Reads not aligning with 10x-ont-visium-hd-3gex preset

[abspangler13]

Hello,

I'm running mixcr analyze 10x-ont-visium-hd-3gex on some Visium HD 3' ont TCR/BCR libraries.

mixcr analyze 10x-ont-visium-hd-3gex \
     --species hsa \
     "${SAMPLE}" \
     "MiXCR_ouput_${SLURM_ARRAY_TASK_ID}"

Looking at the align.report.txt output I can see that the vast majority of the reads are not aligned either because a barcode wasn't found or because the barcode didn't match the whitelist. Here's a snippet of the report file.

Total sequencing reads: 157503935
Successfully aligned reads: 9133376 (5.8%)
Coverage (percent of successfully aligned):
  CDR3: 8964887 (98.16%)
  FR3_TO_FR4: 9007318 (98.62%)
  CDR2_TO_FR4: 8956053 (98.06%)
  FR2_TO_FR4: 8492877 (92.99%)
  CDR1_TO_FR4: 8137348 (89.09%)
  VDJRegion: 7331898 (80.28%)
Alignment failed: no hits (not TCR/IG?): 31086485 (19.74%)
Alignment failed: absent barcode: 38966180 (24.74%)
Alignment failed: sample not matched: 78317894 (49.72%)
Overlapped: 0 (0%)
Overlapped and aligned: 0 (0%)
Overlapped and not aligned: 0 (0%)

Does anyone have any idea what's going on with the barcode alignment process that would cause this result?

Thanks,

Abby

[mizraelson]

Hi, is it a publicly available dataset?

[abspangler13]

No, it's patient data and it can't be shared unfortunately. What can I look for to try to troubleshoot this problem? How is your barcode demultiplexing different from other methods such as Flexiplex?

[mizraelson]

The way MiXCR handles it:

1. The following pattern is applied to extract the barcodes ^*ctcttccgatctN{0:5}(UMI:N{9})N{0:3}(CELL1:gN{13:14}g)(CELL2:N{13:14}g)N{0:5}T{7:}(R1:*)
Alignment failed: absent barcode: 38966180 (24.74%) in your data basically means that 25% of the reads do not carry that pattern.
2. CELL1 and CELL2 are checked against the list to match with the coordinates. You can find the sequences here
   This step is where you lose Alignment failed: sample not matched: 78317894 (49.72%)

One of the potential reasons is the sequencing quality. An indel withing the pattern can make it impossible to identify the barcodes correctly.

What I would do:

1. Add --output-not-used-reads to the command to export not aligned reads in FASTQ files
2. Manually inspect a few randomly picked reads. Can you identify the barcodes there? If yes, then you can maybe change the patter to include it.