molssiai-hub/pubchemqc-b3lyp

PubChemQC-B3LYP/6-31G\*//PM6 Dataset

Table of Contents

• —PubChemQC-B3LYP/6-31G\*//PM6 Dataset
• —Table of Contents
• —Dataset Description
• —Dataset Summary
• —Dataset Structure
• —Data Instances
• —Data Fields
• —Data Splits and Configurations
• —How to Use the Dataset
• —Prerequisites
• —Accessing the Data
• —Dataset Creation
• —Curation Rationale
• —Source Data
• —Initial Data Collection and Normalization
• —Personal and Sensitive Information
• —Considerations for Using the Data
• —Social Impact of Dataset
• —Additional Information
• —Dataset Curators
• —Licensing Information
• —Citation Information
• —Contributions

Dataset Description

• —Homepage: https://nakatamaho.riken.jp/pubchemqc.riken.jp/b3lyppm6datasets.html
• —Repository: https://nakatamaho.riken.jp/pubchemqc.riken.jp/b3lyppm6datasets.html
• —Paper: https://doi.org/10.1021/acs.jcim.3c00899
• —Point of Contact: Maho Nakata
• —Point of Contact: Mohammad Mostafanejad
• —Point of Contact: MolSSI-AI Hub

• —Important Note (Jul. 22, 2025): According to Hugging Face, the usage of data load scripts and the corresponding trust_remote_code parameter in the load_dataset() is deprecated due to security risks. Users should temporarily use datasets<4.0.0 to be able to use our datasets. We are working on this issue to reformat our datasets and make them compliant to the new requirements.

Dataset Summary

The presented *PubChemQC B3LYP/6-31G\//PM6* data set is composed of the electronic properties of 85,938,443 molecules, encompassing a broad spectrum of molecules from essential compounds to biomolecules with a molecular weight up to 1000. These molecules account for 94.0% of the original PubChem Compound catalog as of August 29, 2016. The electronic properties, including orbitals, orbital energies, total energies, dipole moments, and other pertinent properties, were computed by using the B3LYP/6-31G\ and PM6 methods.

Dataset Structure

Data Instances

An example of a data instance is as follows:

{'cid': 1,
  'state': 'S0',
  'pubchem-inchi': 'InChI=1S/C9H17NO4/c1-7(11)14-8(5-9(12)13)6-10(2,3)4/h8H,5-6H2,1-4H3',
  'pubchem-charge': 0,
  'pubchem-version': '20160829',
  'name': '000000001.B3LYP@PM6.S0',
  'coordinates': [
    4.543149670829423,
   -2.8411897941733857,
   -1.6418598810432616,
   ...,
   4.345629685137421
   ],
  'atomic-numbers': [
   6,
   6,
   8,
   ...,
   1
   ],
  'atom-count': 31,
  'heavy-atom-count': 14,
  'core-electrons': [
   0,
   0,
   0,
   ...,
   0
   ],
  'bond-order': [
   1,
   1,
   1,
   ...,
   1
   ],
  'connection-indices': [
   15,
   1,
   17,
   ...,
   30],
  'formula': 'C9H17NO4',
  'version': '1.0',
  'obabel-inchi': 'InChI=1S/C9H17NO4/c1-7(11)14-8(5-9(12)13)6-10(2,3)4/h8H,5-6H2,1-4H3/t8-/m0/s1',
  'pm6-obabel-canonical-smiles': '[O]C(=O)C[C@@H](C[N](C)(C)C)OC(=O)C',
  'charge': 0,
  'energy-beta-gap': 4.34837933099,
  'energy-beta-homo': -4.60960862747,
  'energy-beta-lumo': -0.2612292964799998,
  'energy-alpha-gap': 4.34837933099,
  'energy-alpha-homo': -4.60960862747,
  'energy-alpha-lumo': -0.2612292964799998,
  'total-energy': -19286.973573267132,
  'homos': [54],
  'orbital-energies': [
    [
      -522.303488065215,
      -521.209590386205,
      -518.042185166385,
      ...,
      127.37105114203999
    ]
  ],
  'mo-count': 244,
  'basis-count': 244,
  'multiplicity': 1,
  'molecular-mass': 203.23557999999983,
  'number-of-atoms': 31,
  'lowdin-partial-charges': [
   -0.459759,
   0.210106,
   -0.286001,
   ...,
   0.169819
   ],
  'mulliken-partial-charges': [
   -0.542286,
   0.622923,
   -0.486172,
   ...,
   0.185706
  ],
  'dipole-moment': 11.419443262233626,
  'pubchem-multiplicity': 1,
  'pubchem-obabel-canonical-smiles': '[O-]C(=O)CC(C[N+](C)(C)C)OC(=O)C',
  'pubchem-isomeric-smiles': 'CC(=O)OC(CC(=O)[O-])C[N+](C)(C)C',
  'pubchem-molecular-weight': 203.23558,
  'pubchem-molecular-formula': 'C9H17NO4'
}

Data Fields

Data Splits and Configurations

The dataset has only one train split. The PubChemQC B3LYP/6-31G\*//PM6 dataset has six configurations/subsets:

• —b3lyp_pm6 (default)
• —b3lyp_pm6_chon300nosalt
• —b3lyp_pm6_chon500nosalt
• —b3lyp_pm6_chnopsfcl300nosalt
• —b3lyp_pm6_chnopsfcl500nosalt
• —b3lyp_pm6_chnopsfclnakmgca500

How to Use the Dataset

Prerequisites

We recommend isolating your work in a virtualenv or conda environment. You can create a new conda environment, pubchemqc,

  conda create -n pubchemqc python=3.12

and activate it using the following command

  conda activate pubchemqc

Once the conda environment is activated, you can install the dependencies in it as shown below

  pip install huggingface_hub ijson

Accessing the Data

Once the required packages are installed, you can run the following code to access the data

  # import the modules
  from datasets import load_dataset

  # load the dataset with streaming
  hub_ds = load_dataset(path="molssiai-hub/pubchemqc-b3lyp",
                        name="b3lyp_pm6",
                        split="train",
                        streaming=True,
                        cache_dir="./tmp",
                        trust_remote_code=True)

  # fetch a batch of 32 samples from the dataset
  ds = list(hub_ds.take(32))

The argument name by default is set to b3lyp_pm6 which refers to the entire dataset. Other configurations (subsets), listed in Sec. Data Splits and Configurations, can also be selected.

The split must be set to train as it is the only split in our dataset. We recommend using streaming=True to avoid downloading the entire dataset on disk. The cache_dir allows us to store the Hugging Face datasets' and models' artifacts in a non-default directory (by default, it is set to ~/.cache/huggingface). As we are using a custom load script, the trust_remote_code argument should also be set to True.

Dataset Creation

Curation Rationale

The present version of PubChemQC B3LYP/6-31G\*//PM6 dataset has been extracted from its original Postgresql database, transformed into a dictionary and stored in the .json format.

Source Data

The link to the original PubChemQC B3LYP/6-31G\*//PM6 dataset repository can be found here

Initial Data Collection and Normalization

Other than the changes detailed in Sec. Curation Rationale, no data modification has been performed on the PubChemQC B3LYP/6-31G\*//PM6 dataset.

Personal and Sensitive Information

The PubChemQC B3LYP/6-31G\*//PM6 dataset does not involve any personal or sensitive information.

Considerations for Using the Data

Social Impact of Dataset

The PubChemQC B3LYP/6-31G\*//PM6 dataset paves the way for applications in drug discovery and materials science, among others.

Additional Information

Dataset Curators

• —Maho Nakata, RIKEN Cluster for Pioneering Research, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan
• —Toshiyuki Maeda, Software Technology and Artificial Intelligence Research Laboratory, Chiba Institute of Technology, 2-17-1 Tsudanuma, Narashino, Chiba 275-0016, Japan

Licensing Information

Creative Commons Attribution 4.0 International License

Citation Information

@article{Nakata:2023:5734,
   author = {Maho Nakata and Toshiyuki Maeda},
   doi = {10.1021/ACS.JCIM.3C00899},
   issn = {1549960X},
   issue = {18},
   journal = {Journal of Chemical Information and Modeling},
   pages = {5734-5754},
   publisher = {American Chemical Society},
   title = {{PubChemQC B3LYP/6-31G*//PM6 Data Set: The Electronic Structures of 86 Million Molecules
   Using B3LYP/6-31G* Calculations}},
   volume = {63},
   url = {https://pubs.acs.org/doi/abs/10.1021/acs.jcim.3c00899},
   year = {2023},
}

Contributions

• —Mohammad Mostafanejad, The Molecular Sciences Software Institute (MolSSI)