Deep CANALs: A Deep Learning Approach to Refining the Canalization Theory of Psychopathology

Hello and welcome everyone. It's Active Inference Guest Stream number 48.1 on July 19th, 2023. We're here with Arthur Giuliani and Adam Saffron, also joined by Michael Edward Johnson in discussion. We're going to have a presentation and then a conversation about Deep Canals, a Deep Learning Approach to Refining the Canalization Theory of Psychopathology. So thank you all for joining and looking forward to this presentation and discussion. Thanks, Daniel. I'm very happy to be here with Adam and Mike to talk about this work and to discuss more broadly this general direction that's emerged over the past few years to try to utilize some of the concepts from the active inference world and computational modeling more generally to understand what's going on both in psychopathology and in psychedelic therapy. Yeah, so I'll give a brief presentation and we'll move into a kind of more general discussion of this work. It'll start with some kind of general considerations around psychedelic therapy, some of the neuroscience of psychedelics, and then move kind of more specifically into the model. So there is this class of drugs, the serotonergic psychedelic drugs, which include psilocybin, DMT, LSD, and others. And the past decade has seen a resurgence of interest in psychedelic therapy, psychedelic assisted psychotherapy, where one or multiple of these drugs is used in conjunction with a kind of psychotherapeutic protocol. And there's been a lot of investment of money and time and clinical resources into utilizing these drugs for treatments, depression, anxiety disorders, substance use disorders, and others. And there seems to be quite a lot of positive signals suggesting that these kinds of therapeutic interventions can actually be quite effective. So there's a number of studies suggesting that psilocybin assisted psychotherapy can be helpful for depression, anxiety, substance use disorder, and others. And as this work advances, there are more and more larger clinical trials comparing it, for example, to the standard treatments with SSRIs, with multiple studies suggesting that psilocybin assisted psychotherapy is at least as effective as the Lexapro kind of then care. All of this seems to suggest, and the timelines are pointing at this idea that psilocybin treatment for major depressive disorder may be only a couple of years away of becoming approved as an actual treatment in the United States, which means that it's quite important for us to understand what's going on mechanistically in the brain so that these kinds of treatments can be best optimized, best determined how to be used, for which kinds of mental issues people might be having, and under what kinds of treatment protocols we might expect the most success. I think this will also ultimately hopefully lead to the development of even better drugs, even better treatment protocols, of the treatment protocols, and much more personalized care. So some of the basics of how these drugs work on the brain, for those who are unfamiliar, right, so these are the serotonergic psychedelics. They are serotonergic because they act as agonists for a number of serotonin receptors in the brain and the central nervous system more broadly. So it's thought that most specifically these drugs like psilocybin work through agonizing the serotonin 5-HT2A receptor, and this has been demonstrated multiple times by using blockers, antagonists of the 5-HT2A receptor, which seem to largely remove the effect of psilocybin or other psychedelic drugs. And what this 2-HT2A receptor does when it's agonized is it excites the neuro, making it more likely to fire. And this is kind of essential to some of the downstream effects. So given that it seems that this is this 2-HT2A receptor that psychedelics are mainly targeting, the question then becomes, well, where in the brain are these 2-HT2A receptors most widely expressed? And research has focused on, I would say, three kind of key areas over the past decade or so. The…