Leveraging AI to discover neuron-targeted anti-addiction drugs

Project Summary In this project, we will leverage two artificial intelligence (AI) tools and our company’s unique Historeceptomics approach to rapidly discover an impactful new drug candidate to treat opioid use disorder (OUD) relapse via inhibition of withdrawal-associated negative affect (hyperkatifeia). The premise is that, unlike traditional drug discovery, which is conceptualized starting from drug targets (e.g., mu opioid receptor) and proceeds to phenotypes (e.g., pain relief), we can select specific cells or tissues that control a specific in vivo phenotype, in this case hyperkatifeia, and work backward to identify a drug target specific for those cells, followed by the discovery of a drug-like compound that modulates that drug target. This would be a revolutionary new paradigm for OUD drug discovery, translating extensive neuroscience knowledge about neural circuits that elicit specific analgesic, addiction or withdrawal phenotypes directly into drugs, which has not been direct or efficient before. The critical technical advance that makes this paradigm possible is the capability to screen virtually for drug candidates targeted at any human gene product. In other words, this drug discovery process can start agnostic to the eventual target, with confidence that ligands for that target can rapidly be discovered once it is identified. This advance is made possible by our Historeceptomics technologies, which allow us to query a specific cell or tissue and identify, with statistical significance, gene products exclusive or maximally specific to that tissue. The advance is also made temporally feasible by two historic breakthroughs in AI: 1) AlphaFold2 from DeepMind/Google, which has predicted accurate 3D structures, suitable for virtual chemical library screening, for every human gene product for the first time in history; and 2) breakthrough AI binding affinity prediction for drug candidates to 3D structures developed by our partner Molsoft LLC, which uses a convolutional neural network. We thus aim to 1) adapt our current Historeceptomics Tissue Search software feature to single cell RNA seq data profiling of the human amygdala and identify gene products exclusive to Rspo2+ neurons, which control hyperkatifeia. 2) perform an ultra-large virtual library screen using our AI binding affinity prediction of the 1.5 billion Enamine Real database of drug-like chemical compounds for drug candidates that bind to the AlphaFold2-generated, manually optimized, 3D structure of the thus-identified gene product, 3) test validated hit compounds for their ability to reduce the negative affect of protracted opioid abstinence and relapse in vivo. A drug targeting reinstatement in people suffering from OUD is arguably more impactful than other anti-addiction approaches, because it could synergize with medication-assisted treatment (MAT), psychosocial therapies and community-based recovery supports by neurochemically reducing the impact of triggers conditioned by avoidance of withdrawal symptoms. We will leverage AI to discover such a drug within three years with modest investment, which would both advertise the paradigm of tissue/cell focused drug discovery and the power of AI to accelerate drug discovery for OUD/SUD.