mixcr with 10x Visium data

[lymphomatic]

I am trying to extract TCR and BCR clonotypes from 10x Visium data (standard library without VDJ enrichment), but I am getting no results using the standard 10x-sc-5gex preset (no alignment), even with addition of --tag-pattern "^(CELL:N{16})(UMI:N{12})\^(R2:*) correction.

mixcr analyze 10x-sc-5gex \
--species hsa \
--tag-pattern "^(CELL:N{16})(UMI:N{12})\^(R2:*)"  \
"RS23_10390_10B_S6_R1_001.fastq.gz"  "RS23_10390_10B_S6_R2_001.fastq.gz" \
result

I have tried to use the BAM file from the 10x SpaceRanger output, but I get consistently the error (I have tried with several different BAM files from different samples):

CommandLine$ExecutionException: Error while running command align htsjdk.samtools.SAMFormatException: SAM validation error: ERROR::INVALID_CIGAR:Record 155943416, Read name VH00708:358:22232VMNX:2:2502:32572:54393, CIGAR should have > zero elements for mapped read.

I have also tried using the simple rna-seq preset, and using the alternative --tag-pattern "^(UMI:N{28})\^(R2:*) but without difference.
Am I doing something patently wrong here? Or, alternatively, is it possible there might be a preset for (possibly pseudo-bulked) 10x Visium-specific preset?

Thanks for any insight!
Adam

[mizraelson]

Hi, sorry for the late reply. Generally speaking it is unlikely to recover decent immune repertoire from non targeted 10x data especially with 3' chemistry (default for 10x Visium).

Nevertheless, I would recommend to try the following:

mixcr analyze generic-ht-single-cell-fragmented-with-umi \
​     --rna \
     --species hsa \
     --tag-pattern "^(CELL:N{16})(UMI:N{12})\^(R2:*)" \
     -Massemble.consensusAssemblerParameters.assembler.maxConsensuses=10 \
     sample_R1.fastq.gz \
     sample_R2.fastq.gz \
     sample_result

Since theoretically there can be multiple cells marked with the same spatial barcode I adjust the maxConsensuses number to 10 per chain (TRA, TRB, IGH, IGL, IGK...) (default is 2).

[lymphomatic]

Thank you very much for the suggestion! Despite all attempts, every run ends up with the result:
Reads dropped due to the lack of a clone sequence, percent of total: 750 (0%)
I think I can assume that this is just the lack of adequate sequences in the data.
My only concern is that all alignments on review from the exportAlignmentsPretty module looks very similar:
` <FR4
_ V I L F F F F F F F F F F T T G W F K I F A E G T K
Quality 48888488488888888888888888888448884888848888888888888888888888888888888888888888
Target0 0 CCGTCATTCTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTACCACTGGTTGGTTCAAGATATTTGCTGAAGGGACTAAG 79 Score
TRGJP1*00 21 taccactggttggttcaagatatttgctgaagggactaag 60 250

          L  I  V _   

Quality 8888888888
Target0 80 CTCATAGTAA 89 Score
TRGJP1*00 61 ctcatagtaa 70 250

Read ids: 2520780`

• which makes me worried that perhaps I am missing some parameter that prevents proper assembly.

[mizraelson]

Can you share the alignment, refine and assemble txt report files?

[lymphomatic]

Of course, thank you for taking a look at them!
It seems that all reads are dropped "due to the lack of a clone sequence", so perhaps they are just too short.

res.align.report.txt
res.assemble.report.txt
res.refine.report.txt

[mizraelson]

Given that it's a 3' non-enriched library, we have relatively few reads to begin with:
Successfully aligned reads: 260,361 (0.14%)

89% of these reads cover the region outside the CDR3:
No CDR3 parts alignments, percent of successfully aligned: 231,687 (88.99%)

And 11% partially cover the CDR3 region:
Partially aligned reads, percent of successfully aligned: 28,674 (11.01%)

MiXCR has a partial assembly step that rescues some CDR3s if two pairs of reads combined cover the same CDR3 completely, but that doesn't seem to be the case here. Even after this step, none of the alignments fully cover any complete CDR3 region sequence:
Reads dropped due to the lack of a clone sequence, percent of total: 250,869 (100%)