Entering edit mode
5.0 years ago
bioinformatics2020
▴
830
In our analysis, we are dealing with a suspicious cluster that is 100% representative of one of our samples (we have four total), and also contains an extremely low amount of percent MT genes. Our current MT low cutoff is a quantile of .025 for the percent of MT in each sample (which we will playing around with and seeing how it changes.) Are there any other recommendations or suggestions on how to go about situations like this?
It depends on how your samples are prepared. Usually, sNucSeq samples has much lower MT% than scRNASeq samples.
Gotcha. Do you have an perspective on whether we should use a hard percentage, rather than filtering based upon the quantiles of each sample?
I always use a fixed percentage (such as 10%), since high MT% may indicate dying cells. See: https://kb.10xgenomics.com/hc/en-us/articles/360001086611-Why-do-I-see-a-high-level-of-mitochondrial-gene-expression-