decreased number of clones in mixcr-4.4.0+

[github-wow]

Hi,

When I run one-line analyze solution with a pair of my files I've noticed that
recent mixcr versions (mixcr-4.3.2, -4.2.0, -4.4.1, -4.4.0) produce .vdjca files with similar sizes:

.vdjca        :   1.5 Gb
.refined.vdjca: 560   Mb

but the clones-related files are significantly decreased in size (and # of clones)
in versions with new preset definitions introduced in v.4.4.0
like "milab-human-rna-tcr-umi-race" instead of deprecated "milab-human-tcr-rna-race-cdr3":

mixcr-4.4.1, -4.4.0 (with new preset definitions):

.clns           :  48 **Kb**
.clones_TRAD.tsv:  11 **Kb**

mixcr-4.3.2, -4.2.0 (with previous preset definitions):

.clns           :   6   **Mb**
.clones_TRAD.tsv:  27   **Mb**

Any help is appreciated.

[mizraelson]

Hi,
Can you please share exact commands you used with both version and the length of you reads (R1 and R2)?

[github-wow]

Thank you for your reply. The exact lines when done with slurm:

mixcr_version="mixcr-4.4.0"

$SCRATCH/mixcr/mixcr -Xmx250g analyze -t 32 milab-human-rna-tcr-umi-race \
      $SCRATCH/input/1.fq.gz                                      \
      $SCRATCH/input/2.fq.gz                                      \
      $SCRATCH/output/${mixcr_version}_test_run.result

or

mixcr_version="mixcr-4.3.2"

$SCRATCH/mixcr/mixcr -Xmx250g analyze -t 32 milab-human-tcr-rna-race-cdr3 \
      $SCRATCH/input/1.fq.gz                                      \
      $SCRATCH/input/2.fq.gz                                      \
      $SCRATCH/output/${mixcr_version}_test_run.result

[mizraelson]

and whats the length of R1 and R2?

[github-wow]

150+150

[mizraelson]

By default, milab-human-rna-tcr-umi-race constructs clones based on the VDJRegion. This differs from milab-human-tcr-rna-race-cdr3, which builds clones utilizing the CDR3 region. If you want to replicate the previous behavior entirely, you can either continue using the old preset name—which is deprecated but functional—or use the new preset name and add the --assemble-clonotypes-by CDR3 parameter.
In your case the 150+150 sequencing coverage is insufficient for spanning the entire VDJRegion. Therefore, milab-human-rna-tcr-umi-race doesn't yield significant results by default.

[github-wow]

thank you for the clarification. but despite short coverage it is still sufficient to understand which VDJC genes are used, correct?

[PoslavskySV]

Sure, there is still enough information to identify V/D/J/C gene assignment (and you have corresponding columns in the output)

[github-wow]

thank you for the clarification