Some questions about xCell and xCellView about xcell HOT 2 OPEN

Hi Yiwei,

Thank you for your questions. If it's not clear to you, it's probably not clear to many others...

I don't understand this.
Regarding this issue, scaling the scores by samples is extremely dangerous and will inevitably will result in false interpretations.
To my understanding, we often use row "scale" to visulize the results in heatmap, such as the Fig.4a from the xCell paper. So I don't get what your mean here.

Since xCell is not a deconvolution method, but a cell type enrichment method, it does not claim that the scores across cell types are comparable - for example, for a given sample, it cannot be claimed based on xCell that CD4+ T-cells > CD8+ T-cells. Thus, what I mean in this note is not to do z-score on samples. If you must scale the scores do it on the cell types (all samples for a given cell type). However, this is only needed for heatmap visualizations, for other analyses, it is best to leave the scores as is.

There are 5 seletions of "Choose gene signatures" of xCell web server. The paper of xCell used "xCell (N=64)", and what's the others? Which one should be used?

The other signatures are signatures from other studies we compared xCell with in the manuscript.

xCellView has an option "Filter weak signatures". How does this work?

Not all signatures were born equal - some cell types were not available in sequencing-based methods, some had only a handful of samples, and some are highly variable across datasets. Thus, the results from some scores are not as relaible as others. The flitering is manual filtering based on my observations of problematic cell types.

from xcell.

Thank you for your quick and informative reply! I understand now.

Adding some usage examples in the documents, especially the interpretation of the results, will certainly be helpful to beginners like me. 😄

from xcell.