Depression is one of the most common psychiatric disorders, affecting an estimated 322 million individuals world-wide. Depression burdens society, impacting the quality of life for both individuals and families. It is also an economic burden, costing an estimated $210.5 billion per year in health care and workplace-related costs in the United States.
Scientists have spent decades researching the underlying causes of depression. However, there is a lack of consistency in published findings. One reason for this lack of consistency may be that depression is a variable disorder, with no single profile of symptoms or biological changes that are the same across everyone. Individuals with depression have variable symptoms, differences in the number and duration of depressive episodes, variable ages of onset, and different underlying causes that may have led to depression across development. In this project, we will take an approach that incorporates data from the brain, body, and behavior of individuals with depression to better understand variability in the disorder.
We will apply the concept of “degeneracy” to investigate depression across these different domains. “Degeneracy” refers to the ability of structurally different components to result in the same outcome. Under this framework, the lack of consistency in published findings becomes meaningful: two depressed individuals may have the same symptoms, with different circumstances during development that resulted in different biological profiles, but both of which lead to the same functional end of a depression diagnosis. The main aim of this project is to investigate the possibility that there are multiple profiles of depression using data from the brain, body, and behavior.
To investigate the possibility of multiple subgroups of depressed individuals, we will use analytical methods that group data based on similar underlying patterns. Specifically, we will use unsupervised machine learning, a tool that groups data without using any labels (e.g., data are not labeled as “depressed” or “control”). This allows for the potential to discover multiple sub-groups within a single diagnostic category, rather than a single “depressed” group. As input to our analysis, we will use data from the brain (e.g., structural MRI, resting state fMRI), body (e.g., metabolic, immune, cardiovascular), and self-report (e.g., symptom reports, family history). Our results have the potential to inform early diagnosis and prevention of depression, and may inform individualized treatment of the disorder. We plan to complete this project over a three-year period.