search

broadinstitute / CellBender

CellBender is a software package for eliminating technical artifacts from high-throughput single-cell RNA sequencing (scRNA-seq) data.
https://cellbender.rtfd.io
BSD 3-Clause "New" or "Revised" License
271 stars 50 forks source link

Problems interpreting results #338

Open MarP2606 opened 4 months ago

MarP2606 commented 4 months ago

Hello, thank you for the tool! I am a first time user and I am struggling a bit to interpret the results plots and to change the settings adequatly for the next run.

We performed songle-nuclei RNA sequencing from tissue. Therefore we expect high ambient RNA levels. The cellranger pre-processing recovered about 9.800 cells which fits with our expectations of targeting 10.000 cells. cellranger_report

I ran cellbender with the default settings. I think the learning curve looks ok from what I've gathered from other issues. learning_curve

I don't know what to expect for the top genes removed, but could it be that the fractions removed seem a bit low? genes_removed

Further, cellbender identified 5 times more cells than expected. cell_probability

So now I am a bit unsure which settings are the appropriate ones to change. 1) Should I change the FPR? Would 0.05 be the next try? 2) Should I change total-droplets-included (higher?) and expected-cells (10.000)? I also saw that changing the parameter low-count-threshold was suggested, which would be probably 1.000 in my case. What of those two options is better?

Thank you for your help!