Leveraging predicted gene expression data for recapitulation of gene coexpression network analysis associations with AD pathology and cognitive decline: Genetics/omics and systems biology. (7th December 2020)

• Record Type: Journal Article
• Authors: Seto, Mabel; Janve, Vaibhav A.; Logsdon, Benjamin A.; Mostafavi, Sara; Dumitrescu, Logan; Mahoney, Emily R.; Gaiteri, Chris; Schneider, Julie A.; Bennett, David A.; De Jager, Philip L.; Hohman, Timothy J.
• 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.046394
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Gene co‐expression network (GCN) analysis is an approach in which biologically relevant modules can be identified from gene expression data, allowing for the elucidation of biological functions that are associated with Alzheimer's disease (AD). Here, we evaluate the associations between generated from dorsolateral prefrontal cortex RNA sequencing (RNAseq) to validate previous reports (Mostafavi et al., 2018) and to identify novel module‐trait associations. As RNAseq data is often limited, we also apply an emerging technique to generate the same modules to provide a proof‐of‐concept for how transcriptomic reference panels can be leveraged to apply GCNs in the context of genome‐wide association analyses. Method: Genotype, neuropsychological data, autopsy measures of AD pathology, and RNAseq were obtained from the Religious Orders Study and Rush Memory and Aging Project. Global cognitive composite scores were calculated from 17 neuropsychological tests. Predicted gene expression data were generated using . Using 66 reported AD co‐expression modules (Logsdon et al. 2019; Mostafavi et al. 2018), we performed linear regression analyses assessing the association between the first principle component of the and cognitive decline and AD neuropathology. Covariates included age of death, sex, education, and post‐mortem interval. Correction for multiple comparisons was completed using the false discovery rate procedure. Result: We recapitulated all reported associations. We also identified 15 novel associations with AD phenotypes using module definitions reported by Logsdon et al. Using predicted gene expression data, we replicated grey60 associations with tangles (p = 0.009) and cognitive decline (p = 0.004). The hub gene of grey60 was PAPOLA, which is a nominated target from AMP‐AD. PrediXcan analyses also recapitulated the m111 association with cognitive decline (p = 0.03, hub gene = TCF12) . Both modules were enriched for transcriptional regulation genes, and higher gene expression in both modules was associated with slower cognitive decline. Conclusion: Predicted gene expression can serve as a surrogate for recapitulating aspects of GCN analyses when RNAseq data is limited. Our findings provide additional evidence that gene networks involved in transcriptional regulation contribute to the neuropathology and cognitive decline observed in AD.