SmartAD for Intelligent Alzheimer’s Disease(AD) Personalized Combination Therapy

点击翻译按钮获取中文摘要

Alzheimer’s Disease (AD) is a complex neurodegenerative disease that causes progressive memory loss and cognitive impairment. While current treatments have shown some amelioration of symptoms, the effects have been transient and limited to a small percentage of AD patients. Moreover, disease-modifying drugs based on current understanding of disease mechanisms have all shown negative results in clinical trials. Part of the failure is due to the heterogeneity in the disease mechanism, of which we do not yet have a clear understanding. Increasing evidence has indicated that medical comorbidities share common disease pathways with AD, and the medications used for these diseases can also alter the cognitive functions of AD patients. However, limited studies have assessed combinations of these medications as treatments for AD with common comorbidities. Thus, the goal of this proposal is to develop artificial intelligence (AI) analytics models and a SmartAD app to facilitate cognitive function evaluation and personalized treatment plans for AD patients with the most common comorbidities, such as cardiovascular diseases (CVD)/hypertension (HTN), diabetes mellitus (DM), and depression (DPN). To achieve our goal, we will carry out retrospective analysis of observational clinical data collected by the University of Pittsburgh Alzheimer’s Disease Research Center (ADRC). First, we will statistically investigate the effects of different comorbidity medications when used in combination with anti-AD medications on the trajectory of cognitive decline (Aim1). By identifying specific drug combination(s) that have a synergistic effect against cognitive decline, we will then study the underlying mechanisms using molecular systems pharmacology methods and validate the findings using in vitro iPSC and other bioassays as needed (Aim2). Subsequently, we will build a clinical decision support system, SmartAD, that will facilitate cognitive function evaluation and individualized treatment for AD patients with these common comorbidities. We will build a Bayesian Network model that can predict patient-tailored disease progression and treatment information provided by ADRC at the University of Pittsburgh (Aims 3 & 4). This model will be intelligently machine-learned and trained on the ADRC dataset using causal machine-learning approaches. Methodologies of decision theory will then be applied to search for a treatment combination that leads to the optimal outcome for that patient. Finally, we will use external medical data from AD Neuroimaging Initiative (ADNI) and National Alzheimer’s Coordinating Center (NACC) for model systems test validation (Aims 3 and 4). Taken all together, these studies will contribute to the discovery of novel drug combinations for AD patients with comorbidities and develop SmartAD as an intelligent clinical decision support system that can facilitate paperless cognitive function evaluation, progression prediction, as well as assist optimal personalized medication for Alzheimer’s patients.