Sdrf:templates

SDRF Template System

Templates define which columns are required for a given experiment type. Each SDRF can declare one or more templates via comment[sdrf template] columns.

Specification Data (always read from source)

The authoritative source for all template information is in the spec/ submodule:

• Template manifest: Read spec/sdrf-proteomics/sdrf-templates/templates.yaml
• Individual templates: Read spec/sdrf-proteomics/sdrf-templates/{name}/{version}/{name}.yaml
• Column definitions: Read spec/sdrf-proteomics/TERMS.tsv (the usage field shows which templates include each column)

Always read templates.yaml when answering questions about templates, versions, inheritance, or mutual exclusivity. Never rely on memorized template data — the spec evolves.

How to Read templates.yaml

The manifest file lists every template with these fields:

• name — template identifier (e.g., ms-proteomics, human)
• version — current version (e.g., 1.1.0)
• extends — parent template with version constraint (e.g., sample-metadata@>=1.0.0)
• description — what the template adds
• usable_alone — whether it can be used without other templates (only ms-proteomics and affinity-proteomics)
• excludes — templates that are mutually exclusive with this one
• layer — which selection layer it belongs to

How to Read Individual Template YAMLs

Each template has a YAML file at spec/sdrf-proteomics/sdrf-templates/{name}/{version}/{name}.yaml. These define the columns the template adds, with requirement levels (required/recommended/optional).

How to Find Columns for a Template

Two ways:

1. Read the individual template YAML → lists columns with requirement levels
2. Read TERMS.tsv → filter rows where usage contains the template name

Template Layers (Methodology — stable across versions)

Templates are organized into layers. Each layer serves a different purpose:

1. Technology (REQUIRED — pick exactly one): The measurement technology used.

• ms-proteomics — mass spectrometry experiments
• affinity-proteomics — Olink, SomaScan, and other affinity platforms
• These are mutually exclusive

1. Sample/Organism (RECOMMENDED — pick at most one organism template):

• human — Homo sapiens samples
• vertebrates — mouse, rat, zebrafish, etc.
• invertebrates — Drosophila, C. elegans, insects
• plants — Arabidopsis, crops
• Organism templates are mutually exclusive with each other
• clinical-metadata — clinical studies (can combine with organism templates)
• oncology-metadata — cancer studies (extends clinical-metadata)

1. Experiment (OPTIONAL — pick any applicable):

• cell-lines — cultured cell lines
• dia-acquisition — DIA/SWATH (extends ms-proteomics)
• single-cell — single-cell proteomics (extends ms-proteomics)
• immunopeptidomics — MHC peptide studies (extends ms-proteomics)
• crosslinking — XL-MS (extends ms-proteomics)
• olink — Olink PEA (extends affinity-proteomics)
• somascan — SomaScan (extends affinity-proteomics)

1. Metaproteomics (SPECIAL — uses its own sample scheme):

• metaproteomics — environmental/microbiome base (excludes sample-metadata)
• human-gut — host-associated microbiome
• soil — soil metaproteomics
• water — aquatic metaproteomics

Mutual Exclusivity Rules (Methodology)

1. ms-proteomics ↔ affinity-proteomics — different technologies
2. human ↔ vertebrates ↔ invertebrates ↔ plants — pick at most one organism
3. metaproteomics excludes sample-metadata — uses its own sample scheme
4. olink and somascan extend affinity-proteomics — cannot combine with ms-proteomics
5. dia-acquisition, single-cell, immunopeptidomics, crosslinking extend ms-proteomics — cannot combine with affinity-proteomics

Read templates.yaml for the full excludes field on each template to verify mutual exclusivity.

Template Selection Decision Tree (Methodology)

Is it mass spectrometry?
├── YES → ms-proteomics
│   ├── Human samples? → + human
│   │   ├── Cancer study? → + oncology-metadata
│   │   ├── Clinical trial / drug treatment? → + clinical-metadata
│   │   └── Cell lines from human? → + human + cell-lines
│   ├── Mouse/rat/zebrafish? → + vertebrates
│   │   └── Cell lines from animal? → + vertebrates + cell-lines
│   ├── Drosophila/C. elegans? → + invertebrates
│   ├── Plant? → + plants
│   ├── Environmental/microbiome? → metaproteomics (REPLACES organism layer)
│   │   ├── Human gut? → + human-gut
│   │   ├── Soil? → + soil
│   │   └── Water? → + water
│   ├── DIA/SWATH/diaPASEF? → + dia-acquisition
│   ├── Single-cell proteomics? → + single-cell
│   ├── MHC/immunopeptidome? → + immunopeptidomics
│   └── Cross-linking MS? → + crosslinking
│
└── NO (affinity-based) → affinity-proteomics
    ├── Olink? → + olink
    └── SomaScan? → + somascan

Template Inheritance (Methodology)

When templates are combined, the validator merges all columns from all ancestors. If a parent says OPTIONAL but a child says REQUIRED → REQUIRED wins (strictest requirement).

Read templates.yaml to see the full inheritance tree via the extends field on each template. The general structure is:

• base → sample-metadata → technology + organism + experiment templates
• base → metaproteomics → environment-specific templates (excludes sample-metadata)

Common Template Combinations (Methodology)

| Experiment Type | Templates | |-----------------|-----------| | Human tissue DDA (label-free or TMT) | ms-proteomics, human | | Human cancer clinical trial | ms-proteomics, human, clinical-metadata, oncology-metadata | | Mouse tissue DIA | ms-proteomics, vertebrates, dia-acquisition | | Human cell line study | ms-proteomics, human, cell-lines | | Single-cell proteomics (human) | ms-proteomics, human, single-cell | | Immunopeptidomics (human) | ms-proteomics, human, immunopeptidomics | | Cross-linking MS (human) | ms-proteomics, human, crosslinking | | Gut metaproteomics | ms-proteomics, metaproteomics, human-gut | | Soil metaproteomics | ms-proteomics, metaproteomics, soil | | Olink plasma study | affinity-proteomics, human, olink | | SomaScan serum study | affinity-proteomics, human, somascan | | Drosophila DDA | ms-proteomics, invertebrates | | Arabidopsis study | ms-proteomics, plants | | Drug treatment study (human cells) | ms-proteomics, human, clinical-metadata, cell-lines |

How to Detect Templates from Existing SDRF

When an SDRF file already exists, detect templates from:

1. Metadata column: comment[sdrf template] → e.g., NT=ms-proteomics;VV=v1.1.0
2. Organism: characteristics[organism] → Homo sapiens = human, Mus musculus = vertebrates
3. Technology type: "proteomic profiling by mass spectrometry" → ms-proteomics
4. Acquisition method: comment[proteomics data acquisition method] → DIA = dia-acquisition
5. Cell line columns present: characteristics[cell line] → cell-lines
6. MHC columns present: characteristics[mhc protein complex] → immunopeptidomics
7. Crosslinker columns: comment[cross-linker] → crosslinking
8. Single cell columns: characteristics[single cell isolation protocol] → single-cell
9. Environmental columns: characteristics[environmental sample type] → metaproteomics
10. Oncology columns: characteristics[tumor grading] → oncology-metadata
11. Olink columns: comment[panel name] (or legacy comment[olink panel]) → olink
12. SomaScan columns: comment[somascan menu] → somascan

How to Respond to User Queries

If they describe an experiment:

1. Walk through the decision tree and recommend a specific template combination
2. Read templates.yaml to confirm templates exist and get current versions
3. Read TERMS.tsv to list the columns the combination adds
4. Explain WHY each template was chosen

If they ask about a specific template:

1. Read its entry in templates.yaml for version, extends, excludes, description
2. Read its individual YAML for the columns it adds (with requirement levels)
3. Explain when to use it, what it inherits from, and what it's mutually exclusive with

If they ask about differences between templates:

Compare side by side — read both template YAMLs for their column lists.

If they provide an SDRF and ask "what templates should this use?":

Auto-detect from the content using the detection rules above.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: bigbio
• Source: bigbio/sdrf-skills
• License: MIT
• Homepage: https://sdrf.quantms.org

Install and usage instructions live in the source repository linked above.