N‐of‐1‐pathways transcriptomic analysis reveals distinct subtypes of Alzheimer's disease: Development of new models and analysis methods/novel assays and technologies. (7th December 2020)

• Record Type: Journal Article
• Authors: Uyar, Asli; Pandey, Ravi S.; Preuss, Christoph; Kotredes, Kevin P.; Oblak, Adrian L.; Logsdon, Benjamin A.; Rizzo, Stacey J. Sukoff; Howell, Gareth; Lamb, Bruce T.; Sasner, Michael; Carter, Gregory W.
• Is Part Of: Alzheimer's & dementia. Volume 16(2020)Supplement 2
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-12-07
• 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.046575
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Data from recent studies supports the hypothesis that Alzheimer's disease (AD) is a heterogenous spectrum disorder characterized by multiple genotypes and phenotypes. Several subtypes of AD can be identified based on genetic makeup together with clinical phenotypes, neuroimaging and molecular signatures at different stages of the disease. N‐of‐1‐pathways approaches have been proposed for single‐subject gene expression analysis enabling disease subtyping at the transcriptome level. Method: We analyzed existing RNA‐Seq data generated from post‐mortem brain tissue in the Religious Order Study and the Memory and Aging Project (ROSMAP). We calculated Mahalanobis distance based similarity measures between pair‐wise subjects per gene, and also per KEGG pathway. Using these similarity measures, we applied k‐means clustering algorithm to identify distinct subtypes of AD‐relevant transcriptomic signatures. Alignment of the subtypes with eight different mouse models of AD enabled determination of the mouse model(s) best representing each subtype. Result: In the ROSMAP cohort, 144 and 81 subjects were labeled as AD and Control, respectively, according to predefined thresholds for COGDX, BRAAK and CERAD scores. A total of 186 KEGG pathways were assessed and AD subjects were clustered into four subtypes based on the Mahalanobis distance n‐of‐1‐pathway scores. Neuropathology related pathways, including neurotrophin signaling, axon guidance and long‐term potentiation, were altered in Subtype3 (n = 15). Transcriptomic signatures in the brain of 5XFAD and CR1 mouse models represented similarity with Subtype3. Subtype2 (n = 45) revealed a strong metabolic response with altered lysine and glycan metabolisms similar to ABCA7 and TREM2 mouse models. Immunity related pathways were enriched in Subtype3 (n = 48) mapping to ABCA7 and APOE4 mouse models. Subtype4 (n = 36) did not reveal a significant pathway enrichment. Conclusion: Accurate identification of disease subtypes is critical for understanding the distinct molecular mechanisms of AD to provide effective personalized treatment options. To our knowledge, this is the first time that n‐of‐1‐pathways approach has been utilized for subtyping AD based on single‐subject transcriptomic signatures. The AD‐relevant pathways altered in each subtype was associated with specific mouse models suggesting a potential for utilization of the matching "mouse model/AD subtype" pairs for preclinical testing of therapeutic targets.