Confusion about "Top Features table" in cellranger web summary output for CITE-Seq libraries

Question

[Chris Dean]

I am reviewing the Web Summary (count) documentation on the 10X genomics website and became confused as to what is being displayed in the Top Features by Cluster table under the Antibody tab (see screenshot below). Why are Ensembl gene IDs being displayed instead of Antibody IDs for a CITE-Seq library? It looks like cellranger performs the clustering by the antibody counts, but uses the gene expression data to perform differential expression between clusters?

[Trivas]

I'd assume that CellRanger already converted from antibody barcode space to human readable ensembl IDs/gene symbols.

[ATpoint]

I'd assume (as you suggest) it clusters in surface feature space and then identifies genes that mark the cluster. Honestly, I personally always turn all of this postprocessing off and do my own analysis. CellRanger is a blackbox, you essentially have no clue what exactly it does, so I don't trust any of it beyond the count matrix. What @Trivas said is unlikely simply because shown genes are not surface features and it appears utterly unlikely that CellRanger makes an unsolicitedattempt to convert the names in the surface feature CSV file to any genes, since the csv file can be named at will, you could even call the HTOs foo1, foo2, foo3 and so on.