Source repository for the quantmsdiann methods paper. quantmsdiann is the SDRF-driven Nextflow/DIA-NN pipeline (https://github.com/bigbio/quantmsdiann); this repo holds the LaTeX manuscript and the analysis scripts that generate its figures.
- Manuscript (MCP /
elsarticle):paper/— build withcd paper && make pdf. - Datasets: see the table below; MANIFEST.md adds instrument / file-count detail.
- Reproducibility: REPRODUCIBILITY.md maps every paper number → source URL → generator command.
- Design docs: docs/superpowers/specs/.
Every figure number is computed from public data. Our reanalyses are deposited on the
PRIDE FTP under the benchmarks base ($QB below); the original deposits we compare against
live on PRIDE Archive / MassIVE. DIA-NN version trees are v1_8_1, v2_5_1, v2_5_1_enterprise,
each with quant_tables/diann_report.{parquet,tsv} (+ …site_report.parquet for phospho).
$QB = https://ftp.pride.ebi.ac.uk/pub/databases/pride/resources/proteomes/quantmsdiann-benchmarks
| Dataset (paper) | Accession | Role in paper | Our reanalysis (under $QB/) |
Original deposit |
|---|---|---|---|---|
| HeLa Astral single-cell | PXD046357 | Fig 3 (single-cell) | single-cell/PXD046357/v*/ |
PRIDE |
| HeLa One-Tip single-cell | PXD044991 | Support only³ | single-cell/PXD044991_one-tip/v*/ |
PRIDE |
| Oocyte plexDIA | MSV000093870 | Support only³ | single-cell/MSV000093870/v*/ |
MassIVE; matrix SlavovLab/single_cell_oocyte |
| NCI-60 (Guo 2019) | PXD003539 | Fig 4 (reanalysis + atlas) | quantms-collections/absolute-expression-2.0/cell-lines/PXD003539/¹ |
PRIDE |
| ProCan-DepMapSanger | PXD030304 | Fig 4 (reanalysis + atlas) | cell-lines/PXD030304/v2_5_1/² |
PRIDE |
| Sun breast PCT-SWATH | PXD004701 | Fig 4 (reanalysis + atlas) | (cell-line panel) | PRIDE |
| Tognetti breast | PXD017199 | Fig 4 (atlas only) | (cell-line panel) | PRIDE |
| MultiPro batch testbed | PXD041421 | Fig 4 (atlas only) | (cell-line panel) | PRIDE |
| Spatial DVP (MYCN) | PXD064049 | Fig 4 (reanalysis) | spatial/PXD064049/v2_5_1_enterprise/ |
PRIDE |
| Phospho (NK Fe-NTA) | PXD049692 | Support only³ | phospho/PXD049692/v*/ |
PRIDE |
| Phospho-enriched | PXD034128 | Support only³ | phospho/PXD034128-{biological-study,highspeed-DIA}/v*/ |
PRIDE |
| Galectin-1 phospho | PXD034623 | Support only³ | phospho/PXD034623/v*/ |
PRIDE |
| Scaling sweep | PXD071075 | Fig 1 (scaling) | PXD071075_cluster_sizes/ |
PRIDE |
| ProteoBench Module 7 (Astral) | — | Fig 2 (equivalence) | proteobench/ProteoBench_Module_7/v*/ |
ProteoBench |
| ProteoBench Module 9 (SC Astral) | PXD049412 | Fig 2 (equivalence); Fig 3 (A549/H460) | proteobench/PXD049412/v*/ |
PRIDE |
| ProteoBench Module 5 (diaPASEF) | PXD062685 | Fig 2 (equivalence) | proteobench/PXD062685/v*/ |
PRIDE |
| ProteoBench Module 10 (ZenoTOF) | PXD070049 | Fig 2 (equivalence) | proteobench/PXD070049/v*/ |
PRIDE |
³ Support only: processed end-to-end to demonstrate workflow support, but not benchmarked for identifications (DIA-NN counts are not comparable across versions for these acquisition modes). They appear only in the runtime figures (Fig 2b / Supp Fig S3), which show every modality finishes in minutes to hours.
¹ NCI-60 is served from the earlier quantms-collections deposition. ² ProCan is also mirrored
under $QB/cell-lines/PXD030304/v2_5_1/. Tooling: pipeline bigbio/quantmsdiann,
SDRF adapter bigbio/sdrf-pipelines (convert-diann),
containers bigbio/quantms-containers.
conda env create -f environment.yml # pins numpy<2, bundles rsvg-convert
conda activate quantmsdiann(pip/venv alternative: pip install -r analysis/requirements.txt, plus a system
rsvg-convert/librsvg for the manuscript PDF build.)
All analysis logic lives in one self-contained script, scripts/rebuild.py,
exposed as named stages:
python -m scripts.rebuild --list # list every stage + what it produces
python -m scripts.rebuild --all # data prep -> all figures -> PDFs
python -m scripts.rebuild --only single_cell_combined # rebuild one figure
python -m scripts.rebuild --only paper_numbers # re-aggregate every manuscript number
pytest tests/ # schema/regression testsEach stage writes SVGs to analysis/figures/<group>/ and derived tables to
analysis/figures/<group>/data/ (or data/). Inputs are cached under data/
(git-ignored, re-downloaded on demand; see MANIFEST.md). The PDF
build runs only when every prior stage succeeds.
All reported identification counts follow a single rule (methods.md §1):
exactly one admissible q-value filter per quantity and nothing else. No
contaminant/target filter, no positive-quantity filter (zeros are counted), decoys
dropped. Per-run: protein groups at PG.Q.Value <= 0.01, precursors at
Q.Value <= 0.01. Global (dataset totals): protein groups at Lib.PG.Q.Value <= 0.01,
precursors at Lib.Q.Value <= 0.01. The counting primitive is count_report in
scripts/rebuild.py. plexDIA and phosphoproteomics are
processed end-to-end to demonstrate workflow support, but are not benchmarked
across DIA-NN versions (their identification counts are not comparable, so they
appear only in the runtime figures).