Search

Treatment for major depressive disorder (depression) often has partial efficacy and a large portion of patients are treatment resistant. Recent studies implicate reduced somatostatin (SST) interneuron inhibition in depression, and new pharmacology boosting this inhibition via positive allosteric modulators of α5-GABAA receptors (α5-PAM) offers a promising effective treatment. However, testing the effect of α5-PAM on human brain activity is limited, meriting the use of detailed simulations. We utilized our previous detailed computational models of human depression microcircuits with reduced SST interneuron inhibition and α5-PAM effects, to simulate EEG of individual microcircuits across depression severity and α5-PAM doses. We developed machine learning models that predicted optimal dose from EEG with high accuracy and recovered microcircuit activity and EEG. This study provides dose prediction models for α5-PAM administration based on EEG biomarkers of depression severity. Given limitations in doing the above in the living human brain, the results and tools we developed will facilitate translation of α5-PAM treatment to clinical use. Author summary: Inhibition from somatostatin-expressing (SST) interneurons is reduced in depression, and new pharmacology (α5-PAM) can selectively recover this inhibition to healthy level. However, α5-PAM has not been tested in humans, meriting the use of detailed simulations for dose prediction from EEG biomarkers of depression severity. We simulated a set of individual human neuronal microcircuits varying in severity of SST interneuron inhibition loss and across a range of α5-PAM doses. We then trained machine learning models to predict optimal dose from simulated EEG features, which recovered EEG profile and microcircuit activity to the healthy levels. Our study thus provides new tools for dose prediction and monitoring of α5-PAM efficacy using non-invasive EEG biomarkers of depression severity. [ABSTRACT FROM AUTHOR]