Proteomics Data Retrieval

Find and retrieve metadata for publicly available proteomics datasets from MassIVE and ProteomeXchange repositories. Supports searching by species, keyword, or accession, and returns detailed dataset metadata including instruments, publications, species, and post-translational modifications.

When to Use This Skill

Triggers:

• "Find proteomics datasets for [organism/disease/protein]"
• "Search MassIVE for [keyword]"
• "Get details for PXD000001" or "Look up MSV000079514"
• "What public mass spectrometry datasets exist for [topic]?"
• "Find MS datasets with [PTM type] data"
• "List recent human proteomics datasets"

Use Cases:

1. Dataset Discovery: Search repositories for proteomics experiments related to a research topic
2. Accession Lookup: Get full metadata for a known dataset accession (PXD or MSV)
3. Species-Filtered Search: Find all datasets for a specific organism
4. Cross-Repository Search: Query both MassIVE and ProteomeXchange for comprehensive coverage
5. Experimental Context: Find published datasets to validate or complement in-house results

COMPUTE, DON'T DESCRIBE

When analysis requires computation (statistics, data processing, scoring, enrichment), write and run Python code via Bash. Don't describe what you would do — execute it and report actual results. Use ToolUniverse tools to retrieve data, then Python (pandas, scipy, statsmodels, matplotlib) to analyze it.

KEY PRINCIPLES

1. ProteomeXchange is the aggregator -- it indexes datasets from PRIDE, MassIVE, PeptideAtlas, jPOST, and iProX
2. MassIVE has richer metadata -- includes summaries, keywords, modifications, and contacts
3. Search both repositories -- ProteomeXchange for breadth, MassIVE for detail
4. Species uses NCBI taxonomy IDs -- human = 9606, mouse = 10090, rat = 10116
5. Accession formats: PXD (ProteomeXchange), MSV (MassIVE) -- both accepted by MassIVE_get_dataset
6. LOOK UP DON'T GUESS -- Never assume which datasets exist, their accessions, or their instrument types. Always search and retrieve metadata to confirm.

Domain Reasoning: Dataset Quality Assessment

Dataset quality depends on instrument, sample preparation, and quantification method. TMT/iTRAQ (isobaric labeling) datasets have ratio compression and co-isolation interference biases that differ from label-free quantification (LFQ). DIA datasets require different analysis pipelines than DDA. Check the original publication for methods before reusing data in a meta-analysis or cross-study comparison. Instrument resolution (Orbitrap > ion trap) and acquisition mode (DIA > DDA for completeness) directly affect how many proteins are quantified and at what confidence.

Core Repositories Integrated

PeptideAtlas: Peptide-Level Observation Evidence

Unlike the dataset-discovery tools above (which answer "what studies exist"), PeptideAtlas_get_observed_peptides answers "for this specific protein, which peptides have actually been detected, and with what confidence" -- a direct protein-level query, not a dataset search. Constrain by biosequence_name (a UniProt accession, e.g. P02768 for serum albumin) to get that protein's peptides, or omit it with a small row_limit to sample the whole build (the default build indexes millions of peptides). Each result gives peptide_sequence, n_observations (spectral count across the whole atlas), n_samples, best_probability (PeptideProphet), and empirical_proteotypic_score (how uniquely this peptide identifies the protein, vs. one shared across a protein family) -- live-verified against real UniProt accessions. Use this to check whether a specific tryptic peptide is actually detectable before designing a targeted-MS (SRM/PRM) assay around it, or to gauge how well-characterized a protein is by MS overall.

Cancer Proteomics via PDC

Unlike MassIVE/ProteomeXchange (general-purpose MS dataset registries returning metadata/file listings), PDC is scoped to cancer proteogenomics programs and can return the actual quantitative data and per-patient clinical metadata directly — useful when the goal is a cancer-cohort reanalysis rather than a generic dataset search.

Workflow: PDC_search_studies(query="<disease/site/program>") → confirm scope with PDC_get_study_summary(pdc_study_id=...) → PDC_get_clinical_data for patient context → PDC_get_quant_data_matrix for the actual abundance values to analyze. Feed the resulting matrix into tooluniverse-proteomics-analysis (Phase 2 onward: preprocessing, differential expression, enrichment) rather than re-deriving that logic here.

PRIDE Archive: Full-Text Search, File Downloads, and Reverse Protein Lookup

PRIDE (EBI's proteomics repository) is one of the submission repositories ProteomeXchange indexes -- the same relationship MassIVE has to ProteomeXchange, just a different partner archive. Reach for PRIDE specifically (rather than, or in addition to, ProteomeXchange/MassIVE) when you need: real full-text search (ProteomeXchange's query param matches thinly against title/accession; PRIDE's PRIDE_search_proteomics matches against title, description, species, and tissue, so a query like "phosphorylation" or "COVID-19 proteome" returns far more relevant hits), actual file download URLs (FTP and Aspera links per file -- none of MassIVE/ProteomeXchange/PDC's tools in this skill return downloadable links, only metadata), or a protein-first search ("which datasets contain protein X" has no equivalent in MassIVE or ProteomeXchange here).

Workflow: PRIDE_search_proteomics(query="<keyword>") → PRIDE_get_project(accession=...) for methods/DOI detail → PRIDE_get_project_files(accession=...) for actual download links. Or, protein-first: PRIDE_get_projects_for_protein(accession="<UniProt AC>") → PRIDE_get_project on each hit to evaluate relevance before downloading.

Verified live: PRIDE_search_proteomics(query="breast cancer", page_size=3) returned real recent submissions (e.g. PXD083320, publication date 2026-08-28) with full instrument/organism/disease metadata already populated in the search response itself -- no separate detail call needed to triage relevance.

Workflow Overview

Query (keyword / species / accession)
|
+-- PHASE 0: Input Resolution
|   Determine search type: keyword, species, or accession lookup
|
+-- PHASE 1: Repository Search
|   Search MassIVE and/or ProteomeXchange based on query type
|
+-- PHASE 2: Dataset Detail Retrieval
|   Get full metadata for promising hits
|
+-- PHASE 3: Result Synthesis
    Compile datasets with metadata, publications, and relevance assessment

Phase 0: Input Resolution

Objective: Determine the query type and prepare appropriate search parameters.

Decision Logic

• Accession provided (e.g., PXD000001, MSV000079514):
• PXD accession: call ProteomeXchange_get_dataset and optionally MassIVE_get_dataset
• MSV accession: call MassIVE_get_dataset
• Skip Phase 1, go directly to Phase 2
• Species name provided (e.g., "human", "mouse"):
• Map to NCBI taxonomy ID: human=9606, mouse=10090, rat=10116, yeast=559292, zebrafish=7955, fly=7227, worm=6239, arabidopsis=3702
• Use MassIVE_search_datasets with species filter
• Keyword provided (e.g., "phosphoproteomics", "breast cancer"):
• Use ProteomeXchange_search_datasets with query parameter
• MassIVE does not support keyword search -- use ProteomeXchange for keyword queries

Phase 1: Repository Search

Objective: Find relevant datasets across repositories.

Tools

MassIVE_search_datasets:

• page_size: Number of results to return (integer, max 100, default 10)
• species: NCBI taxonomy ID string to filter by species (e.g., "9606" for human)
• Returns: Array of dataset objects with accessions (array), title, summary, species, instruments, keywords
• Note: No keyword/text search parameter -- filtering is by species only

ProteomeXchange_search_datasets:

• query: Optional search filter -- keyword or dataset accession (e.g., "phosphoproteomics", "PXD")
• limit: Max results (1-50, default 10)
• Returns: {data: [{accession, title, species}], metadata: {source, total_returned, query}}

Workflow

1. For species-specific search:

• Call MassIVE_search_datasets(page_size=20, species="9606") for species-filtered results
• Call ProteomeXchange_search_datasets(limit=20) for broader listing

1. For keyword search:

• Call ProteomeXchange_search_datasets(query="keyword", limit=20)
• Review titles for relevance

1. For comprehensive discovery:

• Call both tools in parallel
• Merge results, deduplicate by accession (PXD accessions may appear in both)

Response Format Notes

• MassIVE_search_datasets: Returns a direct array (no {data: ...} wrapper)
• ProteomeXchange_search_datasets: Returns {data: [...], metadata: {...}}

Phase 2: Dataset Detail Retrieval

Objective: Get full metadata for datasets of interest.

Tools

MassIVE_get_dataset:

• accession: Dataset accession -- accepts both MSV and PXD formats (e.g., "MSV000079514", "PXD003971")
• Returns: Object with accessions, title, summary, species, instruments, keywords, contacts, publications, modifications

ProteomeXchange_get_dataset:

• px_id: ProteomeXchange identifier in PXD format (e.g., "PXD000001")
• Returns: {data: {px_id, title, species, identifiers, instruments, publications, file_count}, metadata: {...}}

Workflow

1. For each promising dataset from Phase 1, call the appropriate detail tool
2. Extract key metadata: title, species, instruments, publications (PubMed/DOI), modifications
3. For PXD accessions: prefer ProteomeXchange_get_dataset for file count; use MassIVE_get_dataset for richer summary/keywords

Key Fields to Extract

• title: Dataset name/description
• species: Organism(s) studied
• instruments: Mass spectrometer(s) used (e.g., Orbitrap, Q Exactive, TripleTOF)
• publications: PubMed IDs and DOIs for associated papers
• modifications: PTMs studied (from MassIVE only)
• file_count: Number of raw files (from ProteomeXchange only)
• keywords: Topic tags (from MassIVE only)

Phase 3: Result Synthesis

Objective: Compile and present dataset results in a structured format.

Report Format

# Proteomics Dataset Search Results
**Query**: [original query]
**Date**: YYYY-MM-DD
**Repositories searched**: MassIVE, ProteomeXchange

## Summary
Found N datasets matching [criteria].

## Datasets

### 1. [Title]
- **Accession**: PXD/MSV number
- **Species**: [organism]
- **Instruments**: [MS platforms]
- **Publications**: [PubMed IDs / DOIs]
- **Modifications**: [PTMs if available]
- **Files**: [count if available]
- **Summary**: [brief description]

### 2. [Title]
...

## Data Gaps
[Note any limitations in search coverage]

Tool Parameter Reference

Response Format Notes:

• MassIVE_search_datasets: Returns direct array of dataset objects (no wrapper)
• MassIVE_get_dataset: Returns direct object (no wrapper)
• ProteomeXchange_search_datasets: Returns {data: [...], metadata: {...}}
• ProteomeXchange_get_dataset: Returns {data: {...}, metadata: {...}}

Fallback Strategies

Common Species Taxonomy IDs

Interpretation Framework

Interpreting dataset search results:

• Datasets with both MassIVE and ProteomeXchange accessions generally have richer metadata; MassIVE provides summaries and keywords while ProteomeXchange provides file counts -- cross-reference both for a complete picture.
• Instrument type determines data quality ceiling: high-resolution instruments (Orbitrap, timsTOF) produce higher mass accuracy and more reliable quantification than older ion trap platforms.
• A dataset lacking a peer-reviewed publication may still be valuable, but its experimental design and processing pipeline cannot be independently verified -- weight such datasets lower in meta-analyses.

Synthesis questions to address in the report:

1. Do multiple independent datasets for the same organism/condition show consistent protein identifications, or do discrepancies suggest batch effects?
2. Is the instrument platform appropriate for the analysis type (e.g., DIA requires high-resolution; TMT requires MS3 or calibrated MS2)?
3. Are the reported PTM types and species consistent with the user's research question, or is additional filtering needed?

Limitations

• MassIVE: No keyword/text search -- only species-based filtering via species parameter
• ProteomeXchange: Limited metadata in search results (no summaries or keywords); get details via Dataverse_get_dataset
• No full-text search: Cannot search within dataset descriptions or abstracts across repositories
• No download via MassIVE/ProteomeXchange/PDC: those tools retrieve metadata only, not raw data files. PRIDE is the exception -- PRIDE_get_project_files returns real FTP/Aspera URLs per file, verified live
• Rate limits: Both APIs may throttle under heavy load; keep page_size/limit reasonable
• Coverage: ProteomeXchange is the most comprehensive but may lag behind individual repositories for very recent submissions

Integration with Other Skills

References

• MassIVE: https://massive.ucsd.edu
• ProteomeXchange: http://www.proteomexchange.org
• PRIDE: https://www.ebi.ac.uk/pride
• ProXI API: https://github.com/PRIDE-Archive/proxi-schemas