docs: clarify params.json vs SDRF-derived parameters

params_<timestamp>.json records launch-time pipeline parameters only.
Modifications, enzyme and mass tolerances are taken from the SDRF per
sample at runtime and override the defaults shown there (fixed mods and
enzyme have no parameter at all, so they never appear in params.json).
Document that the authoritative resolved values live in
results/sdrf/diann_config.cfg + the SDRF, and add an 'SDRF parsing
outputs' section describing sdrf/. No behavior change.

Author: ypriverol

docs/output.md

@@ -136,6 +136,26 @@ mdata = mu.read("results/qpx/PXD019909.h5mu")
 - `execution_trace.txt` - Detailed execution trace
 - `pipeline_dag.html` - DAG visualization
 - `software_versions.yml` - Software versions used
+- `params_<timestamp>.json` - The pipeline parameters **as supplied at launch** (CLI + config defaults)
+
+> [!IMPORTANT]
+> **`params_<timestamp>.json` records launch-time parameters, not the per-sample values resolved from the SDRF.**
+> Several search parameters — **modifications** (fixed and variable), **enzyme**, and **precursor/fragment mass tolerances** — are taken from the **SDRF at runtime, per sample**, and override the pipeline defaults. So an entry like `variable_mods: "Oxidation (M)"` or `precursor_mass_tolerance: 5` in `params.json` is the **default/fallback** and may not be what a given run actually used. Fixed modifications and enzyme have no pipeline parameter at all and therefore never appear in `params.json`.
+>
+> The **authoritative record of what DIA-NN actually used** is in [`sdrf/`](#sdrf-parsing-outputs):
+>
+> - `sdrf/diann_config.cfg` — the resolved DIA-NN flags (`--var-mod`, `--fixed-mod`, enzyme, etc.).
+> - `sdrf/diann_design.tsv` and the published SDRF — the per-sample experimental design.
+>
+> Per-sample mass tolerances and m/z windows are applied directly on the DIA-NN command line for each file (visible in the verbose `diann_preprocessing/` logs / each task's `.command.sh`).
+
+### SDRF parsing outputs
+
+`sdrf/` contains the parsed experimental design and the configuration derived from the input SDRF — the **source of truth for modifications, enzyme, labelling, and tolerances** actually used by the run:
+
+- `diann_config.cfg` - DIA-NN configuration generated by `parse_sdrf convert-diann`, including the fixed/variable modifications (`--fixed-mod` / `--var-mod`) and enzyme.
+- `diann_design.tsv` - DIA-NN experiment design table.
+- the input SDRF (`*.sdrf.tsv`) - the original annotation the above is derived from.
 
 ### pmultiqc
 

docs/parameters.md

@@ -2,6 +2,9 @@
 
 This document lists every pipeline parameter organised by category. Default values come from `nextflow.config`; types and constraints come from `nextflow_schema.json`.
 
+> [!IMPORTANT]
+> **Some parameters are taken from the SDRF per sample at runtime and override the defaults below.** Modifications (variable via `--variable_mods`; fixed modifications and **enzyme** have no parameter at all) and precursor/fragment **mass tolerances** are read from the SDRF for each file when present, falling back to the value here only if the SDRF does not specify them. As a result, the value shown here (and in `pipeline_info/params_<timestamp>.json`) is the **launch-time default/fallback**, not necessarily what a given run used. The authoritative resolved values are written to `results/sdrf/diann_config.cfg` (modifications, enzyme) and applied per file on the DIA-NN command line (mass tolerances, m/z windows). See [Output: Nextflow pipeline info](output.md#nextflow-pipeline-info).
+
 ## 1. Input/Output Options
 
 | Parameter            | Type                    | Default     | Description                                                                                                                                                                                                                                                                        |