PROSPECT: Labeled Tandem Mass Spectrometry Dataset for Proteomics

PROSPECT is a large annotated dataset for machine learning in proteomics, leveraging raw data from the ProteomeTools project. It consists of 12 packages with metadata and annotation files, where annotations are sub-organized into pools of approximately 1,000 peptides each. The dataset was created by Omar Shouman at the Technical University of Munich and is available via the paperswithcode platform.

Use Cases

• Training peptide identification models based on annotated tandem mass spectrometry spectra.
• Developing machine learning methods for proteomics based on the organized pool structure.
• Benchmarking spectral prediction algorithms using the provided unique identifiers to trace raw data.
• Studying peptide fragmentation patterns using the large-scale annotated compendium.

Strengths

• Dataset is structured into 12 distinct packages, facilitating organized access.
• Annotation files are sub-organized into pools of approximately 1,000 peptides to maintain high identification rates.
• Provides unique identifiers linking each example to its original raw data file in the ProteomeTools repository.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific modeling tasks.
• Last update date is unknown; freshness unverified.

Provenance

Source

Raw data originates from the ProteomeTools project, available on the PRIDE repository in three parts.

Collection Method

Dataset was created by annotating and structuring the raw ProteomeTools data into metadata and annotation files.