quantmsdiann manuscript: analysis code and run provenance

Analysis scripts, run inputs and figure sources for the quantmsdiann paper. quantmsdiann is the SDRF-driven Nextflow/DIA-NN workflow at https://github.com/bigbio/quantmsdiann (v2.2.0 was used for this paper).

Folder	Content
sdrf/	The SDRF used by every run, verbatim; per-run Nextflow parameters (params.json, launch.txt, diann_config.cfg); runs.tsv (index of all 38 runs); the FASTA databases (fasta/). Supplementary Table 1 with links: sdrf/README.md.
scripts/rebuild.py	All main-text and Supplementary Note 4-6 figures and manuscript numbers, as named stages.
scripts/review_*.py	Supplementary Figures S6-S8 and Tables S3-S5 (added in revision).
scripts/build_*.py, stage_pride_submission.py, zenodo_upload.py	Table S6 and the PRIDE / Zenodo deposition of the results.
analysis/figures/	Figure sources (SVG) and the small derived tables they are drawn from.
methods.md	The counting rule used for every identification number.
REPRODUCIBILITY.md	Figure/table -> public source -> command.

Data

Every number is computed from public data. Our reanalysis results (DIA-NN reports, matrices, logs) are on the PRIDE FTP under https://ftp.pride.ebi.ac.uk/pub/databases/pride/resources/proteomes/quantmsdiann-benchmarks/ (and quantms-collections/ for three cell-line cohorts), one folder per dataset and DIA-NN version; they are also deposited as a PRIDE reanalysis dataset and on Zenodo (see the paper's Data Availability). The PRIDE dataset is available at https://www.ebi.ac.uk/pride/archive/projects/PXD085732. The original deposits we compare against are on PRIDE / MassIVE; the accessions are listed in sdrf/README.md. Large inputs are never committed: each stage downloads what it needs into data/ (git-ignored) on first use.

What is in PXD085732

The PRIDE dataset holds 88 files; raw files are not re-uploaded.

PRIDE type	Files	Content
EXPERIMENTAL_DESIGN	38	The SDRF of every run, as used (<dataset>_<version>_<name>.sdrf.tsv).
SEARCH	35	One <dataset>_<version>.zip per run with a DIA-NN search: quant_tables/ (report, matrices, stats), pipeline_info/, sdrf/ configs, pmultiqc/, run_metadata.json, nextflow.log.
OTHER	14	9 <PXD>.qpx.zip exports (PXD046357 and PXD049412 have one per DIA-NN version), and 5 PXD071075_v2_5_0_sweep_*cores.zip with only launch.txt and nextflow_trace.txt.
RAW	1	raw_files.raw.zip: raw_files.txt and raw_files.tsv, every raw file processed with its URL in the original deposit.

There are more SDRFs (38) than search zips (35) because the two sets cover different runs:

• 33 runs have both an SDRF and a DIA-NN search zip.
• 5 SDRFs belong to the PXD071075 node-count sweep (10, 50, 100, 200 and 300 cores, Fig. 2a). These runs measured wall-clock time only, so their zips keep the Nextflow traces and no DIA-NN report (type OTHER). The five SDRFs are the same file, one per sweep run.
• 2 search zips are the standalone DIA-NN runs of the paired comparison (Paired-standalone-PXD049412, Paired-standalone-PXD070049, Supplementary Note 8). Plain DIA-NN takes no SDRF; their quantmsdiann counterparts carry the SDRF.

runs.tsv, extra.config and Table S6 are not in the PRIDE dataset; they are in this repository and its Zenodo record (https://doi.org/10.5281/zenodo.23322361).

Setup and use

conda env create -f environment.yml          # pins numpy<2, bundles rsvg-convert
conda activate quantmsdiann
python -m scripts.rebuild --list             # stages and what they produce
python -m scripts.rebuild --all              # data -> figures -> numbers
python -m scripts.rebuild --only fig2_validation
pytest tests/

(pip alternative: pip install -r analysis/requirements.txt.)

Identification counts follow one rule (methods.md §1): global counts at Lib.Q.Value / Lib.PG.Q.Value <= 0.01, per-run counts at Q.Value / PG.Q.Value <= 0.01, decoys dropped, no other filter. plexDIA and phosphoproteomics runs show workflow support (runtime) only; their identification counts are not benchmarked across DIA-NN versions.