Automated Lead Optimization of MMP-12 Inhibitors Using a Genetic Algorithm

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Traditional lead optimization is time-consuming due to lengthy synthesis and testing cycles.
• Microfluidic platforms offer rapid, automated synthesis and screening, shifting bottlenecks to data analysis and design.
• Algorithm-directed optimization is needed to fully leverage rapid experimental cycles.

Purpose of the Study:

• To validate a genetic algorithm for autonomous, algorithm-directed lead optimization.
• To assess the algorithm's robustness with missing data and inactive compounds.
• To make comprehensive datasets publicly available for advancing optimization methods.

Main Methods:

• A genetic algorithm was developed with features for handling missing data and suggesting retests.
• The algorithm was validated retrospectively on a virtual compound library.
• A prospective experiment involved 10 cycles of algorithm-directed synthesis and screening against MMP-12.

Main Results:

• The algorithm successfully identified active compound regions, outperforming traditional methods in efficiency.
• The optimization process demonstrated robustness against synthesis failures (missing data) and inactive compounds.
• Algorithm-selected compounds showed a strong bias towards higher activity.

Conclusions:

• Genetic algorithms can effectively direct lead optimization, especially when coupled with rapid experimental platforms.
• The developed algorithm is reliable and efficient for autonomous drug discovery workflows.
• Publicly releasing the data will facilitate further research in computational drug design.