PROSPECT & PROSPECT PTM: A Collection of Proteomics Datasets for Machine Learning and Deep Learning

PROSPECT and PROSPECT PTM (PROteometools SPECTrum compendium) are large annotated datasets, leveraging the raw data from ProteomeTools [1] [2], including unmodified and modified peptide sequences (with Post-Translational Modifications, PTMs).

Datasets

The unmodified PROSPECT dataset includes examples without modifications. The PROSPECT PTM datasets include 4 datasets, each representing a common subgroup of modified peptide sequences. Dataset TMT includes peptide sequences labeled with Tandem Mass Tags, which are introduced experimentally for multi-plex analysis of samples. Dataset Multi-PTM includes examples rich with 13 unique PTM-residue pairs occurring naturally on proteins. We also curated and annotated the dataset TMT-PTM, which includes six unique PTM-residue pairs occurring in TMT-labeled sequences. Dataset Test-PTM, comprises 21 unique PTM-residue pairs and is well-suited as a hold-out/test dataset.

Dataset	Packages	Pools	Unique Peptides	Precursors	Sepctra	Annotated Peaks	Raw Peaks
PROSPECT Unmodified dataset	12	983	838 K	1.24 M	61.7 M	5.7 B	24 B
PROSPECT PTM TMT dataset	11	1000	714 K	820 K	28.2 M	1.8 B	11.2 B
PROSPECT PTM Multi-PTM dataset	15	400	307 K	413 K	19.6 M	2 B	6 B
PROSPECT PTM TMT-PTM dataset	10	327	159 K	189 K	7.8 M	511 M	3 B
PROSPECT PTM Test-PTM dataset	25	56	10 K	15.6 K	3 M	193 M	732 M

Features

Access to a collection of large annotated Mass Spectrometry datasets including peptide sequences with Post-Translational Modifications (PTMs).
Utilities for downloading and splitting data.

Installation

Install with:

pip install git+https://github.com/wilhelm-lab/PROSPECT

Usage

Downloading a Dataset

Using `prospectdataset`:

Install and import the package

import prospectdataset as prospect

To download a specficic dataset, pass one or more of the available record names:
- prospect
- tmt
- multi_ptm
- tmt_ptm
- test_ptm
```
prospect.download_dataset(record = 'prospect', save_directory = SAVE_PATH)
```
Available record names and URLs can be displayed as follows:

print(prospect.AVAILABLE_DATASET_URLS)

{'prospect': 'https://zenodo.org/record/6602020',
 'multi_ptm': 'https://zenodo.org/record/7998644',
 'tmt': 'https://zenodo.org/record/8003138',
 'tmt_ptm': 'https://zenodo.org/record/8003152',
 'test_ptm': 'https://zenodo.org/record/8003156'}

To download data for retention time prediction only:

prospect.download_dataset(task = 'retention-time', save_directory = SAVE_PATH)

To download data for MS/MS spectrum prediction (includes both metadata and spectra):
```
prospect.download_dataset(task = 'all', save_directory = SAVE_PATH)
```
To download only one package from a specific dataset (for a faster download and a smaller dataset to experiment with), enter a substring from the package name, package names are in the respective Zenodo URL in the table. For example to download the package TUM_missing_first, the following would download the meta data file for the specfic package from the PROSPECT dataset:
```
prospect.download_dataset(
  record = 'prospect', 
  task = 'retention-time',
  save_directory = SAVE_PATH,
  select_package = 'missing'
) 
```
All downloaded files are in the parquet format. They can be easily read using panda’s pd.read_parquet(). For faster loading, we recommend using fastparquet as an engine, in case it fails for some reason, pyarrow can also be used.

df = pd.read_parquet(PARQUET_FILEPATH, engine='fastparquet')

From Zenodo:

Download and unzip from the respective zenodo record, available records are:

Splitting and Filtering Scripts

The three following bash scripts can be used to perform data splitting, merge files, and to filter the retention time data (iRT). Once the package is installed, they are accessbile as bash commands on the system level. Use the option -h to see the arguments of each script.

shuffle-split-data: Shuffle and split data
merge-files: merge multiple files into one
filter-irt: filter iRT data to prepare for training.

License

The project is licensed under the MIT license.

Dataset Hosting

The dataset is hosted on Zenodo PROSPECT DOI and is licensed under Creative Commons Attribution 4.0 International.

Citation

If you use PROSPECT and/or PROSPECT-PTM, please cite our papers PROSPECT: Labeled Tandem Mass Spectrometry Dataset for Machine Learning in Proteomics:

@inproceedings{prospect,
  title={PROSPECT: Labeled Tandem Mass Spectrometry Dataset for Machine Learning in Proteomics},
  author={Shouman, Omar and Gabriel, Wassim and Giurcoiu, Victor-George and Sternlicht, Vitor and Wilhelm, Mathias},
  booktitle={Thirty-sixth Conference on Neural Information Processing Systems Datasets and Benchmarks Track}
}

References

[1] Daniel P Zolg, Mathias Wilhelm, Karsten Schnatbaum, Johannes Zerweck, Tobias Knaute, Bernard Delanghe, Derek J Bailey, Siegfried Gessulat, Hans-Christian Ehrlich, Maximilian Weininger, et al. Building proteometools based on a complete synthetic human proteome. Nature methods, 14(3):259–262, 2017.

[2] Zolg, D. P., Wilhelm, M., Schmidt, T., Médard, G., Zerweck, J., Knaute, T., … & Kuster, B. (2018). ProteomeTools: Systematic characterization of 21 post-translational protein modifications by liquid chromatography tandem mass spectrometry (LC-MS/MS) using synthetic peptides. Molecular & Cellular Proteomics, 17(9), 1850-1863.