Z-score for scRNA-seq data?

Question

[S]

Hi,

I was reading up on Z-score and how it is commonly used for pathway analysis comparisons in bulk-seq/pseudobulk data. I want to make a heatmap for my pathway analysis results using z-scores but I did not pseudobulk my scRNA-seq data during GSEA analysis.

It seems like it is not possible to calculate z-scores without pseudobulking my data, is that correct? If that is so, is there an equivalent metric I can use to generate a similar heatmap? Would the results be better if I pseudobulk?

Thank you!

[rfran010]

I don't think that's correct. For a given gene, z-score of scRNA-seq data could be expression in one cell scaled by mean expression and standard deviation among all cells. The math is the same, but instead of z-score across conditions, you would do across cells+conditions. A zscore across genes (so gene expression scaled by mean expression and standard deviation among all genes within a cell, or even within the whole population) could also be possible, but I think that's less common since expression is usually normalized already.

[S]

Thank you for your response! And I see what you mean, and I think my question was kind of vague but I am trying to do a pathway analysis comparison across conditions. So the enrichment comparison essentially, and the comparison would be interesting across conditions.

I could use the NES and do a celltype comparison for that pathway, but yes that won't work across conditions. So I was wondering if there was a way to achieve what the z-score achieves, which would be whole pathway comparison across conditions/phenotypes. Thanks again!