Comparing personalized brain-based and genetic risk scores for major depressive disorder in large population samples of adults and adolescents.

Background: Major depressive disorder (MDD) is a polygenic disorder associated with brain alterations but until recently, there have been no brain-based metrics to quantify individual-level variation in brain morphology. Here, we evaluated and compared the performance of a new brain-based 'Regional Vulnerability Index' (RVI) with polygenic risk scores (PRS), in the context of MDD. We assessed associations with syndromal MDD in an adult sample ( = 702, age = 59 ± 10) and with subclinical depressive symptoms in a longitudinal adolescent sample (baseline  = 3,825, age = 10 ± 1; 2-year follow-up  = 2,081, age = 12 ± 1).

Methods: MDD-RVIs quantify the correlation of the individual's corresponding brain metric with the expected pattern for MDD derived in an independent sample. Using the same methodology across samples, subject-specific MDD-PRS and six MDD-RVIs based on different brain modalities (subcortical volume, cortical thickness, cortical surface area, mean diffusivity, fractional anisotropy, and multimodal) were computed.

Results: In adults, MDD-RVIs (based on white matter and multimodal measures) were more strongly associated with MDD ( = 0.099-0.281, P = 0.001-0.043) than MDD-PRS ( = 0.056-0.152, P = 0.140-0.140). In adolescents, depressive symptoms were associated with MDD-PRS at baseline and follow-up ( = 0.084-0.086,  = 1.38 × 10-4.77 × 10) but not with any MDD-RVIs ( < 0.05,  > 0.05).

Conclusions: Our results potentially indicate the ability of brain-based risk scores to capture a broader range of risk exposures than genetic risk scores in adults and are also useful in helping us to understand the temporal origins of depression-related brain features. Longitudinal data, specific to the developmental period and on white matter measures, will be useful in informing risk for subsequent psychiatric illness.