What kind of statistical test does Cellranger count use?

Question

[hannah]

Hi all,

I'm trying to understand how 10X cellranger count/Loupe does clustering statistically. Is this a series of t-tests done one after another? Or by ANOVA?

[ATpoint]

Clustering is not a statistical test with p-values. It is unclear what you mean. Actually, I would just ignore everything that the blackboxish CellRanger does, and just use the count table and analyse yourself. What it does is anyway not reliable as it does not do proper per-cell QC etc as should be done, see for example guided tutorials at OSCA/Bioconductor, Seurat, Scanpy and others.

Answer 1

[fracarb8]

You should read the cellranger documentation to understand how the analysis are performed. For example, it uses graphclust for clustering.

As ATpoint suggested, I would use the analysis from cellranger as a mean to have a general understand of how your experiment perfermed, and use additional tools/protocols for the actual analysis.