Halogen-Pi Interaction Energies for Protein Residues from Quantum Mechanical Calculations

Neural network models for predicting halogen-pi interaction energies across multiple biologically relevant aromatic systems. The dataset contains over 18 million interaction geometries for halobenzene-aromatic complexes, evaluated at the MP2/TZVPP level of theory. Models developed by Marc U. Engelhardt achieve high accuracy (R² > 0.98, RMSE < 0.5 kJ/mol) for targeted interaction domains.

Use Cases

• Predict halogen-pi interaction energies for tyrosine, histidine, and tryptophan residues based on compact geometric descriptors.
• Train or benchmark machine learning models for quantum-mechanical property prediction using systematically generated geometry data.
• Support drug design workflows by enabling near-quantum-mechanical accuracy at low computational cost for specific interaction motifs.

Strengths

• Contains over 18 million systematically generated interaction geometries.
• Models achieve high reported accuracy with R² > 0.98 and RMSE < 0.5 kJ/mol within the targeted domain.
• Performance validated on independent, randomly generated geometry and PDB-derived test sets.

Limitations

• Model performance is limited for geometric arrangements outside the training distribution, such as π···π or C–H···π motifs.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

figshare

Collection Method

Quantum mechanical calculations (MP2/TZVPP level of theory) on systematically generated molecular complexes.

Time Range

null

Freshness

Last updated 2026-04-14 13:05:27; freshness should be verified.

Geography

null