WARNING: 10X Multiome (GEX + ARC) should not use bustools 10xv3 internal barcode whitelist

[davemcg]

tldr: If you want to quantify a 10x Multiome experiment (just the Gene / GEX part) you MUST provide the proper onlist to kb count.

This is not a bug per-se, but rather an oversight for an edge case. The 10X Multiome GEX+ATAC chemistry (at least for the GEX part) uses the same barcode setup (16 + 12) as 10xv3.

BUT

If you run the kb count workflow with -x 10xv3 you may (at least in my case) return far too few cells. I was very miffed until I realized that 10x uses a different barcode whitelist for the multiome compared to the 10xv3.

If you give kb count the 10x multiome barcode whitelist as -w gex_737K-arc-v1.txt you get the proper result.

Versions

I noticed this error on 0.28.2 and confirmed it on 0.29.1 (though I did use a 0.28.2 idx...which I don't think matters)

Retrieve data

# get a gene 10x multiome fastq pair
fasterq-dump SRR29226057; pigz -p 8 SRR29226057*
wget https://teichlab.github.io/scg_lib_structs/data/10X-Genomics/gex_737K-arc-v1.txt.gz
gunzip gex_737K-arc-v1.txt.gz

Run kb count the "default" way

This gives you about 679 cells after sc.pp.filter_cells(adata, min_genes = 300)

# not the exact command, but no one wants to see the full paths
 kb count  --workflow nac   --sum total  -g  t2g.txt -t 12  -x 10xv3  -i index.idx  -c1 t2c.cdna.txt -c2 t2c.unprocessed.txt                        -o SRR29226057_kb_provided_whitelist   --h5ad  SRR29226057_1.fastq.gz SRR29226057_2.fastq.gz

# now in python
import scanpy as sc
adata = sc.read_h5ad('SRR29226057_kb_provided_whitelist/counts_unfiltered/adata.h5ad')
sc.pp.calculate_qc_metrics(adata, percent_top=None, log1p=False, inplace=True)
sc.pp.filter_cells(adata, min_genes=300)
adata 

Give kb count the correct onlist

Same, but adding the -w flag -> now you get 12582 cells

kb count  -w gex_737K-arc-v1.txt --workflow nac   --sum total  -g  t2g.txt -t 12  -x 10xv3  -i index.idx  -c1 t2c.cdna.txt -c2 t2c.unprocessed.txt   -o SRR29226057_arc   --h5ad  SRR29226057_1.fastq.gz SRR29226057_2.fastq.gz 

# python
import scanpy as sc
adata = sc.read_h5ad('SRR29226057_arc/counts_unfiltered/adata.h5ad')
sc.pp.calculate_qc_metrics(adata, percent_top=None, log1p=False, inplace=True)
sc.pp.filter_cells(adata, min_genes=300)
adata

[davemcg]

Related - you may want to add "10xv4" chemistry as a "-x" option as the barcodes are different

[Yenaled]

Thanks! I think we need a new technology string for that.

As for 10xv4, I believe that is already added on the latest version of kb-python.

[davemcg]

I don't see 10xv4 on 0.29.1

I also don't see "10xv4" in this repo (though I don't know whether github only searches the main branch when you use the web gui).

Anyways, have fun adding a string that conveys "10x multiome v1 but only gex"

[Yenaled]

Ah you are correct; I put the 10xv4 into one of the dependencies but forgot to update the kb-python code.