A large number of cells are unassigned & very few n_vars

[nansne]

Hi,
I ran vireo model 2 with my 10X multiome data (joint ATAC+RNA, a pool of 23 samples). But got a large number of unassigned cells, and very few n_vars per cells. I'm confused which step was wrong. Thank you!

[huangyh09]

hi, it says only 1102 out of 149K variants were matched to the donor VCF file and can be used for demultiplexing. You may try the following things:

1. check if your donor VCF has been imputed - you can get more SNP by imputation, e.g., via Sanger Imputation Server. Then, it will give you more SNPs to use.
2. If 1) doesn't improve much, you may consider running cellsnp-lite with the donor VCF you have in 1) again, while you may want only to keep SNPs that donors have different genotypes.

Yuanhua

[nansne]

Thank you, Yuanhua. The donor VCF had been imputed before.
So I followed your second advice, ran cellsnp-lite with the donor VCF, and the code as:
cellsnp-lite
-s ./gex_possorted_bam.bam
-b ./barcodes.tsv.gz
-O ./testvkha_poolvcf_ref
-R ./data_pool1.vcf.gz
-p 16 --minMAF 0.1 --minCOUNT 20 --gzip --UMItag None
it took a long time ([I::main] time spent: 27613 seconds.)
Next, I ran vireo with the code:
vireo
-c ./testvkha_poolvcf_ref
-d ./data_pool1.vcf.gz
-o ./testvkha_poolvcf_ref
-p 16 --randSeed 2 --genoTag GT
however, there were many unassigned cells too

Looking forward to hearing from you!
Best wishes,
Nan

[huangyh09]

Hi Nan, good to hear it's been imputed already. However, the results are still quite concerning, particularly the estimated allelic rate mean, it should be near 0, 0.5, and 1.0, while yours are 0.133, 0.846, and 0.156.

You may double-check that your scRNA BAM and your donor VCF are in the same genome build, e.g., both hg38 or hg37.

Yuanhua