Clarify DIA-NN FDR parameters: rename protein_level_fdr_cutoff and expose matrix-level FDR controls

[ypriverol]

Problem

The pipeline parameter protein_level_fdr_cutoff (default: 0.01) is misleadingly named for DIA workflows. It maps to DIA-NN's --qvalue flag, which per the official documentation controls precursor-level q-value filtering:

--qvalue [X] specifies the precursor-level q-value filtering threshold

The current schema description says "The experiment-wide protein (group)-level FDR cutoff" — this is inaccurate for DIA-NN.

Historical context

The parameter name was originally aligned with the DDA/OpenMS pathway, where it correctly maps to -proteinFDR in ProteomicsLFQ and isobaric workflow modules. Now that quantmsdiann is a standalone workflow, this alignment is no longer necessary and causes confusion.

Missing matrix-level FDR parameters

DIA-NN provides matrix-level FDR controls that are not exposed as pipeline parameters:

DIA-NN flag	Description	Currently exposed?
--qvalue [X]	Precursor-level q-value threshold for main report	Yes, as protein_level_fdr_cutoff (wrong name)
--matrix-qvalue [X]	Q-value threshold for output matrices (pg_matrix, pr_matrix, etc.)	No
--matrix-spec-q [X]	Run-specific protein q-value filtering for protein/gene matrices	No

Per DIA-NN docs:

"They [matrices] are filtered at 1% FDR, using global q-values for protein groups and both global and run-specific q-values for precursors. Additional 5% run-specific protein-level FDR filter is applied to the protein matrices, use --matrix-spec-q to adjust it."

Use case: proteogenomics / variant detection

This is especially important for proteogenomics workflows (e.g., COSMIC variant detection). Variant proteins are typically supported by only 1-2 unique peptides, so:

• Precursor-level FDR (Q.Value) is the relevant filter — variant-specific peptides pass at 1%
• Protein-level FDR (Protein.Q.Value) is always 1.0 for variants because they lack enough unique peptides for independent protein-level confidence
• Users need explicit control over matrix filtering to avoid losing valid variant detections

Proposed changes

Parameter renaming

Current	Proposed	Maps to	Description
protein_level_fdr_cutoff	diann_precursor_qvalue	--qvalue	Precursor-level q-value threshold (DIA-NN main report)
(not exposed)	diann_matrix_qvalue	--matrix-qvalue	Q-value threshold for output matrices
(not exposed)	diann_matrix_spec_q	--matrix-spec-q	Run-specific protein q-value for protein matrices

Note: protein_level_fdr_cutoff should be kept for the DDA/OpenMS pathway where it correctly maps to -proteinFDR. The new DIA-NN-specific parameters only apply when running the DIA workflow.

DIA-NN version compatibility

• --qvalue: Available in all supported versions (1.8.1+)
• --matrix-qvalue: Available since DIA-NN 1.8+
• --matrix-spec-q: Available since DIA-NN 1.8+ (documented in README for adjusting the default 5% run-specific protein-level FDR on matrices)

The pipeline currently supports DIA-NN versions 1.8.1 through 2.3.2, so all three parameters are compatible with all supported versions.

Files to modify

1. nextflow.config — Add new parameter defaults (diann_precursor_qvalue = 0.01, diann_matrix_qvalue = 0.01, diann_matrix_spec_q = 0.05)
2. nextflow_schema.json — Add parameter definitions with accurate descriptions
3. modules/local/diann/final_quantification/main.nf — Use new parameters in diann command, add --matrix-qvalue and --matrix-spec-q to blocked flags list
4. modules/local/diann/diann_msstats/main.nf — Update --qvalue_threshold reference if needed

References

• DIA-NN Command-line reference
• DIA-NN Discussion #1035 — FDR in proteomics & data filtering
• DIA-NN Issue #925 — Protein q-value in command line

[jpfeuffer]

But we also have peptide/precursor level FDR cutoff.
And it would be nice if you could unify matrix qvalue and the protein qvalue cutoff from openms.

[ypriverol]

This is moving to quantmsdiann.