Transition-based reweighting and neural rational inference analysis of effects of new psychoactive substances on cannabinoid receptors

New psychoactive substances (NPS) targeting cannabinoid receptor 1 (CB(1)) pose a significant threat to society as recreational abusive drugs that can avoid detection and have higher physiological side effects. These physiological side effects of NPS are shown to be linked to the higher β-arrestin signaling. We hypothesize that the difference in conformational dynamics of the NPxxY motif causes the distinct downstream signaling of NPS contrary to the classical cannabinoids. To compare the dynamic effects of the NPS and classical cannabinoid binding on the NPxxY conformational ensemble, we simulate (un)binding process of NPS MDMB-Fubinaca and classical cannabinoid HU-210 from CB(1) using unbiased and biased molecular dynamics simulations. The transition-based reweighing method (TRAM) is used to combine multi-ensemble simulations for the estimation of transition rates and underlying thermodynamics of (un)binding processes of ligands with nanomolar affinities, where it is more expensive to obtain local reversible sampling. Our analyses suggest that the ligands unbind from the receptors using the same pathway but by a different mechanism. Further analyses reveal higher conformational fluctuation in the NPxxY motif for NPS bound CB(1), supporting our hypothesis. The observation is further validated using a Variational autoencoder (VAE) based on Neural rational inference, which shows higher dynamic allostery-based interactions between the binding pocket residues and NPxxY for NPS bound CB(1). Hence, in this work, MD simulation, data-driven statistical methods, and deep learning point out the significant differences in (un)binding and downstream signaling of NPS and classical cannabinoids.