v0.3.0

v0.3.0 - 2026-06-29

Incoporated multiple requests .

Requests

• Added alevin_usa: alevin-fry quantifies in USA mode and adds spliced, unspliced and ambiguous assays. Requires alevin_sketch.
• Documented that BD sample tags come from the WTA reads, with no separate tag FASTQ input.
• The transcriptome input now accepts a plain .fa, not only .fa.gz.
• A sample can list several fastqs per mate; they are concatenated before processing, paired by order.
• Sample tags can be called without starsolo, via an alignment-free search that writes the same count table.
• The cross-pipeline comparison report is built only with two or more aligners; single-aligner runs skip it.
• Fixed sample tag alignment that rejected every read as too short; STAR now filters on matched bases.
• Added per-sample sampletags to declare which tags a sample carries, as a list or tag-to-label mapping.
• Each aligner's counts are split into one HDF5 SCE per tag; demultiplexing moved from the report into demux_sampletags.R.
• Downgraded the setuptools pin in the pyroe env to fix a pkg_resources ModuleNotFoundError.
• Sped up the per-sampletag split with sampletag_split_backend: memory | delayed (default memory); memory reads the source once.
• Added output_format: sce | h5ad | both (default sce); h5ad written via anndataR next to each SCE and split.

Extras

• Extended the integration tests: memory and delayed sampletag scenarios, dry-run coverage, an integration PR label, and failure artifact uploads.
• Fixed the simulated sampletag path: sampletag_fa resolved from the wrong directory, so no sampletag reads were produced.
• Pinned the self-compiled kallisto to v0.52.0 and bustools to 0.45.1 from the kallisto conda env.
• get_txp2gene reads transcript_id and gene_id GTF attributes by name, so any attribute order works.
• Added cell_filtering: none to keep every observed barcode across STARsolo, alevin and kallisto.