A polygenic-informed approach to a predictive EEG signature empowers antidepressant treatment prediction: A proof-of-concept study

The treatment of major depressive disorder (MDD) is hampered by low chances of treatment response in each treatment step, which is partly due to a lack of firmly established outcomepredictive biomarkers. Here, we hypothesize that polygenic-informed EEG signatures may help predict antidepressant treatment response. Using a polygenic-informed electroencephalography (EEG) data-driven, data-reduction approach, we identify a brain network in a large cohort (N = 1,123), and discover it is sex-specifically (male patients, N = 617) associated with polygenic risk score (PRS) of antidepressant response. Subsequently, we demonstrate in three independent datasets the utility of the network in predicting response to antidepressant medication (male, N = 232) as well as repetitive transcranial magnetic stimulation (rTMS) and concurrent psychotherapy (male, N = 95). This network significantly improves a treatment response prediction model with age and baseline severity data (area under the curve, AUC = 0.623 for medica-ton; AUC = 0.719 for rTMS). A predictive model for MDD patients, aimed at increasing the like-lihood of being a responder to antidepressants or rTMS and concurrent psychotherapy based on only this network, yields a positive predictive value (PPV) of 69% for medication and 77% for rTMS. Finally, blinded out-of-sample validation of the network as predictor for psychother-apy response in another independent dataset (male, N = 50) results in a within-subsample re-sponse rate of 50% (improvement of 56%). Overall, the findings provide a first proof-of-concept of a combined genetic and neurophysiological approach in the search for clinically-relevant biomarkers in psychiatric disorders, and should encourage researchers to incorporate genetic information, such as PRS, in their search for clinically relevant neuroimaging biomarkers.(c) 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ )