Quantitative binding site model generation: compass applied to multiple chemotypes targeting the 5-HT1A receptor

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Drug discovery

Background:

• Structure-activity relationships (SAR) are crucial for drug discovery.
• Predictive modeling of receptor binding aids in identifying potent drug candidates.
• Understanding inter-chemotype relationships can improve model generalizability.

Purpose of the Study:

• To enhance the Compass algorithm for automated deduction of inter-chemotype relationships.
• To generate predictive quantitative models of receptor binding using only structure-activity data.
• To apply and validate the enhanced algorithm on 5-HT1A receptor binding data.

Main Methods:

• Developed enhancements to the Compass algorithm.
• Utilized structure-activity data for model construction.
• Applied the model to predict affinities and bioactive conformations of new compounds.
• Assessed prediction accuracy using mean error.

Main Results:

• Successfully predicted affinities and bioactive conformations for 35 new compounds with high accuracy (0.5 log units mean error).
• The model demonstrated effectiveness on compounds with novel scaffolds and functional groups.
• Identified interpretable hypotheses for receptor binding determinants and chemotype geometric relationships.

Conclusions:

• The enhanced Compass algorithm accurately predicts receptor binding for diverse chemical structures.
• The method provides insights into binding determinants and chemotype relationships.
• This approach facilitates efficient drug discovery by leveraging existing SAR data.