Module configuration

Authored documents

A module is a TOML file and the SDRF-Proteomics table it names. The SDRF owns the sample design: one row per run, its data file, assay name, condition, and the quantity of every species in the mixture. The TOML holds only the scoring settings an SDRF cannot express.

sdrf = "module.sdrf.tsv"

[species.YEAST]
organism = "saccharomyces cerevisiae"
suffix = "_YEAST"
color = "#88ccef"

[species.HUMAN]
organism = "homo sapiens"
suffix = "_HUMAN"

[general]
min_count_multispec = 1
level = "ion"
default_cutoff_min_feature = 1
max_nr_observed = 6

[run_aliases]
"run_A1.raw" = ["run_A1_uncalibrated"]

The SDRF path is relative to the TOML's directory. APB2 reads it through apb2.api.SdrfSource.read; the columns ProteoBench uses are:

Fields used by scoring

• Each species' expected A/B ratio is the quotient of its condition A and condition B quantities. Every scored species needs one quantity per condition, in one unit, identical across that condition's rows.
• [species.<NAME>] links an SDRF organism to the FASTA entry-name suffix that marks its proteins, and to an optional plot color. Every spiked organism needs an entry.
• Species and contaminants come from apb-fasta's varm["fasta_validation"], so scoring requires peptide verification. A feature belongs to a species when a FASTA protein containing its peptide has that organism (suffix without _, e.g. HUMAN). It is a contaminant when any such protein is, or when apb2 marked it apb_Contaminant; it is dropped as a decoy when apb2 marked it apb_Decoy.
• [run_aliases] lists further run names that vendor tables report, keyed by SDRF data file. An observation may match a run's identifier, an alias, or its sample name; successful matching always records the module's canonical metadata.
• min_count_multispec controls exclusion of features assigned to multiple species.
• level selects the single APB2 quantification level to annotate and score.
• default_cutoff_min_feature selects the aggregate projection shown at the top of the score.
• max_nr_observed controls the complete cutoff-indexed score range.

Unknown TOML sections, such as ProteoBench's [reference_database] and [validation], are accepted. Only the resolved sample design and scoring fields are embedded in APB metadata. The packaged SDRFs pass the official sdrf-pipelines structural validation; APB itself does not validate SDRF structure or ontology terms.

Supported quantitative modules

HY omits ECOLI from the SDRF and the [species] table; HYE includes it. The quantitative scorer supports the eight HYE/HY modules used by the legacy APB integration and the plasma module:

Load from Python

Load them by name without locating package files:

from apb_proteobench.api import load_packaged_module

module = load_packaged_module("dia_singlecell")

load_packaged_module() accepts only these nine modules, which are validated for the current quantitative scorer; any other name raises and lists them.

Packaged for planned support

Two newer ProteoBench module documents are also packaged so this repository owns the complete upstream catalogue:

load_packaged_module() refuses both. load_packaged_entrapment_module("entrapment_dia_astral") loads the entrapment module for EntrapmentAnalyzer; see the Python API.

The nine supported modules are derived from ProteoBench's module TOMLs, with their sample design moved into packaged SDRFs; the two others are unchanged copies. All files are pinned by checksum. Their source revision, original paths, and support status are recorded in MODULES_NOTICE.md.

Persisted form

The embedded sample design is normalized as a column mapping (raw_file, sample_name, and condition arrays), rather than TOML's list of tables, so the same JSON-compatible value round-trips through AnnData's HDF5 representation. Authored aliases participate in annotation and remain traceable to the checksum-identified source module; source.sdrf records the SDRF's name and checksum beside the TOML's.

Continue to Result layout for the physical and storage-neutral metadata locations.