datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
claude-protein-binder-design
Claude protein binder design — data release v1.0
1,440 de novo miniprotein binders (50 to 120 residues) against 16 targets, designed by two Claude models operating as autonomous protein-design agents (Mythos Preview, 900 designs; Opus 4.8, 540 designs) and characterized at two contract research organizations, Adaptyv Bio (cell-free expression; SPR/BLI kinetics with the design immobilized) and Twist Bioscience (Fc-fusion expression; capture SPR with a six-point antigen… See the full description on the dataset page: https://huggingface.co/datasets/Anthropic/claude-protein-binder-design.Protein-FN
Do Protein Transformers Have Biological Intelligence?
Figure 1: Illustration of two key protein motifs, i.e., His94-His96-His119and Ser29-His107-Tyr194, identified by our approach.
Deep neural networks, particularly Transformers, have been widely adopted for predicting the functional properties of proteins. In this work, we focus on exploring whether Protein Transformers can capture biological intelligence among protein sequences. To achieve our goal, we first introduce a protein… See the full description on the dataset page: https://huggingface.co/datasets/Protein-FN/Protein-FN.protein-docs
Protein Documents (Parquet)
Structured text documents encoding protein residue sequences and 3D contact maps from AlphaFold Database v4 predicted structures, stored as Parquet files. Each row is one protein document with metadata.
Source structures: timodonnell/afdb-24M and timodonnell/afdb-1.6M
Document Schemes
Each subdirectory contains documents generated with a different scheme. All schemes share leakage-resistant train/val/test splits based on structural… See the full description on the dataset page: https://huggingface.co/datasets/timodonnell/protein-docs.initial-dynamic-proteins
Initial 2,000-protein dataset
This is the canonical local root for the first complete router dataset: 1,000
single-dominant structured-state proteins (label 0) and 1,000 dynamic or
heterogeneous-state proteins (label 1). The fixed split is 1,400 train, 300
validation, and 300 test proteins.
Place Colab's completed ESMFold result files (<sequence_sha256>.npz) in
esmfold_results/. Then import them with:
uv run python scripts/esmfold_dataset.py import
The importer validates every… See the full description on the dataset page: https://huggingface.co/datasets/archiitecture/initial-dynamic-proteins.protein_data_testsplit 1, 2 -> for sequences
split 3, 4 -> for residues
ProteinsProteinGym_v1
ProteinGym
ProteinGym is a benchmark suite for evaluating protein fitness prediction and design models. It includes both substitution and indel mutations, a wide variety of experimentally assayed proteins, and clinically annotated mutations that are relevant to human disease. In total, ProteinGym includes nearly 3 million different mutations.
Dataset Details
ProteinGym is split into four separate benchmarks, based on the prediction target and the type of mutation… See the full description on the dataset page: https://huggingface.co/datasets/OATML-Markslab/ProteinGym_v1.synthetic-protein-folding-matrices
Dataset Card for CGSC Synthetic Protein Folding Matrices
Dataset Summary
This is the secondary repository for the CGSC, strictly dedicated to archiving high-resolution, uncompressed molecular dynamics (MD) trajectories. The dataset contains continuous temporal matrices representing synthetic protein folding simulations at sub-angstrom resolution.
Simulating atomic interactions over time generates colossal amounts of data. To preserve the micro-fluctuations and raw… See the full description on the dataset page: https://huggingface.co/datasets/comp-genomics-consortium/synthetic-protein-folding-matrices.viral-protein-structuresFolded/Extracted structures from PDB, AF2, ESMAtlas, and additional structures folded via AlphaFold2 on the Kempner Institute H100 GPUs.
ProteinLMBench
ProteinLMBench
A benchmark dataset for LLMs in protein related tasks.
marinfold-exp11-protein-docs-seq
marinfold-exp11-pdocs-seq
Sequence-only derivative of
eczech/marinfold-exp11-protein-docs.
For every row, the document field has been reduced to just the amino-acid sequence
portion: the <begin_sequence> tag followed by the per-residue three-letter tokens
(e.g. <begin_sequence> <MET> <LYS> <ASN> ...). The <contacts-and-distances-v1>
document-type prefix and everything from <begin_statements> onward (contacts and
distances) are removed. The token format is preserved verbatim so… See the full description on the dataset page: https://huggingface.co/datasets/eczech/marinfold-exp11-protein-docs-seq.ProteinGYM-DMS
ProteinGYM DMS Benchmark
DMS Benchmark includes three types of mutations: indels, single substitution and multiple substitutions.
Each tsv file contains all data for a single task, representing all possible mutations on a specific protein sequence. The filename corresponds to the task name. Here we use randomly cross-validation scheme, where the data for each task is randomly divided into five folds and the fold_id column indicates the fold assignment. The labels are continuous… See the full description on the dataset page: https://huggingface.co/datasets/genbio-ai/ProteinGYM-DMS.bacbench-antibiotic-resistance-protein-sequences
Dataset for antibiotic resistance prediction from whole-bacterial genomes (protein sequences)
A dataset of 25,032 bacterial genomes across 39 species with antimicrobial resistance labels.
The genome protein sequences have been extracted from GenBank. Each row contains whole bacterial genome, with spaces
separating different contigs present in the genome.
The antimicrobial resistance labels have been extracted from Antibiotic Susceptibility Test (AST) Browser, accessed 23 Oct, 2024.)… See the full description on the dataset page: https://huggingface.co/datasets/macwiatrak/bacbench-antibiotic-resistance-protein-sequences.marinfold-exp11-protein-docs
marinfold-exp11-pdocs
Quality-bucketed re-publication of the contacts-and-distances-v1-5x config from
timodonnell/protein-docs,
partitioned by the source round column:
Config
Source rounds
Approx rows
high
round 0
~1.68M
medium
round 1
~1.42M
low
round 2–4
~2.29M
Train/val/test split assignment is inherited from the source dataset (leakage-resistant
structural-cluster hashing). All columns from the source are preserved; rows are simply
partitioned by round.
See… See the full description on the dataset page: https://huggingface.co/datasets/eczech/marinfold-exp11-protein-docs.ProteinGym_v0.1
ProteinGym benchmarks overview
ProteinGym is an extensive set of Deep Mutational Scanning (DMS) assays curated to enable thorough comparisons of various mutation effect predictors indifferent regimes. It is comprised of two benchmarks: 1) a substitution benchmark which consists of the experimental characterisation of ∼1.5M missense variants across 87 DMS assays 2) an indel benchmark that includes ∼300k mutants across 7 DMS assays.
Each processed file in each benchmark corresponds to… See the full description on the dataset page: https://huggingface.co/datasets/OATML-Markslab/ProteinGym_v0.1.group_mpnn
Curated ProteinMPNN training dataset
The multi-chain training data for ProteinMPNN
Quickstart Usage
Install HuggingFace Datasets package
Each subset can be loaded into python using the Huggingface datasets library.
First, from the command line install the datasets library
$ pip install datasets
Optionally set the cache directory, e.g.
$ HF_HOME=${HOME}/.cache/huggingface/
$ export HF_HOME
then, from within python load the datasets library
>>> import datasets… See the full description on the dataset page: https://huggingface.co/datasets/ProteinMPNN/group_mpnn.protein-datasetProteinMPNN
Curated ProteinMPNN training dataset
The multi-chain training data for ProteinMPNN
Quickstart Usage
Install HuggingFace Datasets package
Each subset can be loaded into python using the Huggingface datasets library.
First, from the command line install the datasets library
$ pip install datasets
Optionally set the cache directory, e.g.
$ HF_HOME=${HOME}/.cache/huggingface/
$ export HF_HOME
then, from within python load the datasets library
>>> import datasets… See the full description on the dataset page: https://huggingface.co/datasets/RosettaCommons/ProteinMPNN.casp14-casp15-cameo-test-proteinsprotein_data_test_2protein_ligand_cofolding_posebusters
Protein-Ligand Cofolding PoseBusters
Frozen data for the Autoresearch task
protein_ligand_cofolding_posebusters.
visible/: 20 development cases.
heldout/: 42 evaluation cases.
This revision is a storage migration of the accepted 62-case union. It does
not change any case bytes, metric, threshold, or weighting. Task images
download a pinned dataset revision during image construction and run without
Hugging Face access.
secondary_structure_predictionProteinGYM-DMS-RAGProteinGYM DMS Benchmark for AIDO.RAGProteinProteinGym
ProteinGym benchmarks overview
ProteinGym is an extensive set of Deep Mutational Scanning (DMS) assays curated to enable thorough comparisons of various mutation effect predictors indifferent regimes. It is comprised of two benchmarks: 1) a substitution benchmark which consists of the experimental characterisation of ∼1.5M missense variants across 87 DMS assays 2) an indel benchmark that includes ∼300k mutants across 7 DMS assays.
Each processed file in each benchmark corresponds to… See the full description on the dataset page: https://huggingface.co/datasets/ICML2022/ProteinGym.bacbench-antibiotic-resistance-protein-sequences
Dataset for antibiotic resistance prediction from whole-bacterial genomes (protein sequences)
A dataset of 25,032 bacterial genomes across 39 species with antimicrobial resistance labels.
The genome protein sequences have been extracted from GenBank. Each row contains whole bacterial genome, with spaces
separating different contigs present in the genome.
The antimicrobial resistance labels have been extracted from Antibiotic Susceptibility Test (AST) Browser, accessed 23 Oct, 2024.)… See the full description on the dataset page: https://huggingface.co/datasets/mbafca2/bacbench-antibiotic-resistance-protein-sequences.fluorescence_prediction
Dataset Card for Fluorescence Prediction Dataset
Dataset Summary
The Fluorescence Prediction task focuses on predicting the fluorescence intensity of green fluorescent protein mutants, a crucial function in biology that allows researchers to infer the presence of proteins within cell lines and living organisms. This regression task utilizes training and evaluation datasets that feature mutants with three or fewer mutations, contrasting the testing dataset, which comprises… See the full description on the dataset page: https://huggingface.co/datasets/proteinglm/fluorescence_prediction.pdb_protein_ligand_complexes
How to use the data sets
This dataset contains about 36,000 unique pairs of protein sequences and ligand SMILES, and the coordinates
of their complexes from the PDB.
SMILES are assumed to be tokenized by the regex from P. Schwaller.
Ligand selection criteria
Only ligands
that have at least 3 atoms,
a molecular weight >= 100 Da,
and which are not among the 280 most common ligands in the PDB (this includes common additives like PEG, ADP, ..)
are considered.
Use… See the full description on the dataset page: https://huggingface.co/datasets/jglaser/pdb_protein_ligand_complexes.ProteinConformers
ProteinConformers Dataset
This repository contains the ProteinConformers dataset — a large-scale benchmark dataset comprising over 381,000 physically realistic conformations for 87 CASP targets.
📁 Contents in This Repository
This Hugging Face repo contains:
87 .zip files
The evalution code ProteinConformers_code.zip file
📬 Citation
Please cite our paper if you use this dataset:
Zhou Y., Wei C. et al. ProteinConformers: Benchmark Dataset for Simulating… See the full description on the dataset page: https://huggingface.co/datasets/Jim990908/ProteinConformers.ProteinChainsproteinLM-mixed-pretraining-v1
Pretraining mix for Protein language models
In order to construct a solid pre-training data mixture for protein language models we sample a mix of proteins from 3 sources:
MG_Prot50 (https://huggingface.co/datasets/tattabio/OMG_prot50): meta-genomic proteins created by clustering the Open MetaGenomic dataset (OMG) at 50% sequence identity.
UniRef50: UniProt proteins from across all species clustered to 50% sequences identity, downloaded form UniProt on 31th of March 2025… See the full description on the dataset page: https://huggingface.co/datasets/MichelNivard/proteinLM-mixed-pretraining-v1.