datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
many-peptides-md
[!IMPORTANT]
Critical Update
The original 8AA TICA models within subsampled_trajectories/*/8AA/*.npz employed a CA-only atom selection. These models are not valid for comparison to results in our paper.
Updated files (uploaded 15/12/2025) now contain corrected models. If you previously downloaded this dataset, please re-download to ensure accurate results.
Note: Codebase references to tica_features_ca must now be replaced with tica_features. This was resolved in our codebase by PR #26.
Note:… See the full description on the dataset page: https://huggingface.co/datasets/transferable-samplers/many-peptides-md.vacuum_peptides
Vacuum Peptide Ensembles
Molecular-dynamics trajectories derived from ManyPeptidesMD. This repository provides vacuum ensembles for a small subset of peptides, initialized from ManyPeptidesMD conformers.
Overview
Path
Model
Use
trajectories/v1/
Amber14, peptide only
Gas-phase sampling
webdatasets/v1/300K/
Streamable samples from the 300 K state
Model training
webdatasets/v1/300-500K/
Streamable samples from all REMD states
Temperature-conditioned… See the full description on the dataset page: https://huggingface.co/datasets/niklastr/vacuum_peptides.microsolvated_peptides
Microsolvated Peptide Ensembles
This repository provides explicit-water microsolvated structures for a subset of peptides initialized from ManyPeptidesMD conformers.
Overview
Path
Model
Use
trajectories/v1/
Amber14 in TIP3P, cropped to the nearest 128 waters
Peptide-solvent sampling
webdatasets/v1/train_300K/
Streamable samples from the 300 K state
Model training
webdatasets/v1/train_300-500K/
Streamable samples from all REMD states… See the full description on the dataset page: https://huggingface.co/datasets/niklastr/microsolvated_peptides.LRGB_Peptides-func
LRGB Peptides-func
Peptides-func (Peptides functional) dataset, part of Long Range Graph Benchmark (LRGB) [1]. It is intended to be used through
scikit-fingerprints library.
The task is to predict functional properties of peptides.
Characteristic
Description
Tasks
10
Task type
classification
Total samples
15535
Recommended split
stratified random
Recommended metric
AUPRC
References
[1]
Dwivedi, Vijay Prakash, et al.
"Long Range Graph Benchmark"… See the full description on the dataset page: https://huggingface.co/datasets/scikit-fingerprints/LRGB_Peptides-func.LRGB_Peptides-struct
LRGB Peptides-struct
Peptides-struct (Peptides structural) dataset, part of Long Range Graph Benchmark (LRGB) [1]. It is intended to be used through
scikit-fingerprints library.
The task is to predict structural properties of peptides. Note that this is raw data, whereas the original paper [1] specifies
that targets should be standardized (mean 0, standard deviation 1) before training and evaluation. scikit-fingerprints does this by
default in the loader function, otherwise this… See the full description on the dataset page: https://huggingface.co/datasets/scikit-fingerprints/LRGB_Peptides-struct.solvated_peptides
Neutral Peptide Ensembles
Molecular-dynamics ensembles combining matched neutral peptide samples from vacuum REMD, ManyPeptidesMD OBC1 implicit-solvent trajectories, and neutral microsolvated peptide frames. The vacuum and microsolvated ensembles are initialized from ManyPeptidesMD conformers.
Overview
Path
Model
Use
webdatasets/v1/joint_neutral/base_temperature/
Amber14 vacuum, Amber14/OBC1, and Amber14/TIP3P microsolvated clusters
300 K training… See the full description on the dataset page: https://huggingface.co/datasets/niklastr/solvated_peptides.peptides_soluble
Dataset Details
Dataset Description
Solubility was estimated by retrospective analysis of electronic laboratory notebooks.
The notebooks were part of a large effort called the Protein Structure Initiative and consider sequences
linearly through the following stages: Selected, Cloned, Expressed, Soluble, Purified, Crystallized,
HSQC (heteronuclear single quantum coherence), Structure, and deposited in PDB. The peptides were identified
as soluble or insoluble by… See the full description on the dataset page: https://huggingface.co/datasets/jablonkagroup/peptides_soluble.peptides_nonfouling
Dataset Details
Dataset Description
Non-fouling is defined as resistance to non-specific interactions.
A non-fouling peptide (positive example) is defined using the mechanism proposed in
ref white2012decoding. Briefly, ref white2012decoding, showed that the exterior surfaces
of proteins have a significantly different frequency of amino acids, and this increases
in aggregation prone environments, like the cytoplasm. Synthesizing self-assembling peptides
that follow this… See the full description on the dataset page: https://huggingface.co/datasets/jablonkagroup/peptides_nonfouling.peptides_hemolytic
Dataset Details
Dataset Description
Hemolysis is referred to the disruption of erythrocyte
membranes that decrease the life span of red blood cells and causes
the release of Hemoglobin. It is critical to identify non-hemolytic
antimicrobial peptides as a non-toxic and safe measure against bacterial
infections. However, distinguishing between hemolytic and non-hemolytic
peptides is a challenge, since they primarily exert their activity at the
charged surface of the… See the full description on the dataset page: https://huggingface.co/datasets/jablonkagroup/peptides_hemolytic.many_peptides_forces
Many Peptides Forces
A fork of the ManyPeptidesMD
single-frame WebDataset in which every frame carries AMBER14 + OBC1
implicit-solvent potential energies and forces in addition to the original
atom coordinates.
Release Status
This dataset is complete. All 5,000 shards are uploaded and hash-verified.
All 434,000,000 frames passed structure, coordinate-identity, and finite-value
checks.
Planned Contents
5,000 shard tars single_frames_forces/0000.tar… See the full description on the dataset page: https://huggingface.co/datasets/niklastr/many_peptides_forces.autonlp-data-peptidesDeep learning the collisional cross sections of the peptide universe from a million experimental values
Data generated from MaxQuant output
wget https://ftp.pride.ebi.ac.uk/pride/data/archive/2020/12/PXD017703/HeLa_200ng_Library_MaxQuant.zip
unzip HeLa_200ng_Library_MaxQuant.zip
awk -F '\t' '{print $1,",",$40}' evidence.txt > pepCCS.csv
wc pepCCS.csv
352111 1056333 12736697 pepCCS.csv
Code
peptides-functional
peptides-functional
Dataset Summary
Dataset
Domain
Task
Node Feat. (dim)
Edge Feat. (dim)
Perf. Metric
Peptides-func
Chemistry
Graph Classification
Atom Encoder (9)
Bond Encoder (3)
AP
Dataset
# Graphs
# Nodes
μ Nodes
μ Deg.
# Edges
μ Edges
μ Short. Path
μ Diameter
Peptides-func
15,535
2,344,859
150.94
2.04
4,773,974
307.30
20.89±9.79
56.99±28.72
Additional Information
Dataset Curators
Vijay Prakash Dwivedi… See the full description on the dataset page: https://huggingface.co/datasets/LRGB/peptides-functional.Dataset-Signal-Peptides
Description
This dataset contains 25693 amino acid sequences and labels on each amino acid.
Protein Format: AA sequence
Splits
traing: 20490
valid: 2569
test: 2634
Related paper
The dataset is from SignalP 6.0 predicts all five types of signal
peptides using protein language models.
Label
Each amino acid has 7 classes:
S (0): Sec/SPI signal peptide | T (1): Tat/SPI or Tat/SPII signal peptide | L (2): Sec/SPII signal peptide |
P (3): Sec/SPIII signal… See the full description on the dataset page: https://huggingface.co/datasets/SaProtHub/Dataset-Signal-Peptides.peptides-structural
peptides-functional
Dataset Summary
Dataset
Domain
Task
Node Feat. (dim)
Edge Feat. (dim)
Perf. Metric
Peptides-struct
Chemistry
Graph Regression
Atom Encoder (9)
Bond Encoder (3)
MAE
Dataset
# Graphs
# Nodes
μ Nodes
μ Deg.
# Edges
μ Edges
μ Short. Path
μ Diameter
Peptides-struct
15,535
2,344,859
150.94
2.04
4,773,974
307.30
20.89±9.79
56.99±28.72
Additional Information
Dataset Curators
Vijay Prakash Dwivedi… See the full description on the dataset page: https://huggingface.co/datasets/LRGB/peptides-structural.bbb-peptides
BBB Peptide Dataset (TFG)
Curated blood–brain barrier (BBB) permeability dataset for peptide sequences with Boltz-predicted 3D structures and physicochemical descriptors.
Summary
Field
Value
Rows
825
With structure
825
BBB+
410
BBB−
415
Variant
full
Files
peptides.parquet — one row per peptide: sequence, label, splits, physicochemical features, structure quality metrics, relative structure paths… See the full description on the dataset page: https://huggingface.co/datasets/manumartinm/bbb-peptides.peptidespeptides-new-datapeptides-v2peptides-fullpeptides-v4peptides-v3research-peptides-reference
Research Peptides Reference Dataset
A clean, machine-readable reference table of research-grade peptides commonly
discussed in biochemistry and drug-discovery literature. Each entry combines a
curated research category with verified physicochemical properties pulled from
PubChem (PUG REST): PubChem CID, molecular
formula, molecular weight, canonical SMILES and IUPAC name.
The goal is a small, high-signal starting point for cheminformatics, tabular ML,
educational tooling and… See the full description on the dataset page: https://huggingface.co/datasets/PeptidosSuplementos/research-peptides-reference.peptides-v5-two-classDataset-Signal-Peptides
Description
This dataset contains 25693 amino acid sequences and labels on each amino acid.
Protein Format: AA sequence
Splits
traing: 20490
valid: 2569
test: 2634
Related paper
The dataset is from SignalP 6.0 predicts all five types of signal
peptides using protein language models.
Label
Each amino acid has 7 classes:
S (0): Sec/SPI signal peptide | T (1): Tat/SPI or Tat/SPII signal peptide | L (2): Sec/SPII signal peptide |
P (3): Sec/SPIII signal… See the full description on the dataset page: https://huggingface.co/datasets/SeprotHub/Dataset-Signal-Peptides.peptides-full-tokenizerbiased_peptides
Biased Peptides
Prefix-sampled training WebDatasets derived from ManyPeptidesMD. Each configuration keeps the ManyPeptidesMD train split and the single-frame WebDataset layout, but draws unique frames only from the first 1/d of every 200,000-frame trajectory. Use these sets to measure how incomplete equilibration changes model training.
Configurations
Path
Prefix
Unique frames / molecule
Shards
Members / shard
prefix_1_0/
1
20,000
5,000
86,800… See the full description on the dataset page: https://huggingface.co/datasets/niklastr/biased_peptides.peptidestestWorkflow-prose-peptidespeptides-completeamino-club-peptides-promo-code