ProteomeTools

training · DDA · 9 versions · 53 papers

Synthetic peptides covering the human proteome, measured deliberately rather than harvested, so the ground truth is the peptide that was synthesised. The main source of large labelled training sets in this field.
Kind training
Acquisition DDA
Organisms Homo sapiens (synthetic)
Home https://www.proteometools.org/

Synthetic peptides covering the human proteome, measured deliberately rather than harvested, so the ground truth is the peptide that was synthesised. The main source of large labelled training sets in this field.

Versions

Parts I-III

The three PRIDE submissions as deposited.

released 2017-02-09

Where it lives:

  • PRIDE · PXD004732 Part I: complete synthetic human proteome
  • PRIDE · PXD010595 Part II: Prosit
  • PRIDE · PXD021013 Part III: HLA class I & II and non-tryptic peptides

high-confidence (InstaNovo)

The all-confidence search results reduced to best PSMs and published as a split parquet dataset with its own DOI.

spectra 2,655,403 · train 2,132,847 · validation 257,187 · test 265,369 · released 2024-12-01

Where it lives:

Assembled from 3 third-party submissions:

21-PTM subset

Synthetic peptides carrying 21 distinct post-translational modifications, the usual source for evaluating modified-peptide sequencing.

spectra 41,158 · released 2018-04-10

Where it lives:

Bruker timsTOF

The same peptide pools re-run on a Bruker timsTOF Pro, so a model trained on the Orbitrap runs can be tested for instrument transfer rather than only for sequence.

Where it lives:

non-tryptic (Bruker)

HLA class I and II ligands plus Asp-N and Lys-N peptides on the Bruker instrument, the non-tryptic counterpart of Part III.

Where it lives:

TMT 6-plex

Tryptic and non-tryptic sequences carrying TMT 6-plex labels, where the label itself changes the fragment ladder a de novo model has to read.

Where it lives:

HC-PT (NovoBench)

NovoBench’s ~10% subsample of the high-confidence InstaNovo split, 1 PTM, mean peptide length 12.53. Papers call it “the HC-PT dataset” and cite the InstaNovo paper for it, so a number reported on HC-PT is a number on a tenth of that corpus under NovoBench’s split.

train 213,284 · validation 25,718 · test 26,536 · introduced by NovoBench: Benchmarking Deep Learning-based De Novo Peptide Sequencing Methods in Proteomics

No public address: this version is named in the literature but cannot be downloaded.

all-confidence (InstaNovo)

Every PSM of the ProteomeTools MaxQuant searches regardless of quality, about 28M spectra over the same 742k peptides as the high-confidence split. InstaNovo evaluates on it as AC-PT.

released 2023-08-30 · introduced by De novo peptide sequencing with InstaNovo: Accurate, database-free peptide identification for large scale proteomics experiments

Assembled from 3 third-party submissions:

single amino-acid variants (SAAV)

Synthetic peptides each carrying one residue substitution relative to a human reference peptide, searched with MaxQuant; PepPr evaluates on 74,575 PSMs of it. PRIDE PXD081052 was private on 2026-10-28, available to peer reviewers on request.

Where it lives:

Deposited by (2)

More than one paper here means one piece of work published twice, a preprint and its version of record; a deposit itself happens once.

Trained on by (25)

Evaluated on by (32)

Used by (16)

Papers that ran on it where the catalog has not recorded whether for training or for evaluation.

Methods on these papers (36)

Taken from the describing links only, so a paper that merely ran a tool on this data does not make that tool a method of it.

Seen in the charts

Back to the full map

Back to top