Assembles longest possible TCR/Ig receptor contig sequences.
This step may be used in the following cases:
- for non-enriched RNA-Seq (shotgun) data in order to reconstruct the longest possible VDJ contigs
- for single cell fragmented libraries (like 10x Genomics) to assemble full V-D-J consensus sequences in each cell
- for poor quality targeted amplicon data with no UMI barcodes and with a high rates of sequencing and PCR errors in order to build full-length (or as long as possible) V-D-J contigs using consensus algorithms
In the first two cases the data is fragmented, reads do not have a fixed position on the reference, so it's not possible to specify a fixed
assemblingFeature. MiXCR uses alignment-guided assembly algorithms to build the full-length consensus V-D-J sequence.
For single cell data MiXCR utilizes all alignments with the corresponding cell barcodes, even if they do not cover CDR3 region:
In the last case of poor quality targeted libraries, though the exact assembling feature may be known in advance, high mutation rate may lead to the situation when there aren't any true (error-free) full-length V-D-J sequences in the data. Error correction algorithms used in clonotype assembly will not have any "anchor" sequence to determine the original clone. In this case we might benefit from assembling clonotypes by CDR3 region at first and then use a consensus algorithm to determine true clones and distinguish errors from hypermutations.
The algorithm iterates through alignments that were used to build a particular clonotype, and aggregates information per each position in V-D-J reference. It takes into account cell barcodes (if present) to assembly full-length contigs inside cells. It also takes care of hypermutations and splits clonotypes into hypermutated variants (see
Note that since
assembleContigs uses original alignments, it takes
.clna (clones & alignments) file as input, produced by
assemble command with option
Command line options
mixcr assembleContigs [--ignore-tags] [--assemble-contigs-by <gene_features>] [-O <key=value>]... [--report <path>] [--json-report <path>] [--threads <n>] [--force-overwrite] [--no-warnings] [--verbose] [--help] clones.clna clones.clns
The command returns a highly-compressed, memory- and CPU-efficient binary
.clns file that holds exhaustive information on consensus clonotype contigs. Clonotype table may be further extracted in a human-readable form using
exportClonesPretty or in a tabular form using
exportClones. It is also possible to impute uncovered V-D-J contig parts from germline (marking such nucleotides lowercase). Additionally, MiXCR produces a comprehensive report.
Basic command line options are:
- Path to input clna file
- Path where to write assembled clones.
- Ignore tags (UMIs, cell-barcodes). Default value determined by the preset.
- Selects the region of interest for the action. Clones will be separated if inconsistent nucleotides will be detected in the region, assembling procedure will be limited to the region, and only clonotypes that fully cover the region will be outputted, others will be filtered out.
- Overrides for the assembler parameters.
-r, --report <path>
- Report file (human readable version, see
-j / --json-reportfor machine readable report).
-j, --json-report <path>
- JSON formatted report file.
-t, --threads <n>
- Processing threads
- Force overwrite of output file(s).
- Suppress all warning messages.
- Verbose messages.
- Show this help message and exit.
Here is the typical workflow for full receptor assembly of e.g. mouse B-cells.
Align raw sequences using kAligner2 specifically designed to work with highly mutated data:
> mixcr align \ --species mmu \ -p kAligner2 \ --report report.txt \ input_R1.fq \ input_R2.fq \ alignments.vdjca
Assemble default CDR3 clonotypes (note:
--write-alignments is required for further contig assembly):
> mixcr assemble \ -OassemblingFeatures=CDR3 \ --write-alignments \ --report report.txt \ alignments.vdjca \ clones.clna
Assembly full BCR receptor consensus sequences and clonotype into hypermutated variants:
> mixcr assembleContigs \ --report report.txt \ -OsubCloningRegion=VDJRegion \ clones.clna \ contigs.clns
Export full BCR receptors sequences and impute uncovered regions from germline (marked lowercase):
mixcr exportClones \ --chains IG \ --preset fullImputed \ contigs.clns \ contigs.tsv
Full sequence assembler parameters
Full sequence assembler parameters that may be tuned:
- Region where clonotype variants are allowed (typically that are hypermutated variants);
nullstands for no variants allowed (one consensus per one clone)
- Minimal contiguous sequence length
- Assemble only parts of sequences covered by alignments
- Minimal quality fraction (variant may be marked significant if variantQuality > totalSumQuality * branchingMinimalQualityShare
- Minimal variant quality threshold (variant may be marked significant if variantQuality > branchingMinimalSumQuality
- Variant quality that guaranties that variant will be marked significant (even if other criteria are not satisfied)
- Positions having quality share less then this value, will not be represented in the output
- Positions having sum quality less then this value, will not be represented in the output
- Maximal number of not aligned nucleotides at the edge of sequence so that sequence is still considered aligned “to the end”
- Number of nucleotides at the edges of alignments (with almost fully aligned seq2) that are “not trusted”
- Minimal fraction of non-edge points in variant that must be reached to consider the variant significant
- Gene feature limiting the set of positions where sufficient number of different nucleotides may split input into several clonotypes. If position is not covered by the region, and significant disagreement between nucleotides is observed, algorithm will produce "N" letter in the corresponding contig position to indicate the ambiguity. Null - means no subcloning region, and guarantees one to one input to output clonotype correspondence. Default - null
- Limits the region of the sequence to assemble during the procedure, no nucleotides will be assembled outside it. Null will result in assembly of the longest possible contig sequence. Default - null
- Used only if
assemblingRegionsis not null. Sets filtering criteria to apply before outputting the resulting clonotypes.
NoFiltering- don't filter output clonotypes.
OnlyFullyAssembled- only clonotypes completely covering
assemblingRegionswill be retained.
OnlyFullyDefined- only clonotypes completely covering
assemblingRegionsand having no "N" letters will be retained. Default -
- Discard ambiguous calls (no use of 'N's).