MiXCR aligns immune-repertoire reads to V/D/J/C germline and assembles them into clonotypes (CDR3 + V + J). In MiXCR 4.x correctness is dominated by one decision: the preset, which encodes the library chemistry (RNA vs gDNA, 5'RACE vs multiplex primers, bulk vs 10x single-cell, UMI vs no-UMI). The wrong preset does not error -- it silently produces plausible-but-wrong clonotypes (truncated CDR3, mis-called V, inflated diversity from uncollapsed UMIs), so the analytical work is choosing and auditing the preset, then reporting the correct quantitation unit (reads for non-UMI bulk, molecules for UMI, cells for single-cell). MiXCR 4.x also requires an activated license (free for academics) or it refuses to run, and the D-segment call in TRB/IGH is near-unassignable and must never key a clonotype.
conda install -c bioconda mixcr
# or download the release jar: https://github.com/milaboratory/mixcr/releases
# MiXCR 4.x needs Java 17 (mixcr --version prints the JVM)
# License is mandatory for 4.x (academic use is free):
# get a key at https://platforma.bio/getlicense
mixcr activate-license # paste the key, or:
export MI_LICENSE_FILE=/path/to/mi.license
Tell your AI agent what you want to do: - "Pick the right MiXCR preset for my Takara SMARTer TCR kit and run it" - "Process my 10x VDJ FASTQ with MiXCR and pair the chains" - "Assemble clonotypes from bulk TCR-seq and export CDR3 with V/J usage" - "Mine TCR clonotypes from my bulk RNA-seq BAM/FASTQ" - "Export my clonotypes as AIRR TSV for Immcantation" - "Check the alignment rate and chain composition QC on my run"
"This is a 5'RACE template-switch TCR library with a 12nt UMI -- which MiXCR preset, and run it on human data."
"Run MiXCR on my multiplex-primer BCR amplicon and set the floating boundary on the primer side."
"Process my 10x Genomics single-cell VDJ FASTQ and assemble paired-chain cells."
"Export TRB clonotypes with CDR3 nucleotide and amino-acid sequences, V and J genes, and the UMI count."
"Give me AIRR rearrangement TSV from these clones so I can run Immcantation lineage analysis."
"Restrict the export to productive clonotypes only."
"Show the MiXCR alignment-rate QC and tell me if the preset matches the chemistry."
"Check the chain-usage QC for cross-contamination between my TRB samples."
"Why did my RNA-seq run produce so few clonotypes?"
mixcr exportPreset.mixcr analyze <preset> (or the hand-run align -> refineTagsAndSort -> assemblePartial/extend -> assemble -> assembleCells chain), adding --assemble-clonotypes-by when the preset lacks an intrinsic assembling feature.exportClones) or AIRR (exportAirr), selecting the correct quantitation field (reads vs UMIs vs cells).mixcr qc and exportQc align/chainUsage and interpret alignment rate and chain composition against the chemistry.mixcr exportPreset --preset-name <name>.uniqueMoleculeCount for UMI libraries, cells for single-cell. Reporting reads on a UMI library re-adds PCR bias.refineTagsAndSort -- uncollapsed barcode errors become fake clonotypes and inflate diversity.--keep-non-CDR3-alignments, run assemblePartial twice, then extend; judge success by absolute clonotype yield, not percent aligned.--assemble-clonotypes-by CDR3 (or VDJRegion) if a preset has no intrinsic assembling feature.extend on BCR can fabricate germline over hypermutated ends; use cautiously and prefer longer reads for SHM work.