Can targeted metabolomics predict depression recovery? Results from the CO-MED trial

Abstract

Metabolomics is a developing and promising tool for exploring molecular pathways underlying symptoms of depression and predicting depression recovery. The AbsoluteIDQ™ p180 kit was used to investigate whether plasma metabolites (sphingomyelins, lysophosphatidylcholines, phosphatidylcholines, and acylcarnitines) from a subset of participants in the Combining Medications to Enhance Depression Outcomes (CO-MED) trial could act as predictors or biologic correlates of depression recovery. Participants in this trial were assigned to one of three pharmacological treatment arms: escitalopram monotherapy, bupropion-escitalopram combination, or venlafaxine-mirtazapine combination. Plasma was collected at baseline in 159 participants and again 12 weeks later at study exit in 83 of these participants. Metabolite concentrations were measured and combined with clinical and sociodemographic variables using the hierarchical lasso to simultaneously model whether specific metabolites are particularly informative of depressive recovery. Increased baseline concentrations of phosphatidylcholine C38:1 showed poorer outcome based on change in the Quick Inventory of Depressive Symptoms (QIDS). In contrast, an increased ratio of hydroxylated sphingomyelins relative to non-hydroxylated sphingomyelins at baseline and a change from baseline to exit suggested a better reduction of symptoms as measured by QIDS score. All metabolite-based models performed superior to models only using clinical and sociodemographic variables, suggesting that metabolomics may be a valuable tool for predicting antidepressant outcomes.

Metabolomics in Psychiatric Disorders: What We Learn from Animal Models

“…Previous studies in humans looked at a wide array of blood metabolites including lipids, AA, hormones, and biogenic amines to predict patient response to drug therapy [3]. More specifically, metabolomics has been applied following treatments with specific antidepressants, including sertraline [83], escitalopram [84], ketamine, and esketamine [84], as well as combinations of antidepressants (bupropion-escitalopram combination, venlafaxine-mirtazapine combination [85]). However, despite these studies already conducted in humans, animal models offer further important insights into the study of the therapeutic actions of drugs, thereby facilitating the development of novel treatments.…”

Metabolomics in Radiation Biodosimetry: Current Approaches and Advances

“…A recent meta-analysis [86] found strong associations between lipid metabolites and the prevalence of a depressive state. In fact, there are studies that suggest that the presence of certain metabolic markers can be used to predict both the propensity to develop depression [87], as well as the ability of patients taking anti-depression medications to demonstrate recovery [88]. Another group found that there were significant differences in the parts of the metabolomic profile associated with energy metabolism in the brain that could be linked, along with protein changes, to major depressive disorders [89].…”

Pharmacogenomics‐Driven Prediction of Antidepressant Treatment Outcomes: A Machine‐Learning Approach With Multi‐trial Replication

“…To address this limitation, recent machine‐learning approaches have used genomics and/or metabolomics data to predict SSRI response, with AUC values of 0.68–0.78 . Although the studies demonstrated the feasibility of integrating biological factors with machine learning to achieve improved prediction performance, they were limited by lack of replication across multiple trials and multiple depression rating scales, in addition to lack of significance from a pharmacogenomics perspective …”

“…[8][9][10] Although the studies demonstrated the feasibility of integrating biological factors with machine learning to achieve improved prediction performance, they were limited by lack of replication across multiple trials and multiple depression rating scales, in addition to lack of significance from a pharmacogenomics perspective. [8][9][10][11] In the present study, we used a machine-learning workflow (depicted schematically in Figure 1) to study the capabilities of Figure 1 The two-stage analysis workflow. Our analysis workflow proceeded in two stages.…”

“…These analyses are sex-stratified, given the established evidence of sex differences in the prevalence of depression and increasing evidence of sex differences in antidepressant response. 11,[14][15][16][17][18][19][20][21] The pharmacogenomic biomarkers included in the present study are six SNPs in or near the TSPAN5 (rs10516436), ERICH3 (rs696692), DEFB1 (rs5743467, rs2741130, and rs2702877), and AHR (rs17137566) genes. Each of these SNPs were the "top" SNP in its respective genomewide association study (GWAS) SNP signal, except that for DEFB1, we included the "top" SNP as well as two others in different haplotype blocks.…”