Two peak in histogram before single cell RNA-seq QC & Filtering

Question

[joonhong kwon]

Hi all,

I am analyzing the scRNA-seq data on immune cells in brain and PBMC. First, before QC and Filtering, I checked the distribution of nFeature (nGene) and uCount (nUMI) with histogram. But, in PBMC data, two peaks were identified in the histogram of nFeature. One occurred at very low values, such as 100, and the other occurred at values above 1000.

I'm newbie, so I haven't been able to deal with a lot of data. Is this a common result? Or is there something wrong with the data? In order to remove low quality cells or empty droplets, cells with low expression are usually removed (ex nFeature < 100 ~ 500). However, I don't know what to do in this situation. How should the criteria be set for removal of low quality cells or empty droplets?

[joonhong kwon]

Bl is Blood (PBMC), Br is Brain, CT is Control, GT is Case.

Answer 1

[jared.andrews07]

Looks to me like your blood samples are just lower quality/have more empty droplets/dead cells. I'd remove those two lower peaks outright and use a MAD-based threshold approach for the rest.

[joonhong kwon]

Thank you for reply! So, your answer means that the peaks at the low value is not due to wrong data or the characteristics of the blood sample, but just because there are many low quality cells (empty droplets/dead cells). Right?

[jared.andrews07]

That's my interpretation, though other QC metrics (mito %, # cells, nCount) would be helpful as well. Generally, if you fail one metric, you fail multiple for a given droplet.

[joonhong kwon]

Thanks a lot!!