Single Cell RNA-seq least expressed genes with seurat FindVariableFeatures: understanding variance

Question

[NGS_enthusiast]

(edited) Hi, I'm new to single cell analysis. I hope this is not too basic a question. I've followed the pipeline as in the tutorial of Satijalab. Now for the purpose of a control I need to run before doing my own single cell experiment, I'd like to find a list of least expressed genes (that are still "detectable") among the non-variable or least variable ones in a single cell experiment. it appears the function FindVariableFeatures creates the table I need with 3 columns: Mean, variance and variance.standardized. Could you tell me which column is best to defind non- and least variable genes, which threshold should I use? Once I set the threshold, I guess I can just sort on the mean (average expression) column.

thanks a lot for any help, best regards

Answer 1

[NGS_enthusiast]

Here is what I did that seems to do the trick, but I still have a question: I got a matrix with 14000 genes supposedly non-variable but it seems too high. How do I know which genes are actually detected (validated) in single cell? only those with mean >1 (indeed, I don't know how the mean can be 0.5 or 0.2 if it is non-variable...)?

Variance = HVFInfo(object = x)
Non_variable_variance <- Variance [Variance [,"variance.standardized"] < 1,]
Non_variable_sorted <- Non_variable_variance [order(Non_variable_variance[,1],decreasing=FALSE),]

I put In Variance the table with mean and variance and variance_staandardized I figures from plots in tutorial that variance_standardized should be <1 for non-variable genes hence the second line and finally I sort on the mean which is the first column.

But problem as I said above, there are way too many genes. How do we get to know which genes are actually considered "detected"? (and sorry in the meantime I asked a question that seemed different but know that I edited this post it is quite similar in the seurat discussion, I'll let you know if I get an answer there)