Predicting the Signaling Efficacy of Biased Agonists of the Cannabinoid Receptor 2

CB2 has become a popular drug target for cannabinoid derivatives and novel chemical entities (NCEs).  Despite the enthusiasm behind candidates for many indications (inflammation, pain, substance use disorders), few drugs have been approved. Many CB2 agonists (both synthetic and endogenous), demonstrate some level of biased signaling. While the outcomes of CB2 biased signaling are unclear, it is expected that they play a crucial role in candidate translation.


Our Process

We challenged our team to build a machine learning model to predict the signaling efficacy (Emax) of 17 known CB2 agonists for cAMP and β-arrestin-2 signaling5.


With only 2 experimental structures (PDBID: 8gus and 8gur), we built computational models to predict cAMP and β-arrestin-2 signaling within 17% and 15% of experimental values, respectively. 


With data from only 17 compounds, Biagon’s machine learning approach accurately predicted signaling for full agonists, partial agonists, inverse agonists, and biased agonists. 

.

Understanding the Mechanisms of CB2 Biased Agonism

The novelty of Biagon’s approach lies in the identification of GPCR conformations with distinct signaling outcomes. For CB2, our approach identified 20 conformations.  Below, we display how cAMP biased compounds (HU308 and JWH133) causes CB2 to populate conformations 1-7, while β-arrestin-2 biased compounds induce conformations 15, 18, and 20. Inverse agonist, SR144528 primarily induces conformation 13.

Understanding how slightly different ligands push GPCRs into different areas of conformational space is a difficult task. Fortunately, Biagon’s machine learning is transparent, allowing close analysis of conformations. These insights are critical to rationally designing next generation CB2 agonists that stabilize conformations that correspond to a desired signaling profile. 


New Opportunities for Structure-Based Drug Design

Virtual Screening

Lead Optimization

Screen compounds against ML-identified conformations

Accurate affinity predictions with alchemical approaches

Identify new chemical space with desired signaling profile

 

Screen optimizations for improvement in signaling and/or affinity

Rationally design improved compounds with learnings from ML-identified conformational space

 

Internal virtual screening efforts are underway. If interested in partnering with Biagon for this program, please reach out to our team here.

Scroll to Top