Anti-inflmmation Treatment in Mood Disorder and Deep Learning Prediction Model

Overview

This three-year study will enroll 180 patients with mood disorders (90 patients with major depressive disorder and 90 patients with bipolar disorder) and high pro-inflammatory cytokine levels. They will be randomly assigned to three groups of aspirin, statin and control groups for 12 weeks according to the disease group. The first aim of the study is to compare the efficacy of aspirin and statin in mood disorders. The second aim is to establish a gene-immuno-brain imaging treatment prediction model by deep learning technology, using pretreatment cytokines, neurocognitive function, brain structural/functional connectivity, and telomere length as the predictors.

Multiple lines of evidence support the pathogenic role of neuro-inflammation in mood disorders. Our team has published a series of papers showing the inflammatory cytokines are related to severity of depressive symptoms, could be biomarkers of clinical outcomes, subtype and mood phase of bipolar disorder. Compared with depressive disorder, bipolar disorder is with more severe inflammatory dysregulation, which correlated to brain structure and functional connectivity abnormality. Treatment non-responders tended to have higher baseline inflammatory markers, suggesting that increased levels of inflammation are contributory to treatment resistance. The clinical studies showed that anti-inflammatory drugs combined with traditional treatments, can improve clinical outcomes, including N-Acetylcysteine, infliximab, pioglitazone, celecoxib, aspirin, omega-3 polyunsaturated fatty acids, minocyclin, statin, aspirin. Among them, aspirin and statin have been used for treatment and prevention of cardiovascular metabolic disorders, which are associated with inflammation dysregulation. The clinical and meta-analysis studies of aspirin and statin have shown significant efficacy and good safety. Therefore, aspirin and statin have better clinical feasibility and rationality for augmentation treatment in mood disorders. However, previous anti-inflammatory research is mostly for individual drug studies, comparative research is still quite lacking. In addition, many studies have suggested anti-inflammatory agents will likely be most useful for the subpopulation of patients whose immune dysfunction is a driving pathogenic factor. In this study, we will establish a prediction model of anti-inflammatory drugs for mood disorder. Recent advances in deep learning have demonstrated its power to learn and recognize complex nonlinear hierarchical patterns based on largescale empirical data. A deep learning algorithm for classification applications such as medical treatment in personalized medicine is a procedure for choosing the best hypothesis from a set of alternatives that fit a set of observations. Our series of studies have shown that the severity of inflammation related with brain structure and functional connectivity abnormalities; which may be the outcome predictors. Another possible predictor may be the chromosome telomere length. Telomeres are located at the end of chromosomes and maintain normal function of chromosomes. Previous studies have found that short telomere length is associated with mood disorder, as well as the inflammatory dysregulation. Therefore, telomere length may be a predictor of anti-inflammatory treatment. The study will be the first comparative study of anti-inflammatory treatment, and establish gene-immuno-brain imaging individualized treatment prediction model. The results will provide important scientific and clinical empirical data for the inflammatory pathophysiology and treatment of mood disorders.