Automatic molecular fragmentation by evolutionary optimisation.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Bioinformatics

Background:

• Molecular fragmentation reduces computational complexity in quantum chemistry.
• Current fragmentation methods lack automation and robust quality metrics.
• Automated approaches are needed for practical application in large molecular systems.

Purpose of the Study:

• To introduce Quick Fragmentation via Automated Genetic Search (QFRAGS), a novel automated algorithm for molecular fragmentation.
• To develop an algorithm that generates molecular fragments yielding low energy errors for Many-Body Expansions (MBEs).
• To enhance the automation and accuracy of fragmentation techniques for large-scale quantum chemistry.

Main Methods:

• Developed QFRAGS, an automated fragmentation algorithm employing a genetic optimization procedure.
• Benchmarked QFRAGS on protein systems (<500 and >500 atoms) using MBE2 and MBE3 calculations.
• Compared QFRAGS against manual fragmentation schemes and Fragment Molecular Orbital (FMO) techniques.

Main Results:

• QFRAGS achieved low Mean Absolute Energy Errors (MAEE) on protein systems: 20.6 kJ/mol (MBE2, <500 atoms), 2.2 kJ/mol (MBE3, <500 atoms), 181.5 kJ/mol (MBE2, >500 atoms), and 24.3 kJ/mol (MBE3, >500 atoms).
• On a lipoglycan/glycolipid dataset, QFRAGS yielded MAEs of 7.9 kJ/mol (MBE2) and 0.3 kJ/mol (MBE3).
• QFRAGS demonstrated comparable or superior performance to manual fragmentation and FMO methods in terms of MAEE.

Conclusions:

• QFRAGS effectively automates molecular fragmentation, generating high-quality fragments with minimal energy errors.
• The algorithm significantly improves the accuracy and computational feasibility of quantum chemistry calculations for large molecular systems.
• QFRAGS represents a substantial advancement in computational chemistry, making complex calculations more accessible.