Immunarch will significantly evolve, but it will break things – and we need your help

Software
2

:dna: immunarch 0.10.3 — Release Notes

Highlights

  • Single-cell support. Load single-cell AIRR data, pair chains, and link scRNA-seq or spatial metadata.
  • Repertoire statistics. Key functions for gene usage, diversity, clonality, and publicness are implemented. More will come.
  • Out-of-memory + rich annotation. Work with data bigger than RAM. Annotate receptors with any info (e.g., immunogenicity, gene expression, cluster). Powered by ImmunData from immundata package.
  • Faster at scale. Major speed-ups for large cohorts and single-cell libraries thanks to DuckDB backend.
  • More reproducible. Immutable data objects reduce side effects and make analyses repeatable.
  • Easier to install. Multiple packages have been moved to “Suggests” and will be removed or moved to other packages later, making immunarch much more convenient to install.

Migration notes (0.9 → 0.10)

  • New data layer: ImmunData from immundata. Functions return new objects instead of mutating in place. This helps reproducibility and scaling.
  • API changes: large “meta-functions” like repDiversity() split into small families such as airr_diversity_*. Use the shared prefixes and the family help pages (e.g., ?airr_diversity).
  • Fewer heavy dependencies; faster install.
  • Visualisation with vis() for the new API is still evolving. New vis() is focused on fast and convenient visualisations rather than publication-ready figures. For publication-ready plots, use ggplot2.

Tutorials & How-tos

  • Single-cell tutorial (end-to-end) — loading, QC, gene usage, clonality, diversity, public indices, annotation. [immunomind.github.io]
  • Import old immunarch data — convert repLoad() to ImmunData with from_immunarch(...). [immunomind.github.io]
  • Read AIRR-C (bulk, single-cell), 10x Genomics (single-chain, paired-chain) — how-tos are available on the docs website. [immunomind.github.io]

New / reworked analysis families (API reference)

These are the main v1.0-style families used in 0.10.3. Open the help in R (?family_name) or the API pages below.

airr_stats:            https://immunarch.com/reference/airr_stats.html
airr_public:           https://immunarch.com/reference/airr_public.html
airr_clonality:        https://immunarch.com/reference/airr_clonality.html
airr_diversity:        https://immunarch.com/reference/airr_diversity.html
annotate_clonality:    https://immunarch.com/reference/annotate_clonality.html

(These pages describe all functions in each family, e.g., airr_diversity_shannon(), airr_public_jaccard(), etc.)

Not all features from immunarch 0.9 are implemented in 0.10 yet. I’m working on it.

Notes on immundata status

  • immundata powers ImmunData and lets you analyse large datasets without loading everything in RAM. It supports bulk, single-cell, and spatial workflows. [immunomind.github.io]