Missing gem_classification.csv in Cell Ranger 9.0.0 Output Using cellranger multi with Combined Human-Mouse Reference

Question

[tlallihuacan]

Hi all,

I’m using Cell Ranger 9.0.0 to process some 5′ single-cell RNA-seq samples from NGS mice engrafted with human tumoral cells. The samples include gene expression, VDJ, and OligoHashtag libraries. I’m running cellranger multi with a combined reference (GRCh38 + GRCh38_two_transgenes + GRCm39) to try and detect the human transgenes in the cells.

In previous Cell Ranger versions (e.g., 7.x), I used to get a gem_classification.csv file under the analysis folder, which indicated whether a cell was of mouse or human origin. In Cell Ranger 9, this file no longer appears, and I can't find any clear equivalent. Here is a simplified version of my CSV config:

[gene-expression]
reference,/path/to/GRCh38_and_GRCm39_combined_reference
create-bam,false
force-cells,10000
tenx-cloud-token-path,/path/to/credentials
cell-annotation-model,auto
[feature]
reference,/path/to/barcode_sample_assignment.csv
[vdj]
reference,/path/to/vdj_reference
[libraries]
fastq_id,fastqs,lanes,feature_types
sample1,/path/to/GEX/fastqs,1|2,gene expression
sample1,/path/to/VDJ/fastqs,1|2,vdj
sample1,/path/to/Hashtags/fastqs,1|2,antibody capture

And and example of the barcode_sample_assignment.csv:

id,name,read,pattern,sequence,feature_type
Hash_1,Hash_2,R2,5PNNNNNNNNNN(BC),GTCAAAACCCCAGCG,Antibody Capture
...

Important: I’m not demultiplexing the hashtags at this stage — that step comes later. Question: Does Cell Ranger 9 no longer generate gem_classification.csv or any equivalent file indicating host species classification in multi-species references? Is there a new way to access this information, or am I missing a config option? Any help will be very appreciated!

[GenoMax]

Reach out to 10x tech support since you have not received a response yet. They should provide an official answer, in case this output file has been removed from v.9.x. Sometimes changes are made (this is a major version release) that are not reflected in the documentation right away.

Please post their response here, when you get one.

[tlallihuacan]

Thanks GenoMax I did that the next day I saw I didn't get the answer here, I am coming back now.

This is what 10x answered:

"Thank you for providing the information. We have not tested and do not support analysis of human + mouse samples with data modality besides GEX. It's likely that the gem_classification.csv file is not generated because there are data types in addition to GEX in the particular analysis."

[GenoMax]

You will need to drop the other assays besides GEX to get the that file then or find an alternate solution.