{"success":true,"data":{"author":"Wilhelm Lab, Technical University of Munich","bucketAccessType":"public","bucketName":"Wilhelmlab/prospect-ptms-irt","bucketPrefix":"","category":"Proteomics","cloudProvider":"huggingface","createdAt":"2026-11-17T14:12:21.963Z","description":"A proteomics machine learning dataset from the Wilhelm lab (Technical University of Munich) processed and split for peptide retention time prediction, with indexed retention time (iRT) targets for peptides carrying post-translational modifications. It is derived from the PROSPECT PTMs datasets on Zenodo, which are built on liquid chromatography tandem mass spectrometry runs of the ProteomeTools synthetic peptide libraries in PRIDE. Two configurations: default, with train, validation and test splits drawn from the PROSPECT PTMs sets, and holdout, from the PROSPECT Test-PTM set for final benchmarking and model comparison. Parquet files under data/ and holdout/, about 100 MB; load with load_dataset(\"Wilhelmlab/prospect-ptms-irt\"). Companion to the MS2 intensity and precursor charge-state sets from the same lab, for training Prosit-style retention time predictors used in DIA library generation and PSM rescoring. Curated by the Wilhelm lab under CC BY 4.0; cite the PROSPECT paper and the ProteomeTools papers.","displayTitle":"PROSPECT PTMs: indexed retention time (iRT) prediction benchmark for modified peptides","huggingFaceUrl":"https://huggingface.co/datasets/Wilhelmlab/prospect-ptms-irt","id":"9fc18b62-8777-595e-aba4-8045e678866e","license":"CC-BY-4.0","licenseEvidenceUrl":"https://huggingface.co/datasets/Wilhelmlab/prospect-ptms-irt","objectCountEstimate":6,"probeStatus":"ok","referenceUrl":"https://github.com/wilhelm-lab/PROSPECT","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"Wilhelmlab/prospect-ptms-irt\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"Wilhelmlab/prospect-ptms-irt\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=Wilhelmlab, license:cc-by-4.0, size_categories:1M<n<10M, format:parquet, downloads=816"],"format":"mixed","generatedAt":"2026-11-17T14:23:18.096Z","layout":{"example":"data/test-00000-of-00001.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"data/test-00000-of-00001.parquet","truncatedAt":512}],"sampleObjects":["data/test-00000-of-00001.parquet","data/train-00000-of-00001.parquet","data/val-00000-of-00001.parquet","holdout/test-00000-of-00001.parquet"],"sidecars":[{"bytes":"---\nlicense: cc-by-4.0\ndataset_info:\n- config_name: default\n  features:\n  - name: raw_file\n    dtype: string\n  - name: scan_number\n    dtype: int64\n  - name: retention_time\n    dtype: float64\n  - name: indexed_retention_time\n    dtype: float64\n  - name: package\n    dtype: string\n  - name: modified_sequence\n    dtype: string\n  splits:\n  - name: train\n    num_bytes: 131622696\n    num_examples: 1042693\n  - name: val\n    num_bytes: 37745818\n    num_examples: 299055\n  - name: test\n    num_bytes: 18701783\n    num_examples: 148195\n  download_size: 94258505\n  dataset_size: 188070297\n- config_name: holdout\n  features:\n  - name: raw_file\n    dtype: string\n  - name: scan_number\n    dtype: int64\n  - name: modified_sequence\n    dtype: string\n  - name: retention_time\n    dtype: float64\n  - name: indexed_retention_time\n    dtype: float64\n  splits:\n  - name: test\n    num_bytes: 6092691\n    num_examples: 45264\n  download_size: 1776929\n  dataset_size: 6092691\nconfigs:\n- config_name: default\n  data_files:\n  - split: train\n    path: data/train-*\n  - split: val\n    path: data/val-*\n  - split: test\n    path: data/test-*\n- config_name: holdout\n  data_files:\n  - split: test\n    path: holdout/test-*\ntags:\n- proteomics\n- mass-spectrometry\n- retention-time\n---\n\n# PROSPECT PTMs - Retention Time Prediction \n\nA mass-spectrometry dataset for applied machine learning in proteomics, processed and split for the task of retention time prediction.\n\n## Dataset Details\n\n- **Curated by:** Wilhelmlab - Technical University of Munich - School of Life Sciences - Germany\n- **License:** CC-BY4.0\n\n### Dataset Sources\n\nThe data is based on the PROSPECT PTMs datasets hosted in Zenodo. \n- **Repository:** https://github.com/wilhelm-lab/PROSPECT\n\n## Uses\n\nThe dataset is intended to be used for retention time prediction given a peptide sequence. Optionally, addiional inputs can be used as features to encode PTMs or features describing the experimental setup.\n\n## Dataset Structure\n\nThe dataset has two configurations:\n- default: this is the train/val/test data that is based on all the Zenodo PROSPECT PTMs datasets [3][4][5].\n- holdout: this is the holdout datasets that can be eventually used to evaluate a model capable of processing PTMs. It is based on the PROSPECT Test-PTM dataset on Zenodo [6].\n\nUse one of the following lines to load the respective configuration:\n\n```python\n# main data for training and evaluation; contains train, val, tesst splits\nmain_dataset = load_dataset(\"Wilhelmlab/prospect-ptms-irt\")\n\n# holdout data for final benchmarking and model comparison; contains test split only\nholdout_dataset = load_dataset(\"Wilhelmlab/prospect-ptms-irt\", \"holdout\")\n```\n\n## Dataset Creation\n\n### Curation Rationale\n\nThe dataset is intended to serve as a reference benchmark dataset for retention time prediction, processed, split, and ready-to-use for developing deep learning models for this specific task.\n\n\n### Source Data\n\nThe upstream source data is based on the ProteomeTools datasets available on PRIDE [1][2].\n\n#### Data Collection and Processing\n\n[More Information Needed]\n\n\n### Annotations\n\nThe indexed retention time was calculated based on the retention time measured during Liquid Chromotography.\n\n#### Personal and Sensitive Information\n\nThe dataset does not contain any personal, sensitive, or private data.\n\n### Recommendations\n\nWe recommend using the holdout configuration for solely evaluation models at the end of the research iteration. \n\n## Citation\n\n**BibTeX:**\n\n[More Information Needed]\n\n**APA:**\n\n\n## References\n\n[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.\n\n[2] Daniel Paul Zolg, Mathias Wilhelm, Tobias Schmidt, Guillaume Médard, Johannes Zerweck, Tobias Knaute, Holger Wenschuh, Ulf Reimer, Karsten Schnatbaum, and Bernhard Kuster. Pro- teometools: Systematic characterization of 21 post-translational protein mo","encoding":"utf8","key":"README.md","name":"README.md","size":4920,"truncated":true}],"sourceObjectKey":"data/test-00000-of-00001.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"Wilhelmlab/prospect-ptms-irt\")\nprint(ds)","starterSamplePath":"data/test-00000-of-00001.parquet","totalObjects":6},"summary":"A proteomics machine learning dataset from the Wilhelm lab (Technical University of Munich) processed and split for peptide retention time prediction, with indexed retention time…","title":"PROSPECT PTMs: indexed retention time (iRT) prediction benchmark for modified peptides","updatedAt":"2026-11-17T20:04:43.780Z","visibility":"public","websiteUrl":"https://huggingface.co/datasets/Wilhelmlab/prospect-ptms-irt"}}