deconvolution question (unidentified cell types)

[chrkuo]

Thank you for creating such a beautiful tool. I have a question regarding the deconvoluted results

What are "unidenfitifed cell typs" and "unknown macrophages" are these from the reference cell type? and what are the reference cell type defining genes that are used to annotate the cell types?

Thank you!

[beibeiru]

Hi, @chrkuo ,

Thanks for showing a firm interest in our package.

Q1. What are "unidentified cell types" and "unknown macrophages" are these from the reference cell type? Since the in-house single-cell reference of SpaCET might miss some specific cell-types in the ST sample to be analyzed, an unidentifiable component reflects the cell fraction of these missing cell-types.

Similarly, the unknown macrophage component reflects the cell fraction of other macrophage sublineages other than macrophage M1 and M2.

Q2. What are the reference cell type-defining genes that are used to annotate the cell types? You can download combRef_0.5.rda from https://github.com/data2intelligence/SpaCET/tree/main/inst/extdata. Then, run these commands.

load("combRef_0.5.rda")
Ref$sigGenes

Please let me know if you have any further questions.

Best, Beibei

[chrkuo]

thank you so much for the prompt response. that makes sense. I will play with the tool a bit more and compare integrating a scRNAseq data to see how it compares.

the result of my ligand receptor was interesting but it seems like noting was significant -

This can happen for multitude of reasons correct?

[beibeiru]

Hi, @chrkuo ,

Welcome to try our tool on other data. Please let me know if you have any questions.

In our software, very stringent conditions are used to evaluate cell-cell interactions, including (1) cell-cell colocalization and (2) L-R network score difference. This aims to reduce the false positive rate.

Best, Beibei

[chrkuo]

@beibeiru is there anyway to loosen the criteria? or manipulate it? But from that data-set it seems like no interactions are occuring is that a correct assumption.

[beibeiru]

An alternative way is that you might only use the co-localization analysis results. Then, for the significant co-localized cell-type pairs, you can explore whether these cell-type pairs have several L-R pairs identified from your matched scRNA-Seq data.

Best, Beibei