PM7 Dataset

Dataset Summary

PM7 is a modern semi-empirical method with broad applications in chemical space including organic molecules, organometallics and solid-state crystalline compounds. The parameterization process involved adopting a diverse set of experimental and high-level ab initio reference data. The current dataset, curated by the Molecular Sciences Software Institute, attempts to provide a ready-to-use version of the PM7 dataset for the computational molecular science and machine learning community.

Dataset Structure

Data Instances

An example of a data instance is as follows:

{'title': 'MOPAC_54/PM7_reference',
 'name': '1,1,1-Trifluoroethane',
 'mopac-id': 54,
 'net-charge': 0,
 'spin-multiplicity': 1,
 'openbabel-canonical-smiles': 'CC(F)(F)F',
 'openbabel-version': '3.1.0',
 'rdkit-canonical-smiles': 'CC(F)(F)F',
 'rdkit-version': '2024.03.5',
 'oechem-canonical-smiles': 'CC(F)(F)F',
 'oechem-version': '20240905',
 'mopac-keywords': ['SYMMETRY', 'PM7'],
 'description': '1,1,1-Trifluoroethane\nI=13.8 IR=LLNBS82 D=2.32 H=-178.9 HR=NIST DR=MCC1963 S=66.9 CP=18.69',
 'atomic-indices': [1, 2, 3, 4, 5, 6, 7, 8],
 'atomic-numbers': [6, 6, 9, 9, 9, 1, 1, 1],
 'atomic-symbols': ['C', 'C', 'F', 'F', 'F', 'H', 'H', 'H'],
 'coordinates': [0.0,
  0.0,
  0.0,
  ...,
  -0.41029998660087585,
  -0.8860999941825867,
  0.5116000175476074],
 'bonds': [1, 7, 1, 1, 2, 1, 1, 6, 1, 1, 8, 1, 2, 5, 1, 2, 4, 1, 2, 3, 1],
 'enthalpy-of-formation': -748.5176,
 'enthalpy-of-formation-error': None,
 'enthalpy-of-formation-reference': None,
 'enthalpy-of-formation-units': 'kJ/mol',
 'ionization-energy': 13.8,
 'ionization-energy-error': None,
 'ionization-energy-reference': None,
 'ionization-energy-units': 'eV',
 'entropy': 279.9096,
 'entropy-units': 'J/mol/K',
 'constant-pressure-heat-capacity': 78.19896000000001,
 'constant-pressure-heat-capacity-units': 'J/mol/K',
 'diople-moment': 2.32,
 'dipole-moment-reference': 'MCC1963',
 'dipole-moment-units': 'D',
 'mopac-reference-energy': -731.158184,
 'mopac-reference-energy-units': 'kJ/mol'}

Data Fields

Field	Description
title	Title line in the MOPAC input/SDF file
name	Name of the molecule
mopac-id	MOPAC ID of the molecule
net-charge	Net charge of the molecule
spin-multiplicity	Spin multiplicity of the molecule
openbabel-canonical-smiles	Canonical SMILES string generated by OpenBabel
openbabel-version	Version of OpenBabel used to generate the canonical SMILES
rdkit-canonical-smiles	Canonical SMILES string generated by RDKit
rdkit-version	Version of RDKit used to generate the canonical SMILES
oechem-canonical-smiles	Canonical SMILES string generated by OpenEye OEChem
oechem-version	Version of OpenEye OEChem used to generate the canonical SMILES
mopac-keywords	Keywords used in the MOPAC input file
description	Description of the molecule
atomic-indices	Atomic indices of the atoms in the molecule
atomic-numbers	Atomic numbers of the atoms in the molecule
atomic-symbols	Atomic symbols of the atoms in the molecule
coordinates	Atomic coordinates of the atoms in the molecule
bonds	Bond information of the molecule
enthalpy-of-formation	Enthalpy of formation of the molecule
enthalpy-of-formation-error	Error in the enthalpy of formation
enthalpy-of-formation-reference	Reference for the enthalpy of formation
enthalpy-of-formation-units	Units of the enthalpy of formation
ionization-energy	Ionization energy of the molecule
ionization-energy-error	Error in the ionization energy
ionization-energy-reference	Reference for the ionization energy
ionization-energy-units	Units of the ionization energy
entropy	Entropy of the molecule
entropy-units	Units of the entropy
constant-pressure-heat-capacity	Constant pressure heat capacity of the molecule
constant-pressure-heat-capacity-units	Units of the constant pressure heat capacity
diople-moment	Dipole moment of the molecule
dipole-moment-reference	Reference for the dipole moment
dipole-moment-units	Units of the dipole moment
mopac-reference-energy	Reference energy of the molecule
mopac-reference-energy-units	Units of the reference energy

Data Splits and Configurations

The dataset has only one train split. Currently, the PM7 dataset has one configuration/subset:

• pm7 (default)

How to Use the Dataset

Prerequisites

We recommend isolating your work in a virtualenv or conda environment. You can create a new conda environment, pm7,

  conda create -n pm7 python=3.12

and activate it using the following command

  conda activate pm7

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

  pip install datasets huggingface_hub

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/pm7",
                        name="pm7",
                        split="train",
                        cache_dir="./tmp")
  # fetch a random batch of 32 samples from the dataset
  ds = hub_ds.shuffle(seed=1234).select(range(32))

The argument name by default is set to pm7. The split must be set to train as it is the only split in our dataset. 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). The shuffle method shuffles the dataset with a seed of 1234, and the select method selects the first 32 samples from the shuffled dataset.

Dataset Creation

Curation Rationale

The present version of PM7 dataset has been extracted from the pm7.sdf.gz file, transformed into a dictionary and stored in the .json format. The SDF file was generated using Simulation Environment for Atomistic and Molecular Modeling (SEAMM) software. The SEAMM flowchart for processing the PM7 dataset from the MOPAC input files can be found on Zenodo.

Source Data

The PM7 dataset (and the corresponding SDF file) was generated using the MOPAC input files hosted in MOPAC GitHub repository.

Initial Data Collection and Normalization

Other than the changes detailed in Sec. Curation Rationale, no data modification has been performed on the PM7 dataset.

Personal and Sensitive Information

The PM7 dataset does not involve any personal or sensitive information.

Considerations for Using the Data

Social Impact of Dataset

The PM7 dataset paves the way for applications in (in)organic chemistry, solid-state physics and materials science, among others.

Additional Information

Dataset Curators

• James J. P. Stewart, Stewart Computational Chemistry, 15210 Paddington Circle, Colorado Springs, CO, 80921, USA
• Paul Saxe, The Molecular Sciences Software Institute (MolSSI), Blacksburg, VA, 24060, USA
• Mohammad Mostafanejad, The Molecular Sciences Software Institute (MolSSI), Blacksburg, VA, 24060, USA

Licensing Information

Creative Commons Attribution 4.0 International License

Citation Information

@article{Stewart:2013:1,
   author = {Stewart, James J. P.},
   doi = {10.1007/S00894-012-1667-X},
   journal = {Journal of Molecular Modeling},
   pages = {1-32},
   publisher = {Springer},
   title = {Optimization of parameters for semiempirical methods VI: More modifications to the NDDO approximations and re-optimization of parameters},
   volume = {19},
   url = {https://link.springer.com/article/10.1007/s00894-012-1667-x},
   year = {2013}
}

Contributions

• James J. P. Stewart, Stewart Computational Chemistry
• Paul Saxe, The Molecular Sciences Software Institute (MolSSI)
• Mohammad Mostafanejad, The Molecular Sciences Software Institute (MolSSI)