Screening biochemical changes induced by psychedelic ayahuasca in healthy subjects and patients with depression: a machine learning approach
machine-learning, ayahuasca, psychedelic, biomarkers, depression
Ayahuasca is an ancient psychedelic beverage from Amazonia that has recently demonstrated potential antidepressant effects. In this study, we aimed to investigate Ayahuasca's impact on biochemical markers in healthy subjects (control group - CG, n = 42) and patients with treatment-resistant depression (TRD, n = 28). We employed machine learning techniques, specifically random forest algorithms associated with recursive feature elimination, and 5-fold cross-validation. Firstly, our model selected c-reactive protein (CRP), awakening salivary cortisol, and total cholesterol as the most important features to differentiate between CG and TRD. Next, the model presented low performance in distinguishing Ayahuasca from placebo group on the CG, suggesting a minimal impact of Ayahuasca on these biomarkers in healthy subjects. On the other hand, to distinguish Ayahuasca's effects from placebo in TRD, the model suggests a modulation of glucose, CRP, lymphocytes, LDL cholesterol, red blood cells, BDNF, and AST. Finally, using recursive feature elimination and correlation analysis, BDNF and plasmatic cortisol emerged as significant predictors of Ayahuasca's therapeutic outcomes in TRD. Overall, our findings reinforce previous evidence that Ayahuasca is safe in healthy individuals, and associate biomarker shifts with its effect in patients with TRD.