any deconvolution approach.

[vwtlin]

@RubD Thanks for your reply. I did try using the main cell types for deconvolution but the result did not get improved.

An additional question about the DWLS matrix, what would you suggest to make a signature gene list? I was using differentially expressed genes obtained from DE analysis, together with some marker genes from the literature. Does the signature gene need to be unique to one cell type only?

Following that with a similar question, what would you recommend to get the signature list for PAGE enrichment analysis?

Cheers,
Weitao