Running MiXCR on Overlap Extension Data

[spracers]

I'm hoping to run MiXCR on data collected from an overlap extension RT-PCR library. In this library, the primers were designed to be overlapping such that the TCR alpha and TCR beta transcripts become physically linked together on the same DNA construct. These constructs were then sequenced on an Illumina MiSeq.

The GitHub repository (https://github.com/milaboratory/mixcr/releases/tag/v4.1.1) states that MiXCR has implemented initial support for overlap extension based chain pairing protocols, but I haven't been able to find any documentation or guidance on how to use MiXCR on sequencing data from these overlap extension libraries.

I'd really appreciate guidance on applying MiXCR to this type of data!

[mizraelson]

Hi, can you describe your library structure in more detail? Is it always one chain (e.g. alpha) in R1 and the other (e.g. beta) is in R2? Do you have any UMIs?

[spracers]

Yes, for each sequencing construct, the TCR alpha chain is covered in read1 and the TCR beta chain is covered in read2.

Each sequencing construct (from 5' to 3') consist of the read 1 sequencing adapter, the V alpha sequence, the J alpha sequence, the C alpha sequence, the linker region (generated by the overlap in the primers), the V beta sequence, the D beta sequence, the J beta sequence, and C beta sequence, and the read 2 sequencing adapter.

The library does not have UMIs, but it does have Illumina barcodes at the 3' end of the construct.

[mizraelson]

Is the data already demultiplxed (by sample) or do you want to use the Illumina barcodes for demultiplexing?

[mizraelson]

Please put the attached YAML file either to ~/.mixcr/presets/ or to the directory from where you run mixcr. Then run:

mixcr analyze local:chain-pair-overlap \
--species mmu \
input_R{{CELLSPLIT:n}}.fastq.gz \
output

Put the {{CELLSPLIT:n}} pattern in place of a read number in your input files names.
Also, make sure to select the proper --species.

Let me know how it works.

Sincerely,
Mark
chain-pair-overlap.yaml.zip

[spracers]

I'm running MiXCR version 4.1.1, and I tried to run the above command (with the .yaml file provided), but I first got the error "No preset with name "generic-amplicon"", so I changed the present name to "generic-tcr-amplicon" in the .yaml file.

However, now I'm getting the error "No preset with name "exportClones-single-cell-no-umi-base"" . Is there another preset name I should be using, or do I need to be running a different version of MiXCR?

Also, with regards to the file names, I have two files (sample1_R1.fastq and sample1_R2.fastq). Should I be renaming the files to sample1_R1{{CELLSPLIT:n}}.fastq and sample1_R2{{CELLSPLIT:n}}.fastq?

The data is already demultiplexed by sample, so MiXCR wouldn't have to deal with the demultiplexing.

Thanks so much again for your help!

[spracers]

I installed MiXCR version 4.5.0, and the errors with the presets went away. I can now successfully run the command below:

mixcr analyze local:chain-pair-overlap \
--species hsa \
sample1_R{{CELLSPLIT:n}}.fastq \
output

using the input files sample1_R1.fastq and sample1_R2.fastq.

I'm using sequencing data from this paper that is deposited on the sequencing read archive (SRA). The paper reports that they resolve ~500k paired TCRalpha-TCRbeta clonotypes per sample. However, when I ran MiXCR on the sequencing data, it only resolved ~15k TCRalpha chains and ~28k TCRbeta chains. Additionally, the output clones.tsv file did not link the alpha and beta chains, and instead gave separate rows for the individual chains.

I'd really appreciate your advice with troubleshooting this, as I've tried different settings but haven't been able to resolve this issue. Also, how do I tell MiXCR assemble that I want it to output clones that have paired TCRalpha-TCRbeta chains?

[mizraelson]

Yes, the preset is only compatible with the latest version of MiXCR. You can pair chains based on the value in the tagValueREAD_IDX column. Identical numbers indicate that the clones originate from the same read. If you could share the .txt report files generated by MiXCR, I could examine them for any issues that might be causing the loss of the expected number of clones. Please note that this feature is currently under development, and as of now, we cannot automatically filter out unpaired chains.

[spracers]

Thanks for your help! Attached are the .txt files generated by MiXCR.
assemble.report.txt
refine.report.txt
align.report.txt
qc.txt

Here are the input .fastq files in case it be helpful for you to have them.

The paper reports ~339,000 paired TCRab clonotypes from this data, but when I ran MiXCR with the options provided above, it only returned ~154,000 paired TCRab clonotypes. I'd really appreciate your help with troubleshooting this - are there particular options with regard to the read alignment that might have to be changed?

Also when reviewing the results, I noticed that all of the chains have a readCount equal to 1 or 2, and I want to make sure I'm interpreting this correctly. Does this mean that there is only one or two sequencing reads supporting each TCR chain?