Evaluation of an inverse molecular design algorithm in a model binding site.

Area of Science:

• Computational chemistry
• Drug discovery
• Molecular modeling

Background:

• Virtual screening is a forward approach in drug discovery.
• Inverse approaches construct compounds computationally to optimize properties.
• A new inverse approach using dead-end elimination and A* algorithms was developed.

Purpose of the Study:

• To evaluate a new inverse computational drug design method.
• To assess the method's performance on a known biological target, the W191G mutant of cytochrome c peroxidase.
• To determine the method's ability to identify binders and nonbinders.

Main Methods:

• Utilized dead-end elimination and A* algorithms with a physical potential function.
• Applied the method to combinatorially constructed libraries of small-molecule ligands.
• Tested the approach on the W191G mutant of cytochrome c peroxidase, known for its charged binding pocket.

Main Results:

• The inverse approach demonstrated excellent performance in distinguishing binders from nonbinders.
• Scoring inaccuracies, particularly in solvation energy estimation, led to some false positives.
• The method showed excellent coverage of known binders, a notable achievement for inverse approaches.

Conclusions:

• The new inverse computational drug design method is effective for identifying high-affinity ligands.
• Addressing solvation energy inaccuracies could further improve the method's precision.
• The approach offers a promising alternative to traditional virtual screening in drug discovery.