Epigenetic prediction of major depressive disorder

Objective DNA methylation (DNAm) is associated with environmental risk factors for major depressive disorder (MDD) but has not yet been tested for its ability to discriminate individuals with MDD from unaffected individuals. Methods Using penalized regression based on genome-wide CpG methylation, we trained a DNAm risk score of MDD (DNAm-RS) in 1,223 cases and 1,824 controls and tested in a second independent sample of 363 prevalent cases and 1,417 controls. Using DNA from 1,607 unaffected individuals, we tested whether DNAm-RS could discriminate the 190 incident cases of lifetime MDD from the 1,417 individuals who remained unaffected at follow-up. Results A weighted linear combination of 196 CpG sites were derived from the training sample to form a DNAm-RS. The DNAm-RS explained 1.75% of the variance in MDD risk in an independent case-control sample and significantly predicted future incident episodes of MDD at follow up (R 2 =0.52%). DNAm-RS and MDD polygenic risk scores together additively explained 3.99% of the variance in prevalent MDD. The DNAm-RS was also significantly associated with lifestyle factors associated with MDD, including smoking status (β=0.440, p=<2×10 −16 ) and alcohol use (β=0.092, p=9.85×10 −5 ). The DNAm-RS remained significantly associated with MDD after adjustment for these environmental factors (independent association: β=0.338, p=1.17×10 −7 association post-adjustment: β=0.081, p=0.0006). Conclusions A novel risk score of MDD based on DNAm data significantly discriminated MDD cases from controls in an independent dataset, and controls who would subsequently develop MDD from those who remained unaffected. DNAm-RS captured the effects of exposure to key lifestyle risk factors for MDD, revealing a potential role in risk stratification.