Statistics-based model for basis set superposition error correction in large biomolecules.

Area of Science:

• Computational chemistry
• Biomolecular modeling
• Quantum chemistry

Background:

• Basis set superposition error (BSSE) is a significant artifact in quantum chemical calculations.
• Accurate estimation of BSSE is crucial for large biomolecular systems.
• Existing methods for BSSE correction are computationally expensive for large systems.

Purpose of the Study:

• To develop a computationally efficient and accurate statistical model for estimating BSSE in large biomolecular systems.
• To provide reliable estimates of BSSE and associated uncertainties.
• To apply the developed model to various biomolecular systems and quantum chemical levels.

Main Methods:

• A statistics-based linear model was developed to estimate BSSE.
• Molecular fragment interactions were classified and analyzed using a bimolecular proximity descriptor.
• Models were trained independently for different interaction classes, quantum methods, and basis sets.
• Predicted fragment BSSE values and uncertainties were propagated to the entire system.

Main Results:

• The model successfully estimated BSSE and uncertainties for protein-ligand complexes and proteins.
• The method was demonstrated at MP2/6-31G* and MP2/aug-cc-pVDZ levels of theory.
• Accurate BSSE predictions were achieved for native and decoy folds of the Pin1 WW domain.

Conclusions:

• The developed statistical model provides an efficient approach for BSSE estimation in large biomolecular systems.
• The method offers reliable predictions of BSSE and uncertainty, crucial for accurate computational studies.
• This approach facilitates high-throughput screening and analysis of large biomolecular interactions.