Statistical methods for population-level cell-type-specific analyses of tissue omics data for Alzheimer's disease

点击翻译按钮获取中文摘要

PROJECT SUMMARY/ABSTRACT Alzheimer's disease (AD) accounts for 60-80% of dementia cases and causes progressive neurodegeneration that ultimately leads to death. While the number of US people with late-onset AD is expected to reach 13.8 million by 2050, its prevention and treatment remain only modestly effective. Many efforts have been made to study AD pathophysiology by collecting and curating rich omics data from AD-affected or unaffected human brains, e.g., the National Institute on Aging's Accelerating Medicines Partnership for Alzheimer's Disease (AMP-AD) project. Most of those omics data, such as gene expression, DNA methylation, and proteomics, are collected at the tissue level, and thus the cell-type-specific (CTS) signals are masked. Recently, with the emerging single-cell techniques, single-cell RNA-seq and DNA methylation data have been generated. However, given the difficulty of quantifying a small number of molecules and associated high costs, single-cell data suffer from high technical variation and are constrained to a small number of samples that lack representativity. To address these issues in AD research and accelerate our understanding of cellular multi-omics mechanisms underlying AD, we aim to: 1) Improve estimation of cellular fractions in brain tissue samples by the ensemble over existing methods and considering cell-type hierarchy. 2) Identify CTS differentially methylated regions (DMR) associated with AD. We will consider the spatial correlation of CpG sites and cell-type specificity. 3) We will further build statistical models to systematically integrate those CTS omics estimates via omics-wide association studies and causal mediation analyses. Through extensive analyses of several large cohorts in AMP-AD datasets, we will produce statistically significant and biologically meaningful omics results at an unprecedented population-scale and cell-type resolution, which will improve our understanding of complex AD biology. We will validate our findings using additional data available within and outside the AMP-AD project, including single-cell multi-omics data. The resulting methods will be implemented as efficient computational algorithms via public software readily available to the research community. Successful completion of this project will provide state-of-the-art methods for cell- type deconvolution and integrative multi-omics analyses and advance our knowledge of genes/proteins contributing to AD in selectively vulnerable brain regions and cell types.