This proposal focuses on the challenge of identifying drug repositioning candidates for Alzheimer’s Disease. The
foundation of this work is the ReFRAME library, a set of ~13,000 compounds that includes nearly all small molecules that
have been FDA-approved, reached clinical development, or undergone significant preclinical profiling. The ReFRAME
library is being actively screened against a diverse cross-section of in vitro assays.
This proposal pursues three distinct strategies for identifying repositioning candidates among the ReFRAME collection.
First, we will create and mine a large and heterogeneous biomedical knowledge graph. We will use machine learning
methods to identify repositioning candidates based on properties of the knowledge graph surrounding and joining each
drug and disease. Second, we will mine a massive data set of insurance claims data for associations between drug use
and the incidence or severity of Alzheimer’s Disease. Containing almost 7 billion medical claims and over 2 billion
pharmacy claims, this data set represents the largest source of claims data available. Third, we will use concept of gene
expression complementarity to identify repositioning candidates. We will generate a gene expression signature for
every ReFRAME compound in three cell lines relevant to Alzheimer’s Disease, and we will screen for compounds that
produce a signature that appear to reverse gene expression changes seen in Alzheimer’s Disease.
After assembling repositioning candidates identified through all three of these methods, we will prioritize up to 100
compounds (or compound combinations) for further characterization and validation. These follow-up experiments will
initially investigate the activity of these compounds in five cell-based assays to establish a mechanistic hypothesis on
their mechanism of action in Alzheimer’s Disease. Secondary follow-up experiments may include validation in some
combination of in vitro (including hiPSC-derived cerebrocortical neurons and/or organoids) and in vivo systems.
We believe that the multifaceted approach described in this proposal offers the best possible chance at successfully
identifying AD repositioning candidates. Moreover, this work will create methods and resources that will be useful to
the broader scientific community, both for Alzheimer’s Disease and for other disease areas.