For my Master's capstone project, I developed a metabolomics-based framework to better understand and detect major depressive disorder (MDD) by moving beyond the typical "one-size-fits-all" approach. Most research on depression treats MDD as a single, uniform condition, despite how differently it presents across individuals. I wanted to explore whether specific metabolic biomarkers could be used to reflect varying levels of MDD severity. Using large-scale UK Biobank data, I engineered network pipelines to isolate the most informative metabolic features from over 250 biomarkers, then applied multivariable regression models to assess how these biomarkers change across progressively severe depression phenotypes.
The lipid-related metabolites, notably triglycerides and very-low-density lipoproteins (VLDL), showed stronger associations as MDD severity increased. This suggests metabolic signatures could help distinguish between mild, self-reported depression and more clinically-defined, recurrent forms of the disorder. The real inspiration for this work is it's potential to be used for early-detection, especially given how hard it is to stay in remission once affected by MDD. In addition, it provides the foundation for developing a more affordable, accessible blood-based diagnostic tool for those struggling with MDD. Equally important, this approach avoids generalizing depression into a single phenotype and instead emphasizes stratifying individuals based on severity, allowing for more targeted and effective interventions.