Computational Drug Repurposing for Alzheimer's Disease via Sheaf Theoretic Population-Scale Analysis of snRNA-seq Data
Abstract
Single-cell and single-nucleus RNA sequencing (scRNA-seq /snRNA-seq) are widely used to reveal heterogeneity in cells, showing a growing potential for precision and personalized medicine. Nonetheless, sustainable drug discovery must be based on a population-level understanding of molecular mechanisms, which calls for the population-scale analysis of scRNA-seq/snRNA-seq data. This work introduces a sequential target-drug selection model for drug repurposing against Alzheimer's Disease (AD) targets inferred from population-level snRNA-seq studies of AD progression in microglia cells as well as different cell types taken from an AD affected brain vascular tissue atlas, involving hundreds of thousands of nuclei from multi-patient and multi-regional studies. We utilize Persistent Sheaf Laplacians (PSL) to facilitate a Protein-Protein Interaction (PPI) analysis of AD targets inferred from differential gene expression (DEG), and then use machine learning models to predict repurpose-able DrugBank compounds for molecular targeting. We screen the efficacy of different DrugBank small compounds and further examine their central nervous system (CNS)-relevant ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity), resulting in a list of lead candidates for AD treatment. The list of significant genes establishes a target domain for effective machine learning based AD drug repurposing analysis of DrugBank small compounds to treat AD related molecular targets.