Hi,
I'm, analyzing a single-nuclei dataset with not-so good quality, and I am trying to understand what went wrong.
These are cancer samples. Biopsies of metastases found in the liver of 3 patients. Pre-/post-treatment.
I ran velocyto to get the % intronic reads / total reads on the raw barcode matrix.
I'm assuming that droplets containing mostly ambient RNA coming from the cytoplasm of broken cells should have low %intronic. However, my "empty" droplets range 30%~60% intronic reads.
Which values of %intronic should I expect in cancer cells and in surrounding "healthy" liver tissue from cytoplasmic or intronic origin? I assume that, ambient RNA coming from the (unwashed) broken cytoplasm of otherwise "successfully captured" nuclei should lie at 0% intronic reads, no? What would the % approximately look like if it came from "whole broken cells" (cytoplasm + nucleus) ?
It looks to me like the ambient washout protocols were successful, and what I'm seeing on the empty droplets is... mostly ambient RNA coming from broken nuclei? Maybe because of harsh FACS-sorting?
Any ideas of what I'm seeing here?
1 answer
Rough numbers people usually report: 50-80% intronic for single-nuclei, 20-35% for whole-cell 10x. So whole cell isn't anywhere near zero either.
That's the bit I'd push back on in your reasoning - pure cytoplasmic mRNA doesn't give 0% intronic. Retained introns are common and get more so in tumours, plenty of reads land in an intron of one isoform that's exonic in another, and unannotated 3' UTR extensions overlap intronic space. A realistic floor for cytoplasm-only is more like 10-20%.
With that in mind, empty droplets at 30-60% does look like nuclear-derived ambient to me, so I think your read of it is right. The comparison that would actually settle it is empty-droplet %intronic against your cell-containing droplets: close together means the ambient is the same material as your nuclei and the wash did its job, empties clearly lower means you've got cytoplasm from broken whole cells in there.
Log in to answer this question.