Characterization of genetic expression patterns in MCI using a multiomics approach and neuroimaging endophenotypes: Biomarkers (non‐neuroimaging): EEG and other biomarkers. (7th December 2020)

• Record Type: Journal Article
• Authors: Sanjay, Apoorva Bharthur; Patania, Alice; Yan, Xiaoran; Svaldi, Diana; Apostolova, Liana G.
• Is Part Of: Alzheimer's & dementia. Volume 16(2020)Supplement 5
• 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.046472
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Our goal was to characterize gene expression patterns at the MCI stage using a multiomics approach involving the integration of neuroimaging MRI data and peripheral blood‐based gene expression data. Method: Data from the ImaGene study (Imaging and Genetics Biomarkers study of Alzheimer's Disease, UCLA) was used. This included 160 subjects clinically diagnosed with amnestic MCI (n=70), non‐amnestic MCI (n=38) and normal controls (n=52)(Table 1). Coronal T1‐weighted MPRAGE scans were processed in Freesurfer 6.0 using the standard processing pipeline and atrophy measures were obtained. Quantile‐normalized and log‐transformed mRNA levels were obtained from peripheral blood. Preliminary data reduction was performed by selecting 3420 genes whose expression was significantly associated with hippocampal volume and cortical thickness (p<0.05, linear regression model). Vertex‐wise regressions were performed in SurfStat to map the associations of peripheral blood‐based gene expression with brain atrophy. A persistent‐homology pipeline was applied on the regression‐derived beta‐coefficients to derive betti numbers characterizing the association between gene expression and brain topology (Figure 1). The betti numbers were then used in a kernel‐based clustering solution to cluster genes. PCA was used to represent each cluster by its first principal component. Clusters of genes significantly associated with disease diagnosis were identified. Significant clusters were characterized ontologically to identify enriched biological pathways within the clusters. Cluster level association of genes with brain regions was then assessed using vertex wise regressions in SurfStat (Figure 3). Result: The 3420 genes clustered into 20 clusters with good separation (Figure 2). Three disease‐relevant significant clusters emerged (pFDR‐Corrected <0.05) and were validated using a null model. The biological processes associated with each cluster were mitochondrial fatty acid beta oxidation, as well as regulation of NF‐kappaB signaling and apoptosis (Figure 2). Cluster‐level associations with cortical thickness displayed canonical AD‐like patterns (Figure 3).The apoptosis cluster showed a very strong AD like presentation with respect to atrophy pattern. Conclusion: Using a data‐driven approach, we found a systems‐level disease‐relevant gene expression signature that is sensitive to pre‐clinical AD using peripheral blood showing an AD‐like pattern of association with cortical atrophy. This approach might be useful in developing robust risk assessment platforms.