AIMD-Chig: 2 Million Ab Initio Molecular Dynamics Conformations of Chignolin

Chig-AIMD is a molecular dynamics dataset containing 2 million conformations of the 166-atom protein Chignolin. The data was generated using density functional theory (DFT) with the M062X/6-31G* method and a Berendsen thermostat at 340K, consuming 7,763,146 CPU hours. It was created by Tong Wang and includes coordinates, energies, and forces for each conformation, sampling folded, unfolded, and metastable states.

Use Cases

• Training machine learning force fields based on ab initio energy and force labels.
• Benchmarking molecular dynamics sampling algorithms against a high-quality reference dataset.
• Studying protein folding and unfolding pathways of Chignolin based on conformational coordinates.
• Validating computational chemistry methods using a dataset of 10,000 initial conformations covering the full conformational space.

Strengths

• Contains 2 million conformations generated at the computationally expensive DFT level.
• Sampling includes 10,000 initial conformations covering folded, unfolded, and metastable states.
• Provides atomic coordinates, energies, and forces for each conformation, which are essential for ML model training.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• Row count for the full dataset is unspecified, though the total number of conformations is stated.

Provenance

Source

Tong Wang via paperswithcode.

Collection Method

Ab initio molecular dynamics simulations using the M062X/6-31G* method and a Berendsen thermostat at 340K.