Implicating causal brain imaging endophenotypes in Alzheimer's disease using multivariable IWAS and GWAS summary data

Abstract

Recent evidence suggests the existence of many undiscovered heritable brain phenotypes involved in Alzheimer's Disease (AD) pathogenesis. This finding necessitates methods for the discovery of causal brain changes in AD that integrate Magnetic Resonance Imaging measures and genotypic data. However, existing approaches for causal inference in this setting, such as the univariate Imaging Wide Association Study (UV-IWAS), suffer from inconsistent effect estimation and inflated Type I errors in the presence of genetic pleiotropy, the phenomenon in which a variant affects multiple causal intermediate risk phenotypes. In this study, we implement a multivariate extension to the IWAS model, namely MV-IWAS, to consistently estimate and test for the causal effects of multiple brain imaging endophenotypes from the Alzheimer's Disease Neuroimaging Initiative (ADNI) in the presence of pleiotropic and possibly correlated SNPs. We further extend MV-IWAS to incorporate variant-specific direct effects on AD, analogous to the existing Egger regression Mendelian Randomization approach, which allows for testing of remaining pleiotropy after adjusting for multiple intermediate pathways. We propose a convenient approach for implementing MV-IWAS that solely relies on publicly available GWAS summary data and a reference panel. Through simulations with either individual-level or summary data, we demonstrate the well controlled Type I errors and superior power of MV-IWAS over UV-IWAS in the presence of pleiotropic SNPs. We apply the summary statistic based tests to 1578 heritable imaging derived phenotypes (IDPs) from the UK Biobank. MV-IWAS detected numerous IDPs as possible false positives by UV-IWAS while uncovering many additional causal neuroimaging phenotypes in AD which are strongly supported by the existing literature.

Keywords: Causal inference; Genetic pleiotropy; Instrumental variable; MRI; MV-IWAS; Mendelian randomization; TWAS.

Fig. 1.

a) A causal directed acyclic graph (DAG) representing the assumed model for univariate IV analysis with satisfaction of all 3 instrumental variable assumptions: 1) Relevance: Z is correlated with X, 2) Exchangability: Z is uncorrelated with confounders, and 3) Exclusion Restriction: Z does not affect Y, except through X. Here, C represents the set of additional causal phenotypes for disease which are not associated with the IVs. b) The assumed model under the MV-IWAS framework, allowing the IVs to be pleiotropic for multiple causal phenotypes (X₁ and X₂) for Y. Here, Z₁ and Z_p exhibit horizontal pleiotropy, while Z₂ does not. Under this setting, UV-IWAS estimates for X₁ or X₂ will be inconsistent. c) The assumed model for MV-IWAS-Egger, where SNPs Z are pleiotropic for pathways outside of X₁ and X₂, as represented by the direct pathway from Z to Y.

Fig. 2.

Workflow for applied analysis of 3 different data sources.

Fig. 3.

(a)LD Correlations for the 20 simulated SNPs, estimated using ADNI1 data. (b)LD Correlations for the 20 simulated SNPs, estimated using 1000G data.

Fig. 4.

95% confidence intervals for β₁ from the first 100 iterations from the 2 sample simulations for a quantitative disease trait. The true value for β₁ = 0.

Fig. 5.

Simulation Type I Error and Power across different magnitudes of ${\sigma }_{\tau }^2$ and μ. For each setting of μ: i) Type I error for β₁ ii) Power for β₂ iii) Power for β₃ iv) Power/Type I Error for μ. LD estimated using 1000 Genomes for all Summary Statistic IWAS methods (Summ MVI-WAS and Summ MVIWAS-Egger).

Fig. 5.

Simulation Type I Error and Power across different magnitudes of ${\sigma }_{\tau }^2$ and μ. For each setting of μ: i) Type I error for β₁ ii) Power for β₂ iii) Power for β₃ iv) Power/Type I Error for μ. LD estimated using 1000 Genomes for all Summary Statistic IWAS methods (Summ MVI-WAS and Summ MVIWAS-Egger).

Fig. 5.

Simulation Type I Error and Power across different magnitudes of ${\sigma }_{\tau }^2$ and μ. For each setting of μ: i) Type I error for β₁ ii) Power for β₂ iii) Power for β₃ iv) Power/Type I Error for μ. LD estimated using 1000 Genomes for all Summary Statistic IWAS methods (Summ MVI-WAS and Summ MVIWAS-Egger).

Fig. 6.

Correlations between the 14 Imputed ADNI1 Endophenotypes.

Fig. 7.

Manhattan Plots reflecting Stage 1 Models for the 2 ADNI (left) Endophenotypes significant for MV-IWAS with comparison to comparable UKBB T1 FAST (right) IDPs. Note the difference of y-axis scale for the ADNI and UKBB plots; the notably higher peaks for the UKBB marginal p-values can be explained by the substantial difference in sample size between the two studies. We further note that UKBB IDP 0053 is a measure of only the anterior division of the left inferior temporal gyrus and is thereby not directly comparable to the corresponding ADNI measure.

Fig. 8.

a) Distribution of pairwise pearson correlations for IDPs with unadjusted UV-IWAS p-values below 0.05 by modality. This includes 306 dMRI, 279 sMRI and 146 fMRI imputed IDPs. b) 3 IDP pairs of sMRI IDPs with correlations > 0.75. c) 7 pairwise correlations > 0.75 between imputed dMRI IDPs.

Fig. 9.

Manhattan Plots for Stage 1 SNPs used for each of the IDPs with the greater causal effect estimate for MV-IWAS-Egger.

Fig. 10.

Number of SNPs which are included in 1, 2, 3, 4, 5–50, or 50+ UK Biobank IDP’s Stage 1 IWAS model by modality. Recurrence in more than one Stage 1 IDP model gives evidence for possible pleiotropic effects that will cause inconsistency in the univariate IWAS approach.

Fig. 11.

Comparison of the significant phenotypes identified by the univariate and multivariate IWAS tests for Structural, Diffusion, and Functional MRI IDPs. The quantities given within the red, blue, and grey circles indicate the number of IDPs which were significant for MV-IWAS-Egger, MV-IWAS, and UV-IWAS, respectively. Their intersection reflects IDPs which were significant under all 3 tests.