Analysis of multiple compound-protein interactions reveals novel bioactive molecules

Area of Science:

• Drug discovery and development
• Chemical genomics
• Systems biology

Background:

• Novel bioactive molecules are essential for understanding biological processes and driving drug development innovation.
• Chemical genomics generates large datasets of compound-protein interactions (CPIs).
• While CPI databases help identify new drug targets, their use for novel ligand discovery is underexplored.

Purpose of the Study:

• To demonstrate machine learning's capability in assessing drug polypharmacology.
• To showcase the identification of novel bioactive scaffold-hopping compounds using machine learning on multiple CPIs.
• To explore the potential of chemical genomics data in discovering novel ligands for pharmaceutically important targets.

Main Methods:

• Utilized a machine-learning technique applied to multiple compound-protein interaction (CPI) datasets.
• Applied the method to identify novel lead compounds for G-protein-coupled receptors and protein kinases.
• Compared the machine learning approach with existing computational ligand-screening methods.

Main Results:

• Successfully identified novel lead compounds for G-protein-coupled receptors and protein kinases.
• The identified compounds were not discoverable through conventional computational ligand-screening methods.
• Demonstrated that machine learning on multiple CPIs can effectively assess drug polypharmacology and identify novel ligands.
• Validated the utility of chemical genomics data for exploring chemical space.

Conclusions:

• Machine learning of multiple CPIs is a powerful tool for identifying novel bioactive scaffold-hopping compounds.
• Chemical genomics data, viewed through a systems biology lens, can significantly accelerate drug discovery.
• This approach offers a novel strategy for discovering ligands beyond existing computational methods.