Phenotypic screening and computational modeling for development of combination therapies for Alzheimer's disease

University Of Pittsburgh At Pittsburgh, Pittsburgh PA

Investigators

Abstract

Alzheimer?s disease (AD) is thought to affect 40 million patients worldwide, a likely underestimate given the increasing recognition that the disease begins in the brain 2-3 decades before clinical symptoms are manifest. With an aging population and the staggering economic and emotional burden of patient care, the demand for effective AD therapies continues to grow. Beta amyloid (A?) peptides are thought to be the initial trigger for neurotoxic sequelae that in turn define AD progression; however, the recent clinical failures of a number of investigational drugs aimed at either preventing the production and/or aggregation of A? or enhancing its clearance have highlighted the need to gain a more holistic understanding of the temporal and spatial complexity intrinsic to AD pathophysiology. The pleiotropic effects attributed to A? are a challenge for developing pharmaceutical interventions against neuronal toxicity in AD. In this project, we propose merging phenotypic screening with computational inference to deconstruct A?-induced toxicities into targetable and interacting pathogenic and protective pathways. Our approach iterates experiments with computational modeling and analysis to efficiently identify drug combinations that protect against A?-induced toxicity in human induced pluripotent stem cell (iPSC)-derived AD cell models. We have two specific aims: The first aim is to identify FDA approved drugs that prevent death in iPSC-derived neuronal progenitor cells (NPCs) treated with A?, and then find combinations of these drugs that will synergistically combine to protect cells. Our second aim is to find combinations of approved drugs that target different pathways involved in A?-mediated cell death; here, the drugs considered need not individually demonstrate measurable protection against A? cell death, but we will computationally identify combinations that provide protection. In both aims, hits will be found through cell-based phenotypic screening, and combinations will be predicted using orthogonal computational models, including existing models, novel methods that incorporate drug-target interaction information, and deep learning. We design our study to simultaneously minimize the number of screens required to identify effective combinations, maximize the efficacy of the optimal combinations, and maximize the gain in information about the cellular effects of A?. This project will not only generate novel combinations of FDA approved drugs that slow A?-related cell death in human NPCs, but it will also deepen our understanding of the systems biology of AD pathogenesis, opening the door for development of new therapeutic strategies. At the end of the project, we expect to have several combinations of approved drugs poised for in vivo or clinical testing, as well as a list of pathways that are central to protection of neuronal cells in AD.