Predicting antidepressant non-remission in late-life depression as a function of neurodegenerative and cerebrovascular comorbidities, using genomic machine-learning models
Approved Research ID: 53331
Approval date: October 26th 2020
The biology of late-life depression (LLD) is distinct from early-onset depression. LLD patients present with greater concurrent illness, in large part due to underlying cerebrovascular (e.g., stroke) and neurodegenerative (e.g., Alzheimer's disease, dementia) changes. Globally, more than 50% of LLD patients do not recover sufficiently after treatment with ongoing antidepressant medications, which further predisposes them to cognitive decline, representing an increased risk for dementia and stroke.
Our proposed investigation will use genetic Machine Learning (ML) models, which are algorithms that learn genetic patterns associated with disease. The goal is to identify older adults who are at risk for neurodegenerative and cerebrovascular comorbidities, and antidepressant non-remission, prior to treatment selection. The genomic data is publicly available from large studies of older adults, and from well-characterized clinical samples of older adults treated with antidepressants. By leveraging genomic data and large samples, we aim to discover factors that differentiate and predict early-onset depression from LLD and those that differentiate and predict LLD in patients with and without comorbidities or at risk for cognitive decline from dementia or stroke. Comorbidities and risk of such may affect LLD patients' response to antidepressants. Ultimately, such models will aid in substantially reducing the functional, economic and social burden associated with poorly managed LLD. Within the next years, these findings will provide a platform that will allow us to develop a genetic test to personalize and optimize treatment of LLD and to slow the progression of cerebrovascular and neurodegenerative diseases.