README.md

metadata

license: mit

Summary

This dataset contains conformational ensembles generated by HyRes physics-based molecular dynamics simulations, together with selected sequence properties and simulation-derived observables, for intrinsically disordered proteins and regions (IDPs/IDRs) from the human proteome in UniProt database (release 2026_01) and the DisProt database (release 2025_06 with ambiguous evidence).

Hybrid Resolution (HyRes) force field is hybrid-resolution protein model with atomistic backbone and an intermediate-resolution sidechains. By maintaining detailed backbone structure while simplifying sidechain representation, HyRes enables efficient simulation of IDPs/IDRs systems with reduced computational cost, while preserving key structural and dynamical features necessary for biologically meaningful results. The HyRes model is optimized using 20 A1-LCD variants for radius of gyration (Rg) and 15 diverse IDPs for residual helical propensity.

HyRes has been extensively benchmarked aganist experimental measurements of monomeric IDPs as well as their interactions in complex environments:

• Small-angle X-ray scattering (SAXS): benchmarked on 98 IDPs with sequence lengths ranging from 16 to 477 residues for the Rg, achieving a Pearson correlation of 0.97.
• Single-molecule FRET (smFRET): benchmarked on 16 IDPs for end-to-end distance (Re), yielding a Pearson correlation of 0.88.
• Paramagnetic Relaxation Enhancement (PRE): benchmarked on 9 IDPs, achiving a Pearson correlation of 0.72, demonstrating ability of HyRes to capture long-range interactions.
• NMR chemical shift: 40 IDPs for mean helicity, achiveing a Pearson correlation of 0.83.
• Dynamic interactions of IDPs in diverse environments: H1-ProTα is studied in the unbound, dimer, and condensate enviroment.
• Phase seperation of IDPs: benchmarked on ~64 IDPs, achieving a Pearson correlation of 0.74 for Csat.

This dataset is a valuable resource for researchers working on intrinsically disordered proteins and regions (IDPs/IDRs). It can be used for bioinformatics analyses as well as for developing deep learning models for IDP-related studies.

Data Description

Dataset	Sequences	Sequence length	Derived	Simulation details	Total frames
Human IDRs	27,489	30-999 residues	AlphaFold2 pLDDT scores based	2μ s and 4μ s for <300 and >300 residues respectively.	1000 per Sequence*
DisProt	2,383	30-778 residues	DisProt + IUPRED2a analysis	2μ s and 4μ s for <300 and >300 residues respectively.	1000 per Sequence*
Total	29,872	30-999 residues	-	-	-

* - after discarding inital 10% of the frames

Data Files

Each individual folder contains the following files:

Q9UKF5_153_198/
├── Q9UKF5_153_198.psf
├── Q9UKF5_153_198.dcd
├── Q9UKF5_153_198_Rg.npy
├── Q9UKF5_153_198_E2E.npy
├── Q9UKF5_153_198_DSSP.npy
└── Q9UKF5_153_198.xlsx
    ├── Properties
    ├── Helical_segments
    └── Residual_helicity

1. Structure File (.psf): Protein Structure File (PSF) containing the molecular topology. This file is required together with the trajectory file (.dcd) for structural analysis.

2. Trajectory File (.dcd): Molecular dynamics trajectory file containing atomic coordinates for 1,000 frames, obtained after discarding the initial 10% of the simulation frames.

3. Radius of Gyration (_Rg.npy): NumPy array containing the radius of gyration (Rg) values computed for 1,000 trajectory frames, with one value per frame. The radius of gyration provides a measure of the overall compactness of the IDPs/IDRs throughout the simulation.

4. End-to-End Distance (_E2E.npy): NumPy array containing end-to-end distance (Re) computed for 1,000 trajectory frames, with one value per frame. This quantity measures the distance between the terminal residues during the simulation.

5. DSSP Secondary Structure (_DSSP.npy): NumPy array containing DSSP-derived secondary structure assignments (8 discrete classes). Secondary structure is assigned to each residue in every simulation frame.

   Code	Structure
   H	α-helix
   B	Isolated β-bridge
   E	β-strand
   G	3-10 helix
   I	π-helix
   T	Turn
   S	Bend
   -	Loops

6. Analysis Spreadsheet (.xlsx): Excel workbook containing sequence-derived properties, simulation-derived properties, and protein annotation details, and information on helical segments and residual helicity.

   Properties

   The Properties sheet in each Excel file contains sequence-level physicochemical properties, simulation-derived observables, and biological annotations.

   • Sequence information

     Column	Description
     Name	Unique identifier for the sequence in the HyRes-IDRome dataset
     Seq	Amino acid sequence of the intrinsically disordered region (IDR)
     Start	Starting residue index in the full protein sequence
     End	Ending residue index in the full protein sequence
     Sequence length	Length of the IDR sequence
     Protein_name	Name of the parent protein

   • Sequence-derived properties The following properties were computed using localCIDER.

     Column	Description
     kappa	Extent of charge segregation in the sequence
     FCR	Fraction of charged residues
     NCPR	Net charge per residue
     Neg_Fraction	Fraction of negatively charged residues
     Pos_Fraction	Fraction of positively charged residues
     Fraction_expanding	Fraction of residues predicted to promote chain expansion (E/D/R/K/P)
     AA_fractions	Dictionary containing amino acid composition fractions
     Mean_net_charge	Absolute mean net charge
     Mean_hydropathy	Mean hydropathy computed using the normalized Kyte-Doolittle scale

   • Simulation-derived properties

     Column	Description
     Mean_Rg (Å)	Mean radius of gyration
     Std_Rg (Å)	Standard deviation of radius of gyration
     Mean_Re (Å)	Mean end-to-end distance
     Std_Re (Å)	Standard deviation of end-to-end distance
     Mean_helicity	Mean helicity averaged over residue
     Flory_exponent	Estimated Flory scaling exponent (ν)

   • DisProt-derived annotations The following annotations are available for sequences derived from DisProt.

     Column	Description
     region_id	DisProt region identifier
     acc	UniProt accession identifier
     organism	Source organism
     ncbi_taxon_id	NCBI taxonomy identifier
     disprot_id	DisProt entry identifier
     term_namespace	Controlled vocabulary namespace
     term	Functional or structural ontology term
     term_name	Human-readable ontology term
     ec	Evidence code
     ec_name	Evidence code description
     reference	Literature or database reference

   • Taxonomic annotations

     Column	Description
     Domain	Annotated protein domain
     Kingdom	Taxonomic kingdom
     Phylum	Taxonomic phylum
     Class	Taxonomic class
     Order	Taxonomic order
     Family	Taxonomic family
     Genus	Taxonomic genus
     Species	Taxonomic species annotation

     Note: There are few NaN values for the organism that have not assigned to any taxonomic category. Also, all the Viruses have the Domain "Virus" instead of NaN.

   • Human IDRome-derived annotations The following annotations are available for sequences derived from UniProt.

     Column	Description
     UniProt_ID	UniProt accession identifier
     Gene	Gene name
     Cellular Components	Gene Ontology cellular component annotations
     GO_ID:Cellular Components	GO identifiers corresponding to the cellular component annotations
     GO_ID_depth:Cellular Components	Depth of each cellular component GO term in the GO hierarchy.
     Molecular Functions	Gene Ontology molecular function annotations
     GO_ID:Molecular_Functions	GO identifiers corresponding to the molecular function annotations
     GO_ID_depth:Molecular_Functions	Depth of each molecular function GO term in the GO hierarchy.

   ---

   Helical_segments Helical segment statistics represented as (Derived from 1800 frames):

   Start residue	End residue	Segment length	Count
   1	8	8	160

   where:

   • Start residue = starting residue index of helical segment.
   • End residue = ending residue index of helical segment.
   • Segment length = length of helical segment.
   • Count = total number of frames in which the helical segment occurs.

   ---

   Residual_helicity Per-residue helicity values.

   Residue Index	Residue Name	Helicity
   1	A	0.25

   where:

   • Helicity represents the fraction of frames in which the residue adopts an α-helical conformation.

Usage

Recommended Python packages for loading and analyzing the dataset:

pip install numpy pandas openpyxl MDAnalysis datasets huggingface_hub 

---

# Whole dataset
from datasets import load_dataset
ds = load_dataset("umassmdlab/HyRes-IDRome")

# Specific IDR
from huggingface_hub import snapshot_download
Q9UKF5_153_198_path = snapshot_download(
    repo_id="umassmdlab/HyRes-IDRome",
    allow_patterns="Q9UKF5_153_198/*",
    repo_type="dataset"
)

import numpy as np
import pandas as pd
import MDAnalysis as mda

# Load radius of gyration
Rg = np.load("Q9UKF5_153_198_Rg.npy")

# Load end to end distance
Re = np.load("Q9UKF5_153_198_E2E.npy")

# Load DSSP assignments
dssp = np.load("Q9UKF5_153_198_DSSP.npy")

# Load Excel data
properties = pd.read_excel("Q9UKF5_153_198.xlsx", sheet_name="Properties")
Helical_segments = pd.read_excel("Q9UKF5_153_198.xlsx", sheet_name="Helical_segments")
Residual_helicity = pd.read_excel("Q9UKF5_153_198.xlsx", sheet_name="Residual_helicity")

# Load trajectory
u = mda.Universe("Q9UKF5_153_198.psf", "Q9UKF5_153_198.dcd")

---

Trajectory Conversion

The provided trajectories were generated using the HyRes model. To reconstruct atomistic sidechains from the coarse-grained sidechains, we recommend using CG2ALL.

conda create --name cg2all pip cudatoolkit=11.3 dgl=1.0 -c dglteam/label/cu113

Q9UKF5_153_198 workflow:

convert_cg2all \
  -p Q9UKF5_153_198.psf \
  -d Q9UKF5_153_198.dcd \
  -o Q9UKF5_153_198.aa.dcd \
  -opdb Q9UKF5_153_198.aa.pdb \
  --cg MainchainModel \
  --all

Please refer to the official CG2ALL documentation for installation instructions and advanced usage options.

Additional Info

Reference

Shanlong Li, Shrishti Barethiya, and Jianhan Chen. HyRes: Accurate Physics-Based Simulation of Dynamic Protein Structures and Interactions in Complex Environments at Scale. bioRxiv 2026. DOI: https://doi.org/10.64898/2026.06.23.734133