P4‐027: Combing Evidence Across Multiple Cohorts for Systems‐Based Target Discovery: the AMP‐AD Knowledge Portal

Kristen Dang,Thanneer M. Perumal,Mariet Allen,Corey Funk,Minghui Wang,Jishu Xu,Ben Logsdon,Lei Yu,Sarah Schuyler,Stephen H. Friend,David A. Bennett,Bin Zhang,Eric Schadt,Philip de Jager,Nathan D. Price,Nilufer Ertekin-Taner,Lara M. Mangravite

DOI: https://doi.org/10.1016/j.jalz.2016.06.2116

2016-01-01

Abstract:Accumulating failures in Alzheimer's disease drug development highlight a critical need to improve understanding of the etiology of this disease. The AMP-AD consortium is seeking to advance this understanding and support target discovery through genomic analysis of human disease. To date, the consortium has generated genomic data from blood and postmortem brain samples collected across more than ten sources. The heterogeneity in this collection provides an unprecedented opportunity to identify mechanisms that are robustly observed across multiple representations of this complex disease. Here, we evaluate the continuity in disease-associated differential gene expression in RNA sequencing data collected across four sources. All genomic data generated on brain collections used in AMP-AD was deposited in a publicly accessible to date, centralized repository including RNA sequencing data generated from postmortem brain tissue from four sources: the ROS and MAP cohort studies (N=547), the Mayo Brain Bank (N=550), and the Mt. Sinai Brain Bank (N=448). Experimental design differed across studies in sample ascertainment, criteria for diagnosis, data generation and processing methods. For 10 samples from each source, data generation was repeated across sequencing centers. Gene-level differential expression was calculated separately within each source by linear regression using VOOM normalized data. When multiple brain regions from the same individual were available, data were jointly analyzed using a mixed effects model. Meta-analysis was performed using Fisher's method with Brown's correction. Despite major sources of study-specific technical and biological variation, differences in gene expression between samples classified as Alzheimer's disease versus control significantly overlapped across studies. 11696 genes were differentially expressed in at least one study (FDR=5%). Meta-analysis across all studies identified 9186 differentially expressed genes. 582 genes were differentially expressed in more than 75% of studies and these were enriched for genes previously demonstrated to be involved in neurodegenerative processes. Combined analysis identified a high-confidence set of genes that were differentially expressed in relation to Alzheimer's disease in a consistent manner across a diverse collection of studies. Future analytical efforts that leverage this heterogeneous collection have the potential to identify robust disease mechanisms in support of target discovery.