High number of unassigned alignments and exported nonproductive clones for scBCRseq

[sckers]

I am trying to analyze single-cell Ig-enriched sequencing data for comparison against our current pipeline, but have run into a couple of concerns.

1. It seems we are getting a high number of unassigned alignments that causes the final read count for the clonotypes to be lower than expected. I will share the report files, but in summary, 60k reads successfully aligned, but only 36k reads are used in the clonotypes. 42k reads are dropped due to lack of a clone sequence, and 24k alignments are unassigned. The data is fragmented, so assembleContigs does run, but this doesn't seem to maintain any additional reads because the final total read count when the clones are exported is 36k. In comparison, when we run this sample using BALDR (our current pipeline), the total read count 93k for the same clones. I know the assembly methods are different, but I would still expect everything that successfully aligns to be used in the assembly, based on my understanding of assembleContigs.

2. The particular sample that I'm working with exports with 1 heavy chain and 2 light chain sequences. One of these light chain sequences is nonproductive, and I can't figure out why mixcr isn't recognizing it as nonproductive. There is a stop codon in the CDR3 that mixcr is showing in the final clone report. Running with BALDR removes this sequence due to being nonproductive.

Protocol

We are using SmartSeq to synthesize and amplify cDNA from single-sorted-cells that have been sorted into plates, enrich with Ig constant region primers, and prepare libraries using the Illumina Nextera XT kit. Libraries are sequenced on the NextSeq 2000 using PE 2x151, IR 10x2. More details can be found in the original paper for the method, if needed.

mixcr command

The original data is run with Illumina's bcl2fastq software to demultiplex such that there is one set of read files per well. Then I use trimmomatic to trim the adapters and prepare the fastq files for mixcr. I am using the generic-lt-single-cell-fragmented preset with the development version from this thread. At the moment, I am only running one well as a test.

mixcr analyze generic-lt-single-cell-fragmented \
     --species hsa \
     --assemble-contigs-by VDJRegion \
     ../200-655-113_{{CELL:a}}_S1_L001_R1_001_nexteratrim_{{R}}P.fastq.gz \
     generic-lt-frag

Thank you in advance for your help, and let me know if I've missed any critical details!

generic-lt-frag.align.report.txt
generic-lt-frag.assemble.report.txt
generic-lt-frag.assembleCells.report.txt
generic-lt-frag.assembleContigs.report.txt
generic-lt-frag.assemblePartial.report.txt
generic-lt-frag.extend.report.txt
generic-lt-frag.qc.txt
generic-lt-frag.refine.report.txt