Scaffold-Hopped Compound Identification by Ligand-Based Approaches with a Prospective Affinity Test

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Scaffold-hopped (SH) compounds offer novel chemical diversity.
• Identifying SH compounds is vital for drug discovery.
• Current molecular representations for SH identification lack clarity.

Purpose of the Study:

• To evaluate and compare molecular representations for SH compound identification.
• To validate retrospective and prospective approaches for SH discovery.
• To assess the performance of different computational methods.

Main Methods:

• Retrospective validation using controlled datasets.
• Comparison of four 2D/3D molecular representations with two screening algorithms.
• Prospective screening of the Namiki database for ABL1 inhibitors.
• Surface plasmon resonance and competitive binding assays.

Main Results:

• Support Vector Machine with Extended Connectivity Fingerprint (SVM-ECFP4) and Rapid Overlay of Chemical Structures (SVM-ROCS) showed high performance.
• SVM-ROCS identified compounds lacking substructures of active training compounds.
• SVM-ECFP4 identified mostly recombinant compounds.
• Prospective screening yielded five potential active compounds, but none confirmed in competitive assays.

Conclusions:

• SVM-ECFP4 and SVM-ROCS are promising for early scaffold-hopping identification.
• Further refinement of methods is needed for successful prospective SH compound discovery.
• The study highlights challenges in translating computational predictions to validated hits.