Predicting the Efficacy of Psilocybin in Treating Mental Health and Addiction

Courtney Ledford, Portland State University


Machine Learning is used to predict the efficacy of psilocybin in treating mental health and addiction by using a random forest algorithm. Psilocybin is a chemical compound found in fungi. Following ingestion, it is converted to psilocin which acts as a serotonin agonist, producing altered states of consciousness and hallucinations. Research suggests the use of psilocybin as a potential treatment for mental health disorders and addiction. An analysis of thirty-one experiments from the Altered States Database was inputted into the model. Past meta-analysis performed on this experimental data focuses on healthy participants rather than participants diagnosed with a mental illness. The subjective survey accessed in this model is the 5-Dimensional Altered States of Consciousness (5D-ASC) questionnaire. The forest is composed of 1,000 decision trees randomly assigned data to produce a final leaf node prediction. Ten inputs were added from values determined by the 5D-ASC questionnaire, with the predictive output being the subjects' health. Based on the subjective questionnaires given to participants following the administration of psilocybin, the model predicted through classification, which of the experiments contained data of the clinically diagnosed patients. The model also predicted the variable of most importance as Oceanic Boundlessness, accounting for nearly 25% of the work in classifying an experiment as using clinical or healthy participants. The model is biased towards healthy participants, as only four of the experiments involved a diagnosis related to anxiety, alcohol dependency, or obsessive-compulsive disorder. A limitation of this model is the lack of data to support a split in training and testing of the dataset to validate the model. Current treatment options for these mental health issues depend on medications taken daily, versus psilocybin, which was administered in one to two doses, while sustaining a longer-lasting impact on the individual. A random forest is an appropriate algorithm to use for the small sample size and numerous variables. Machine learning allows for predictions to be made off the available experimental data. As more experimental research is conducted, increasing the amount of data available, machine learning will provide a means to analyze and predict the efficacy of psilocybin in treating mental health.