Detecting outlier in single cell multiome data.

Question

[mandecent.gupta]

Hi,

I am new to single cell sequencing world and working on Single Cell multiomics (ATAC + GEX) data analysis. Can someone please share the relevant method (or publication/R package) for detecting outlier in Single Cell multiome.

Thanks

Answer 1

[V]

Have a look at the second part of this Seurat tutorial, found here.

[mandecent.gupta]

That is just "WNN analysis of 10x Multiome, RNA + ATAC" but no explanation of outlier detection. Please correct me if I am wrong.

[V]

An 'outlier' would be a cell that is a detected and filtered out using multiple metrics. Such as if it has a very high/low number of genes detected, or high mitochondrial content detected, or in your case, a very high or a very low number of peaks detected..... all of which are shown how to detect and filter out in the above.

[mandecent.gupta]

Is there any statistical method for finding the cutoff for low/high number of peaks or detecting high/low level of gene expression?

[V]

As far as I am aware, no. In part because this is very cell type specific, so would be very difficult to benchmark. Additionally different technologies vary in the number of genes they can detect (on average) - and we havent even considered the effects of sequencing depth. So, in short there are too many variables.

[V]

Please dont forget to upvote my original answer as I believe it addresses your question about how to remove outliers. Thanks