Coarse-Grained Models for Automated Fragmentation and Parametrization of Molecular Databases

Area of Science:

• Computational chemistry
• Molecular modeling
• Chemical engineering

Background:

• Accurate prediction of molecular properties is crucial for drug discovery and materials science.
• Traditional methods for molecular simulation are computationally expensive for large datasets.
• Coarse-graining offers a computationally efficient alternative for molecular modeling.

Purpose of the Study:

• To develop and validate novel algorithms for calibrating coarse-grained interaction potentials.
• To enable efficient screening of large molecular datasets using a top-down approach.
• To improve the accuracy and speed of molecular property predictions.

Main Methods:

• Automated molecular fragmentation into coarse-grained units.
• Coarse-Grained Reference Interaction Site Model-Hypernetted Chain (CG RISM-HNC) as a proxy for dissipative particle dynamics (DPD).
• Top-down coarse-grained model based on COSMO-RS activity coefficient theories.
• DPD thermodynamics integration for log P_OW calculations.

Main Results:

• Fragment distributions were found to follow Zipf and Heaps scaling laws.
• Gibbs energy of mixing calculations achieved accuracy within tenths of a kilocalorie per mole.
• Log P_OW for 4627 compounds was calculated with an average error of 0.84 log units using DPD.
• CG RISM-HNC calculations took seconds, while DPD thermodynamic integration took minutes.

Conclusions:

• The developed coarse-graining algorithms provide an accurate and efficient method for molecular property prediction.
• This approach facilitates the screening of large chemical libraries.
• The methods offer a significant speedup compared to traditional all-atom simulations.