[Development of quantitative structure-activity relationships and computer-aided drug design]

Area of Science:

• Computational chemistry
• Cheminformatics
• Medicinal chemistry

Context:

• Quantitative structure-activity relationships (QSARs) are crucial for computer-aided drug design.
• Existing QSAR methods face challenges with non-congeneric datasets.
• Accurate prediction of chemical properties is vital for drug development.

Purpose:

• To develop and present the Fuzzy Adaptive Least-Squares (FALS) method for QSAR analysis.
• To apply FALS for non-congeneric QSAR modeling of carcinogenicity, mutagenicity, and pharmacokinetic properties.
• To introduce a heuristic search method for identifying active molecular conformers.

Summary:

• Fuzzy Adaptive Least-Squares (FALS) is a novel pattern recognition technique for generating QSAR models from structure-activity data.
• FALS utilizes fuzzy membership functions to define sample activity class belongingness.
• The method was successfully applied to predict chemical properties using log P values calculated by Moriguchi's method, demonstrating reliability and simplicity compared to other hydrophobicity descriptors.

Impact:

• FALS provides a robust framework for predictive QSAR modeling, particularly for diverse chemical datasets.
• The developed models aid in drug design by predicting key toxicological and pharmacokinetic properties.
• The proposed heuristic search method enhances the identification of optimal molecular conformations for drug discovery.