Deliver indels, and read two SNPs of one codon as one change

[Paururo]

What changed and why

Two things the pipeline did not deliver: its indels, and the amino-acid change of a codon that carries more than one SNP.

Indels. The main VCF holds SNPs only, and of the indels only the homozygous ones reached a filtered file (the legacy var.homo.indel.vcf). A minority indel was in none, and a minority frameshift is how resistance through Rv0678 often starts. Indels now get the definition a SNP has: each sample's <sample>.<ref>.indels.ann.vcf.gz keeps every indel FreeBayes made, normalised, with FILTER PASS where it meets the hom or het rule and LowSupport where not, so a minority call keeps its fraction instead of vanishing. <samplesheet>_indel_matrix.tsv holds the run's indels the way the SNP matrix holds its SNPs, and the report's matrix panel gains an Indels view with frameshifts marked.

Codon-level changes, with get_MNV. Two SNPs in one codon change one amino acid, and annotated a base at a time they name two changes that are not there. GET_MNV runs get_MNV 1.1.5 on each sample's passing SNPs and indels with the BAM they were called on, and reads every codon whole on the reads that span it. get_MNV also writes a combined change for two SNPs of one codon on different molecules, which no read carries, so <samplesheet>_mnv.tsv keeps only the codons at least --mnv_min_reads reads carry whole. The SNP matrix gains codon_change / codon_change_samples, and the report marks those SNPs and their cells.

On reads simulated from an H37Rv window (rpoB + Rv0678), called with FreeBayes and annotated with SnpEff:

Codon	Per-base annotation says	Read whole	Reads carrying it	Kept
rpoB 445 (CAC>GAG)	His445Asp + His445Gln	His445Glu	103 / 104	yes
rpoB 460 (GAG>TTG), minority	Glu460* (stop) + Glu460Val	Glu460Leu (MNV-masked)	23 / 99	yes
rpoB 430, two SNPs on different molecules	Leu430Met + Leu430Leu	Leu430Ile	0 / 115	no

The same run's indel matrix: Rv0678 c.144dupC p.Glu49fs, c.301delG p.Ala101fs at AF 0.1852, and c.402_404delACG p.Arg135del at AF 0.8061; the sample without them reads AF 0 at the anchor's depth.

A report fix on the way. The report read single-base records only, so every SNP FreeBayes wrote as an MNP (CAC>GAG in one record) was in the SNP matrix and in none of the report's tables. It now reads such a record as the SNPs it is made of.

Outputs and defaults

• New per sample: <sample>.<ref>.indels.ann.vcf.gz (+ .tbi; indels.vcf.gz with --annotate_main_vcf false), var.het.indel(.ann).vcf in the legacy split set, and mnv/ with get_MNV's TSV, VCF, summary and run manifest.
• New per run: <samplesheet>_indel_matrix.tsv and <samplesheet>_mnv.tsv.
• Changed: <samplesheet>_snp_matrix.tsv gains two columns after aa_change, codon_change and codon_change_samples, when get_MNV runs. Anything that reads that file by column position needs to account for them; with --run_mnv false the layout is unchanged.
• New parameters, on by default: --make_indel_matrix, --run_mnv, --mnv_min_reads 2, --mnv_gff_features CDS, --mnv_translation_table 11, --mnv_container.
• New processes: GET_MNV (per sample, 2 CPUs, 2 GB), MNV_TABLE, INDEL_MATRIX, each with a stub.
• Container: GET_MNV runs in quay.io/biocontainers/get_mnv 1.1.5, pinned by digest, until the pipeline image carries it. The Dockerfile now installs get_mnv=1.1.5, but this PR publishes no image (docker-publish runs on a version tag); after the next release, bump params.container and empty --mnv_container. The Singularity profiles keep the docker:// scheme and the docker profile strips it, as for the main image.
• Garnatxa: the launcher runs main, so the first launch after the merge runs GET_MNV and pulls its image from quay.io.
• A gzipped GFF, which the samplesheet accepts, is decompressed for get_MNV, which reads plain text only.

How it was verified

tests/run_tests.sh        # ruff clean · unit 2,218 passed, 39 skipped · JS 120 passed · pipeline 39 passed, and the container test added since, 7 of 7 on its own
mkdocs build --strict     # passes

• New tests: the indel matrix (10), the run's codon table on rows from a real get_MNV run (7), the report's indel view and codon marks including the MNP expansion (7), the SNP matrix's codon columns and its unchanged layout without them (4), the indel soft filter against a real bcftools (1), the GET_MNV script with a gzipped and a plain GFF (2), and the get_MNV image's scheme per profile (1). -profile test_full reaches the three new processes.
• End to end on simulated data: FreeBayes, SnpEff, get_MNV 1.1.5 and the new scripts, with the results above.
• Not run: the full pipeline on real data, and the get_MNV container itself (no Docker daemon locally; the same get_MNV version was run from bioconda).

Checklist

• tests/run_tests.sh passes locally (all four legs)
• ruff check . is clean, and I did not restyle code the change does not touch
• New or changed behaviour has a test, and tests pin what the code does today
• A new process in modules/ has a stub: block, and -profile test_full reaches it
• tests/data/ was not edited by hand (fixtures come from make_test_data.py)
• User-visible changes have a CHANGELOG.md entry and, where relevant, a docs update