Phase 5: DDA data analysis support via DIA-NN (v2.3.2+)

[ypriverol]

Overview

DIA-NN v2.3+ adds beta-stage DDA/DDA-PASEF data analysis via the --dda flag.

Confirmed by Vadim Demichev (Feb 2026):

"2.3.1 and 2.3.2, better just require 2.3.2"

Capabilities & Limitations

Use cases: Legacy DDA reanalysis, offline-fractionation library creation, immunopeptidomics, metaproteomics

Limitations (from DIA-NN docs):

• Only trust: q-values, PEP values, RT/IM values, Ms1.Apex.Area quantity, Normalisation.Factor
• PTM localization probabilities unreliable with DDA data
• MBR requires MS2-level evidence
• No isobaric labeling or reporter-tag quantification

Tasks

New params:

Parameter	Type	Default	Description
diann_dda	boolean	false	Enable DDA analysis mode

• Pass --dda to all DIA-NN steps when enabled
• Version guard: require DIA-NN >= 2.3.2
• Add --dda to blocked list for ALL DIA-NN modules
• Decide integration strategy: new workflow (workflows/dda_diann.nf) or dda_engine parameter in existing DDA workflow

Open Questions

• Should DIA-NN DDA be a separate workflow or integrated into the existing DDA branch?
• Are there additional flags beyond --dda for DDA-PASEF data?

Test Plan

• Run PXD039580 (DDA SILAC HeLa) with --diann_dda true and DIA-NN v2.3.2
• Verify --dda flag passed to all DIA-NN steps
• Version guard blocks DDA with DIA-NN < 2.3.2

Test Datasets:

Dataset	Description	Instrument
PXD039580	DDA + plexDIA SILAC HeLa	Q Exactive HF (has SDRF)
PXD024809	Immunopeptidomics DDA component	Orbitrap Fusion Lumos

[ypriverol]

Research Summary (April 2026)

DIA-NN DDA Mode

• --dda is the only flag needed — pipeline steps (library gen → preliminary → empirical → individual → final) are identical to DIA
• Works library-free (FASTA + in-silico), DDA-PASEF also just uses --dda
• Limitations: only trust q-values, PEP, RT/IM, Ms1.Apex.Area; PTM localization unreliable; MBR limited

Blocker: No DIA-NN 2.3.x Release

• Latest DIA-NN release is 2.2.0 (May 2025). Version 2.3.x does not exist yet on GitHub
• No container available — ghcr.io/bigbio/diann:2.3.2 cannot be built
• Decision: wait until 2.3.2 is released and containerized before implementing

Implementation Plan (ready to execute when unblocked)

1. Build + push ghcr.io/bigbio/diann:2.3.2 container
2. Add diann_dda param (boolean, default: false) + version guard (>= 2.3.2)
3. Pass --dda to all DIA-NN steps when enabled; add to blocked lists
4. Allow DDA acquisition method through create_input_channel validation
5. Add conf/diann_versions/v2_3_2.config + conf/tests/test_dda.config
6. Use existing BSA dataset from bigbio/quantms-test-datasets/testdata/lfq_ci/BSA/ (6 mzML, ~35 MB, has SDRF)
7. Add CI job for DDA test profile
8. Document DDA limitations (PTM, MBR, isobaric labeling not supported)

Test Dataset

The BSA dataset at bigbio/quantms-test-datasets/testdata/lfq_ci/BSA/ is ideal — small, label-free DDA, has SDRF, already hosted.