PREDICTING THE FUTURE DISEASE STATUS OF DEPRESSED PATIENTS FROM DNA METHYLATION PATTERNS IN BLOOD

Background Major Depressive Disorder (MDD) is a severe disorder with a lifetime prevalence of 15%. Recurrent or chronic MDD, which typically requires long-term treatment, occurs in the majority of patients. As previous efforts to predict recurrence/chronicity have met with limited success, novel avenues are required. In this study, we aimed to predict the future disease status of MDD patients from DNA methylation patterns in blood. Methods We assayed 28 million methylation sites in 581 MDD patients from the Netherlands Study of Depression and Anxiety (NESDA). A Machine Learning algorithm condensed all information into a single predictor labeled methylation risk score (MRS). The outcome was MDD disease status six years later. To evaluate the predictive power of our MRS, we obtained an unbiased estimate of the area under the curve (AUC) using k-fold cross validation. Results The AUC of the MRS was 0.724. We compared our MRS with predictions based on a set of five putative MDD biomarkers (e.g., assaying neurotrophic factors or inflammation), genome-wide genetic variant data, and 27 clinical, demographic or lifestyle variables (e.g., MDD symptom severity, childhood trauma, alcohol use). The MRS not only outperformed all these predictors but also seemed to capture their predictive power as the inclusion of any of these sets did not significantly increase the AUC of the MRS. Discussion The current study suggests a novel avenue for predicting future MDD status using DNA methylation patterns in blood. The predictive power of the MRS is comparable to the AUC of the Framingham Risk Score, one of the most widely used clinical tools to predict coronary heart disease. This AUC level can potentially support clinical decisions about treatment strategies by providing empirical information about the likelihood MDD is chronic or will recur in the future.