For "cellranger count," what is the underlying feature counting algorithm?

[olgabot]

Hello, I am making mini-bams per cell, filtering them for certain reads, and then running featureCounts on them. But I can't tell if this is truly comparable with the output of cellranger count, especially for the Unassigned category of reads. How does cellranger count deal with multimapping, and ambiguous/overlapping features?

Here is an example summary from featureCounts:

Assigned        5561
Unassigned_Unmapped     0
Unassigned_MappingQuality       0
Unassigned_Chimera      0
Unassigned_FragmentLength       0
Unassigned_Duplicate    0
Unassigned_MultiMapping 5806
Unassigned_Secondary    0
Unassigned_NonSplit     0
Unassigned_NoFeatures   23
Unassigned_Overlapping_Length   0
Unassigned_Ambiguity    101

Thank you! Warmest, Olga

EDIT: Misspelled my name ...

[evolvedmicrobe]

Hi Olga,

Cell Ranger does a fair bit more than simple counting, you can read about our algorithms here:

https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/algorithms/overview

We account for UMIs, UMI correction, chimera filtering, etc. So while the results are likely qualitatively similar, I would not expect an exact match.

Cheers, Nigel