README.md

metadata

license: cc-by-4.0
dataset_info:
  features:
    - name: Entry ID
      dtype: string
    - name: Experimental Method
      dtype: string
    - name: Matthews Coefficient
      dtype: float64
    - name: Percent Solvent Content
      dtype: float64
    - name: Crystallization Method
      dtype: string
    - name: pH
      dtype: float64
    - name: Crystal Growth Procedure
      dtype: string
    - name: Temp (K)
      dtype: float64
    - name: Deposition Date
      dtype: string
    - name: Release Date
      dtype: string
    - name: Number of Non-Hydrogen Atoms per Deposited Model
      dtype: float64
    - name: Total Number of Polymer Instances (Chains)
      dtype: float64
    - name: Total Number of Polymer Residues per Deposited Model
      dtype: float64
    - name: Number of Water Molecules per Deposited Model
      dtype: float64
    - name: Disulfide Bond Count per Deposited Model
      dtype: float64
    - name: Molecular Weight per Deposited Model
      dtype: float64
    - name: Number of Distinct Protein Entities
      dtype: float64
    - name: Refinement Resolution (Å)
      dtype: string
    - name: Structure Determination Methodology
      dtype: string
    - name: Average B Factor
      dtype: float64
    - name: R Free
      dtype: float64
    - name: R Work
      dtype: float64
    - name: Structure Title
      dtype: string
    - name: Sequence
      dtype: string
    - name: Entity Polymer Type
      dtype: string
    - name: Polymer Entity Sequence Length
      dtype: float64
    - name: Entity Macromolecule Type
      dtype: string
    - name: Total Number of polymer Entity Instances (Chains) per Entity
      dtype: float64
    - name: Molecular Weight (Entity)
      dtype: float64
    - name: Macromolecule Name
      dtype: string
    - name: EC Number
      dtype: string
    - name: EC Provenance Source
      dtype: string
    - name: Source Organism
      dtype: string
    - name: Taxonomy ID
      dtype: float64
    - name: Total Number of Polymer Residues per Assembly
      dtype: float64
    - name: Total Number of Polymer Instances (Chains) per Assembly
      dtype: float64
    - name: Oligomeric Count
      dtype: float64
    - name: Assembly ID
      dtype: float64
    - name: Oligomeric State
      dtype: string
    - name: Stoichiometry
      dtype: string
    - name: Ligand ID
      dtype: string
    - name: Ligand Formula
      dtype: string
    - name: Ligand MW
      dtype: float64
    - name: Ligand Name
      dtype: string
    - name: InChI
      dtype: string
    - name: Ligand SMILES
      dtype: string
    - name: Ligand of Interest
      dtype: string
    - name: __index_level_0__
      dtype: int64
  splits:
    - name: train
      num_bytes: 9342528
      num_examples: 8282
    - name: validation
      num_bytes: 1994645
      num_examples: 1775
    - name: test
      num_bytes: 1998895
      num_examples: 1775
  download_size: 6026997
  dataset_size: 13336068
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*

---

Data Description

This dataset contains information about 11,832 human protein structures derived from the RCSB Protein Data Bank (PDB). It includes data on each structure's experimental method, crystallization conditions, molecular composition, and structural metrics.

Data Version and Change Log

Last Version: 1.0
Last Updated: May 18th, 2025

Data Steward and Owner

Data Steward: Satarupa Deb
Data Owner: RCSB Protein Data Bank

Tables

The dataset consists of three splits: train, test, and validation, stored in the data folder

Data:

Description: This table stores all information about different columns in the dataset.
Rows (total: 11832) correspond to individual protein structure entries, typically registered in the Protein Data Bank (PDB). Each row contains structural, chemical, and biological metadata about a single protein or protein-ligand complex.
Columns (total: 42) correspond to different observables or attributes of the protein structures

Column Name	Data Type	Column Description
Entry ID	string	A unique identifier assigned to each protein structure in the Protein Data Bank.
Experimental Method	string	The technique used to determine the protein structure (e.g., X-ray diffraction, NMR).
Matthews Coefficient	float64	A value that estimates the volume of a crystal occupied by the protein versus the solvent.
Percent Solvent Content	float64	The percentage of solvent (usually water) present in the crystal structure.
Crystallization Method	string	The technique used to grow the crystals used for structure determination (e.g., vapor diffusion).
pH	float64	The acidity or alkalinity level at which the crystallization was performed.
Crystal Growth Procedure	string	The detailed steps followed to grow the crystals used for structure determination.
Temp (K)	float64	The temperature, measured in Kelvin, at which the crystallization experiment was conducted.
Deposition Date	string	The date when the protein structure was submitted to the Protein Data Bank.
Release Date	string	The date when the structure became publicly available in the PDB.
Number of Non-Hydrogen Atoms per Deposited Model	float64	The count of non-hydrogen atoms present in the deposited protein structure.
Total Number of Polymer Instances (Chains)	float64	The number of unique chains or instances of polymers present in the structure.
Total Number of Polymer Residues per Deposited Model	float64	The total number of amino acid residues present in the protein model.
Number of Water Molecules per Deposited Model	float64	The total number of water molecules present in the structure.
Disulfide Bond Count per Deposited Model	float64	The number of disulfide bonds (covalent bonds between cysteine residues) in the protein structure.
Molecular Weight per Deposited Model	float64	The molecular weight of the entire protein structure, including all chains.
Number of Distinct Protein Entities	float64	The number of unique protein entities present in the structure.
Refinement Resolution (Å)	string	The resolution of the X-ray diffraction experiment, indicating the level of detail (in Ångströms).
Structure Determination Methodology	string	The overall approach or methodology used to solve the protein structure.
Average B Factor	float64	A measure of atomic displacement (temperature factor), indicating the flexibility of atoms in the structure.
R Free	float64	A cross-validation metric that assesses the quality of the crystallographic model against experimental data.
R Work	float64	A measure of how well the model fits the experimental X-ray diffraction data during refinement.
Structure Title	string	The name or title assigned to the protein structure by the researchers.
Sequence	string	The linear sequence of amino acids in the protein.
Entity Polymer Type	string	The type of polymer present (typically a polypeptide or protein).
Polymer Entity Sequence Length	float64	The length (number of residues) of the polymer or protein sequence.
Entity Macromolecule Type	string	The classification of the macromolecule (e.g., protein, nucleic acid).
Total Number of polymer Entity Instances (Chains) per Entity	float64	The number of chains for each polymer entity.
Molecular Weight (Entity)	float64	The molecular weight of each individual polymer entity (e.g., a protein or nucleic acid).
Macromolecule Name	string	The name of the macromolecule or protein.
EC Number	string	The Enzyme Commission (EC) number, a numerical classification for enzymes based on their function.
EC Provenance Source	string	The source of the enzyme classification information.
Source Organism	string	The organism from which the protein or macromolecule was derived (e.g., Homo sapiens).
Taxonomy ID	float64	The taxonomy identifier that links to the classification of the source organism.
Total Number of Polymer Residues per Assembly	float64	The total number of residues in the entire protein assembly (multiple chains combined).
Total Number of Polymer Instances (Chains) per Assembly	float64	The total number of polymer chains in the entire assembly.
Oligomeric Count	float64	The number of subunits in the protein's oligomeric state (e.g., monomer, dimer, trimer).
Assembly ID	float64	A unique identifier for the specific assembly in the PDB.
Oligomeric State	string	The oligomeric form of the protein (e.g., monomer, dimer, tetramer), indicating how many subunits are involved in forming the functional unit.
Stoichiometry	string	The specific ratio of components (subunits or molecules) in a protein assembly.
Ligand ID	string	A unique identifier for any ligand (small molecule) bound to the protein in the structure.
Ligand Formula	string	The chemical formula of the ligand.
Ligand MW	float64	The molecular weight of the bound ligand.
Ligand Name	string	The name of the ligand.
InChI	string	The International Chemical Identifier for the ligand, providing a textual representation of its chemical structure.
Ligand SMILES	string	A string representation of the ligand’s structure using the SMILES notation.
Ligand of Interest	string	Specifies whether a particular ligand is of special interest in the structure (e.g., a drug or cofactor).
index_level_0	int64	auto-generated, Not biologically meaningful.

---