Multivariate canonical correlation analysis to delineate ATN specific genetic variants in Alzheimer’s disease

• Published in: Alzheimer's & dementia Vol. 19; no. S12
• Main Authors: Zhuang, Xiaowei, Yang, Zhengshi, Cordes, Dietmar
• Format: Journal Article
• Language: English
• Published: 01.12.2023
• ISSN: 1552-5260; 1552-5279; 1552-5279
• DOI: 10.1002/alz.079292

Background Genome wide association studies (GWAS) in AD have been focused on clinically diagnosed case‐control subjects, and have identified more than 70 AD risk loci1‐3. The biological mechanisms underlying these genetic variants in AD remain unclear. In this study, we investigated the relationship between subjects’ genetic variants and AD biomarkers within the NIA‐AA ATN framework4 using a single holistic multivariate canonical correlation analysis (CCA) approach. Method 932 subjects from the ADNI database5 with at least one 1) clinical visit; 2) 18F‐AV45 amyloid PET scan 3) structural MRI scan; 4) CSF sample; and 5) genotyping data were included in this integrated CCA analysis. More specifically, after imputation and quality control steps, 3323 single‐nucleotide‐polymorphism (SNP) with p‐values<1e‐4 in previous AD‐GWAS1 were included as genetic features. Following the ATN framework, subjects’ amyloid status were quantified by CSF‐derived amyloid beta 42 (Abeta42) values and amyloid PET‐derived standardized uptake value ratio (SUVR) values in major cortices; subjects’ tau status was represented by CSF‐derived phosphorylated tau values; and subjects’ neurodegenerative status were quantified using MRI‐derived cortical thickness measures from AD‐meta‐ROIs6. Total 16 measures were included as ATN biomarkers. Using CCA, we sought to estimate pairs of canonical components along which genetic variants and ATN biomarkers co‐vary in a similar way in these subjects7,8. Statistical significance of each canonical component was determined through the permutation inference9. We further examined SNPs and biomarkers that are most strongly correlated with each significant CCA component identified. Result CCA analysis reveals two significant components that relate AD risk SNVs to ATN biomarkers (r = 0.6381 (p = 0.0002) and r = 0.5776 (p = 0.0046), Fig.1). The first component is most significantly contributed by amyloid biomarkers (Fig.2(A)), and top associated SNVs are centered around the APOE, APOC1, TOMM40, PVRL2 genes on chromosome 19 (Fig.2(B)), which are overrepresented in pathways related to amyloid formation and catabolic process. Furthermore, tau biomarkers significantly contribute to the second CCA component (Fig. 2(A)), and top associated SNVs are additionally involved in the regulation of neurofibrillary tangle assembly pathway. Conclusion Using multivariate CCA method, we identified specific genetic variants covaried with the amyloid (A), tau (T) and neurodegeneration (N) biomarkers in AD, respectively.