arda — Antigen Receptor Domain Annotation

Fast FR/CDR region annotation for TCR and BCR nucleotide and amino-acid sequences. arda builds a pre-aligned IgBLAST reference database once, then maps queries with MMseqs2 and transfers the region markup through the alignment in a small C++ hot path — producing AIRR-formatted output that matches IgBLAST.

Contents

• Installation
  • Development bootstrap
  • What it does, and what it verifies
  • MMseqs2 without conda
  • The segment reference is generated, not shipped
  • IgBLAST without a checkout
• Usage
  • Command line
  • Python library
  • Bulk RNA-seq mode
  • Choosing a mode
  • Junction markup and repair
  • Scaling
  • Run reports
  • Supported organisms
• D segments and tandem D-D
  • How the call is made
  • d_support ranks correctness: the E-value bands
  • Germline geometry: two constraints, applied before the statistics
  • A tandem D-D must run in genomic order
  • Using the output: D-D markup
  • See also
• Somatic hypermutation
  • What arda records
  • Why not just diff the two alignment strings
  • Recounting a table written before the fix
  • Recomputing what you actually want
  • Accuracy against IgBLAST
  • Building a lineage tree
  • See also
• Error correction
  • The abundance test
  • Quality: the evidence abundance does not have
  • Where the abundance model stops working: the ladder and the cliff
  • The quality-directed rescue
  • Modes: --ec-mode
  • The clonotype key: --clonotype-key
  • What the options actually do
  • Calibration: the MIGEC spike-ins
  • What a junction disagreement means
• Run QC, verbosity and logging
  • The table
  • Gene coverage
  • Quality columns
  • Alleles versus SHM
  • Verbosity and logging
  • Streams
  • Peak RSS
• Pipeline integration
  • One-shot commands
  • Nextflow module
• Running on a cluster (SLURM)
  • One command for the whole chain
  • Splitting by hand
  • Very large inputs: shard Stage 1, run Stage 2 once
  • Choosing the regime
  • Resource sizing
• Use cases and analysis guidelines
  • Bulk RNA-seq: extract a repertoire from a transcriptome
  • Targeted amplicon / RepSeq
  • Monoclonal QC: is my cell line what it says it is?
  • Negative controls
  • Low-frequency variants: spike-ins, MRD, minor clones
  • Somatic hypermutation
  • Comparing arda against another tool
  • The one test that needs no external truth
• Reference database build
  • Pipeline
  • Outputs (per organism)
  • Reading frames
• Exporting the reference
  • The three kinds
  • The four formats
  • Worked examples
  • Reproducibility
• API reference
  • Library entry point
  • Runtime annotation
  • Junction markup and repair
  • Bulk RNA-seq
  • Run QC and logging
  • Cluster sharding
  • Reference build
  • External tool wrappers
  • Cache layout and reference fetch

Indices

• Index

• Module Index