A multiple classifier system identifies novel cannabinoid CB2 receptor ligands.

Area of Science:

• Computational Chemistry
• Medicinal Chemistry
• Machine Learning

Background:

• Drug discovery is costly and time-consuming, necessitating efficient screening methods.
• Existing drugs may be insufficient or have adverse effects, driving the search for new therapeutics.
• Virtual screening, particularly using machine learning, offers a cost-effective approach to identify potential drug candidates.

Purpose of the Study:

• To apply a Multiple Classifier System (MCS) for virtual screening, termed D2-MCS, utilizing circular fingerprints.
• To screen a large chemical library for novel cannabinoid CB2 ligands.
• To experimentally validate the identified potential active molecules.

Main Methods:

• Development and application of the D2-MCS algorithm for virtual screening.
• Utilizing circular fingerprints and machine learning classifiers.
• Screening of the Enamine HTS collection (1,834,362 compounds) against cannabinoid CB2 targets.

Main Results:

• D2-MCS identified 48,232 potential active molecules from the screened library.
• Experimental validation confirmed six highly active and five medium active cannabinoid CB2 ligands.
• The D2-MCS method achieved a 29% hit rate for highly active compounds and a 52% overall hit rate.

Conclusions:

• D2-MCS is an effective machine learning approach for virtual screening in drug discovery.
• The study successfully identified novel, experimentally validated cannabinoid CB2 ligands.
• This method demonstrates significant potential for accelerating the identification of promising drug candidates.