Whole Genome Deconvolution Unveils Alzheimer’s Resilient Epigenetic Signature.

Abstract

AbstractBackgroundThe pathological hallmarks of Alzheimer’s disease (AD) are amyloid beta plaques and neurofibrillary tangles that usually progressively affect characteristic brain regions as individuals advance from latent to prodromal to severe AD dementia (ADD). An uncommon exception to coupled progression of disease burden and clinical impairment are resilient individuals who harbor a high pathological load sufficient to cause ADD yet remain cognitively normal. We hypothesize that determining the biological mechanisms of such underlying resilience to AD (RAD) will provide unique insights into therapeutic opportunities.MethodHere, we combined epigenetic sequencing data and an innovative computational algorithm to identify gene regulatory mechanisms and putative mediators distinguishing RAD from normal control and ADD at an unprecedented scale. We developed and validated a novel machine learning method, Cellformer, for deconvoluting bulk ATAC‐seq into cell type‐specific expression across the whole genome, leading to extensive regional and cellular chromatin profiling of RAD.ResultCellformer successfully deconvoluted in silico bulk ATAC‐seq data across cross‐validation iterations, achieving strong performances with a mean Spearman coefficient of 0.82 between deconvoluted and ground truth expressions, outperforming other approaches. Validation was performed using external single nucleus datasets and showed a significant correlation above 0.75 (P‐value<0.05) between deconvoluted cell type‐specific ATAC‐seq and single‐ATAC‐seq expressions for each cell type. Applied on 191 samples from normal control, RAD and ADD’s 3 brain regions, CellFormer predicted cell type‐specific expression of more than 41954 open chromatin regions (OCR), preserving regional and phenotypic variations. Downstream analysis of the derived RAD‐specific epigenetic features pinpointed the hippocampus, a region critical for learning and memory, as the major player in the resilience, modulating synaptic and neuronal process in a cell‐type specific manner.ConclusionCellformer opens new perspectives to advance bulk analysis and understand biological mechanisms using cost effective bulk sequencing. Through this study, we showed that cell type‐resolved data nominated potential epigenetic and transcriptional mediators underlying RAD that may illuminate therapeutic opportunities to limit the cognitive impact of this highly prevalent yet stubbornly incurable disease. Cellformer will be made publicly available to advance analysis of high‐throughput bulk ATAC‐seq in future investigations.