Predicted plasma proteomics from genetic scores and treatment outcomes in major depression: a meta-analysis

Abstract

Proteomics has been scarcely explored for predicting treatment outcomes in major depressive disorder (MDD), due to methodological challenges and costs. Predicting protein levels from genetic scores provides opportunities for exploratory studies and the selection of targeted panels. In this study, we examined the association between genetically predicted plasma proteins and treatment outcomes - including non-response, non-remission, and treatment-resistant depression (TRD) - in 3559 patients with MDD from four clinical samples. Protein levels were predicted from individual-level genotypes using genetic scores from the publicly available OmicsPred database, which estimated genetic scores based on genome-wide genotypes and proteomic measurements from the Olink and SomaScan platforms. Associations between predicted protein levels and treatment outcomes were assessed using logistic regression models, adjusted for potential confounders including population stratification. Results were meta-analysed using a random-effects model. The Bonferroni correction was applied. We analysed 257 proteins for Olink and 1502 for SomaScan; 111 proteins overlapped between the two platforms. Despite no association was significant after multiple-testing correction, many top results were consistent across phenotypes, in particular seven proteins were nominally associated with all the analysed outcomes (CHL1, DUSP13, EVA1C, FCRL2, KITLG, SMAP1, and TIM3/HAVCR2). Additionally, three proteins (CXCL6, IL5RA, and RARRES2) showed consistent nominal associations across both the Olink and SomaScan platforms. The convergence of results across phenotypes is in line with the hypothesis of the involvement of immune-inflammatory mechanisms and neuroplasticity in treatment response. These results can provide hints for guiding the selection of protein panels in future proteomic studies.