AI-based Mapping of Complex Cannabis Extracts in Pain Pathways

点击翻译按钮获取中文摘要

ABSTRACT There is tremendous interest in cannabis (as smoked marijuana or CBD- or THC-dominant extracts) as a therapeutic modality for a variety of health indications (including chronic pain). Given the complexity of cannabis, however, we have little insight into its mechanisms of action in complementary and integrative health approaches. Specifically, there is a prevailing notion that the 100+ cannabinoids and the various terpenoids/flavonoids that comprise cannabis act in concert to create an “Entourage Effect”. A comprehensive analysis is required to better understand the potential of cannabis agents as complementary medicines. We herein propose a novel artificial intelligence-driven approach to address this gap in our knowledge. Not surprisingly, a natural product (e.g., cannabis) that is active in an organism typically works because it acts like endogenous ligands or those known to the organism. We hypothesize that deconstructing ligand structures into specific fragments will allow us to identify targets that bind endogenous tergets containing such fragments. Moreover, we believe that disparate compounds acting in concert will maximally engage selective pathways. We have developed an artificial intelligence (AI)-driven platform, DRIFT (drug-target identification based on chemical similarity), to map ligand compounds (cannabinoids and terpenoids) to molecular targets. Thus, we can illuminate involved cellular pathways, and predict physiological response. We will use DRIFT to profile compounds in a number of different cannabis extracts (e.g., high in CBD, CBG, or THC) with varying analgesic properties to identify therapeutic combinations and their relevant targets. We will undertake three specific aims. In Specific Aim 1, use DRIFT for massive mapping of cannabis constituents to corresponding target proteins. Then, in Specific Aim 2, we will extend the DRIFT platform by evaluating binding affinities between metabolites and their proteins using a new neural network paradigm (NeuralDock). We will use text mining techniques to mine compound-protein relationships from PubMed. Finally, in Specific Aim 3, we will experimentally validate the outputs of the DRIFT platform to predict mechanisms of cannabis on pain. We will test the AI-based results using traditional pharmacology tools and a variety of preclinical animal models of pain. These same models will then be employed to test mechanisms through the complementary use of agonists, antagonists, inhibitors and (where appropriate) gene knockouts to validate mechanisms. When these studies are successful, we will have validated DRIFT as a new and valuable AI tool for studying natural products. Moreover, we will provide important insights into the growing use of cannabis in complementary and integrative health.