Navigating Drug-Like Chemical Space of Anticancer Molecules Using Genetic Algorithms and Counterpropagation...

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Cheminformatics

Background:

• The Binding database contains numerous drug-like anticancer molecules.
• Effective classification of these molecules is crucial for drug discovery.
• Understanding structure-activity relationships (SAR) is key to identifying novel anticancer agents.

Purpose of the Study:

• To analyze a large set of drug-like anticancer molecules.
• To identify relevant molecular descriptors for classification.
• To develop virtual filters for efficient screening of new anticancer drug candidates.

Main Methods:

• Collected 6289 drug-like anticancer molecules from the Binding database.
• Encoded molecules using diverse descriptors representing physical and chemical properties.
• Employed genetic algorithms and counterpropagation artificial neural networks for descriptor selection and classification.

Main Results:

• Selected molecular descriptors define distinct chemical space regions for specific anticancer molecule classes.
• Developed classification rules based on selected descriptors.
• Demonstrated the utility of these rules for precise screening of large compound databases.

Conclusions:

• The identified SAR patterns and classification rules provide valuable insights into anticancer molecule properties.
• These rules act as virtual filters for mining compound databases.
• The approach facilitates the discovery of new anticancer drug candidates.