Multi‐omics data integration reveals clinically meaningful molecular profiles of Alzheimer disease. (December 2021)

• Record Type: Journal Article
• Authors: Eteleeb, Abdallah M; Novotny, Brenna C; Morris, John C.; Bateman, Randall J.; Perrin, Richard J.; Karch, Celeste M.; Cruchaga, Carlos; Benitez, Bruno A.; Harari, Oscar
• Is Part Of: Alzheimer's & dementia. Volume 17(2021)Supplement 3
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-12
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.052942
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Alzheimer disease (AD) is a complex polygenic disease in which multiple molecular pathways and biological processes are affected in distinct cell‐types of the brain. Increasingly, novel machine learning approaches have been applied in other areas of similarly high complexity (e.g., cancer) to integrate different modalities of "‐omics" data and provide novel insights. We sought to apply high‐throughput "multi‐omics" and machine learning approaches to a large assembly of clinically and neuropathologically well‐characterized brain samples to identify genes and biological pathways that may lead to new biomarkers and therapies for AD. Method: We analyzed high‐throughput transcriptomics, proteomics, metabolomics and lipidomics profiles of postmortem parietal cortex samples from the Knight ADRC (n=328) and the DIAN (n =21) participants including APP, PSEN1 and PSEN2 autosomal dominant AD mutation carriers (ADAD, n=28), sporadic AD (sAD, n=282), presymptomatic AD (preAD, n=14), and controls with minimal neuropathologic changes (n=25). We applied a Bayesian integrative clustering method (iClusterBayes), designed to identify disease subtypes, to cluster sAD samples on the basis of 2, 676 transcripts, 1, 067 proteins, 350 metabolites, and 277 lipids. Result: Our analyses identified four different molecular signatures among sAD cases (Figure 1A). Cluster 4 was associated with higher Clinical Dementia Rating (CDR; p=2.2 x10‐20) scores and shorter life span (p=5.6 x10‐3, HR=1.6; Figure 1B). In particular, alpha‐synuclein (SNCA) transcriptomic and proteomic levels were differentially expressed between samples in cluster 4 and others representing sAD, ADAD, preAD, and controls (Figure 1C) suggesting a unique role for SNCA in sAD cases with unfavorable outcomes. The lower levels of soluble SNCA protein in AD samples may be associated with insoluble aggregated SNCA (Lewy bodies) reported in 50‐60% of sAD. Conclusion: By applying machine learning approaches to sAD samples with high‐throughput multi‐omic profiles, we identified novel molecular signatures missed by single layer analyses. These signatures are associated with clinical phenotypes and offer new insights into which molecules may wax or wane as cognition declines. The association of SNCA with poor outcomes suggests that concomitant Lewy bodies may be a negative prognostic factor in sAD. Currently, we are extending these analyses to incorporate replication cohorts and to perform downstream pathway and enrichment analyses.