Conceptual check-in: Avg Pearson residual as gene set score?

[dpcook]

Hey Christoph,

Quick question. I'm just thinking about gene set scoring strategies. As an alternative to AddModuleScore's approach to scoring, do you think it makes sense to make a score from a cells average Pearson residuals for the gene set? ie. colMeans(seurat[["SCT"]][gene_set,])

I did a quick comparison on a data set and it actually looks pretty consistent with AddModuleScore():

[satijalab]

Thats an interesting point. Pearson residuals are basically asking how much of an 'outlier' a cell is for a given gene, given the average expression of that gene in the population. I never thought about it, but AddModuleScore is doing a similar thing (without an underlying statistical model). Great to see that they correlate, and I think it certainly makes sense

[dpcook]

Thanks! Yeah, from my understanding of the methods, I thought they conceptually addressed the same question, but with different approaches. Using SCT for the rest of the analysis, I liked the idea of using the same model for generic gene set scoring, but wanted to make sure I wasn't missing anything obvious.