quality control scRNAseq

Question

[sarahmanderni]

Hi,

I have scRNAseq data from mouse colon (benign tumors) and the mt percent for the cells per samlpe is as in the figure. Considering that the mt percent is relatively high in all samples, what would you conclude about the samples quality. Is it reasonable to remove for example cells with percent.mt > 20 and perform the routine analysis?

Answer 1

[jv]

High mito transcript "contamination", aka mitoDNA%, can be a marker of lysed, poor quality cells. There is a considerable range of mitoDNA% in your samples. It is acceptable to apply a static filter to remove cells with high contamination. What that threshold should be depends on a number of factors, but it has been proposed that 5% is an appropriate baseline recommendation for mouse cell, but there are a few things to consider such as:

• Doing your homework regarding biology of your samples. Some cell types may have greater numbers of mitochondria e.g., muscle, liver, Sertoli cells
  • Experimental treatments, disease conditions, etc. can affect mitochondria
  • Cellular stress, apoptosis increases mtDNA transcription
• Consider cell viability information of the samples
• Visualize the distribution of mitoDNA% across cells and samples
  • Preliminarily split cells into cell type groups and assess mitoDNA% for each group. Don’t immediately filter but instead tag/label cells with high mitoDNA%
  • Do additional exploratory visualization to assess trends related to mitoDNA% e.g., Are their specific cell types with higher contamination? Are high mitoDNA% broadly distributed across clusters? Other strong correlations?

The above points are from a presentation I put together regarding this very topic. Some of the references I used for my presentation are: Lukassen, Soeren, Elisabeth Bosch, Arif B. Ekici, and Andreas Winterpacht. 2018. “Single-Cell RNA Sequencing of Adult Mouse Testes.” Scientific Data 5 (1): 180192. https://doi.org/10.1038/sdata.2018.192

Ma, Anqi, Zuolang Zhu, Meiqin Ye, and Fei Wang. 2019. “EnsembleKQC: An Unsupervised Ensemble Learning Method for Quality Control of Single Cell RNA-Seq Sequencing Data.” In Intelligent Computing Theories and Application, edited by De-Shuang Huang, Kang-Hyun Jo, and Zhi-Kai Huang, 11644:493–504. Lecture Notes in Computer Science. Cham: Springer International Publishing. https://doi.org/10.1007/978-3-030-26969-2_47

Mercer, Tim R., Shane Neph, Marcel E. Dinger, Joanna Crawford, Martin A. Smith, Anne-Marie J. Shearwood, Eric Haugen, et al. 2011. “The Human Mitochondrial Transcriptome.” Cell 146 (4): 645–58. https://doi.org/10.1016/j.cell.2011.06.051

Osorio, Daniel, and James J Cai. 2021. “Systematic Determination of the Mitochondrial Proportion in Human and Mice Tissues for Single-Cell RNA-Sequencing Data Quality Control.” Edited by Anthony Mathelier. Bioinformatics 37 (7): 963–67. https://doi.org/10.1093/bioinformatics/btaa751

“Removal of Dead Cells from Single Cell Suspensions Improves Performance for 10x Genomics Single Cell Applications.” 2017. CG000130 Rev A Technical Note. https://cdn.10xgenomics.com/image/upload/v1660261286/support-documents/CG000130_10x_Technical_Note_DeadCell_Removal_RevA.pdf

[jv]

and cancer cells often show elevated mitoDNA%

Answer 2

[jared.andrews07]

This is normal. There will always be some degree of dead/dying cells, empty droplets with ambient RNA, etc. You look to have plenty of cells with a more appropriate mt% in each sample, which is good. 20% is a pretty high threshold for typical 10X scRNAseq. Some cell types tend to have higher mt% than others, so it's difficult to recommend an appropriate blanket threshold.

Starting with 20% and doing additional filtering from there is a good place to start. Most people use something in the 5-10% range as a threshold, but again, the distribution varies across cell types.