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
299 stars 54 forks source link

raw h5ad OR filtered h5ad #378

Open elegantmedstu opened 3 months ago

elegantmedstu commented 3 months ago

Hi, I would like to ask, for the h5ad file input by cellbender, should I use the raw data or the filtered data? Thank you for your time, thank you!

maltekuehl commented 3 months ago

Hi, not related to Cellbender, but came across this question and will try to answer. Assuming you are talking about Cellranger (or similar) output, you would want to use the unfiltered data since Cellbender needs surely empty droplets to learn the ambient RNA profile and correct for it. Together with the ambient RNA corrected counts, Cellbender will also output a cell probability that you can use to call cells instead of the EmptyDroplets algorithm implemented in other software such as Cellranger, see the Cellbender documentation for reference. While Cellbender should usually perform well for this task, I however usually also add some further selection steps afterwards such as requiring a minimum count, minimum number of UMIs, maximum number of UMIs, etc. What values you should use for this and whether that is necessary depends on the source of your data (e.g. single nuc) and you are best advised to review the count distribution. For more details, you might want to checkout Single Cell Best Practices.