Multimodal data integration via mediation analysis with high‐dimensional exposures and mediators

Abstract

Motivated by an imaging proteomics study for Alzheimer's disease (AD), in this article, we propose a mediation analysis approach with high‐dimensional exposures and high‐dimensional mediators to integrate data collected from multiple platforms. The proposed method combines principal component analysis with penalized least squares estimation for a set of linear structural equation models. The former reduces the dimensionality and produces uncorrelated linear combinations of the exposure variables, whereas the latter achieves simultaneous path selection and effect estimation while allowing the mediators to be correlated. Applying the method to the AD data identifies numerous interesting protein peptides, brain regions, and protein–structure–memory paths, which are in accordance with and also supplement existing findings of AD research. Additional simulations further demonstrate the effective empirical performance of the method.

Instrumental variable-based high-dimensional mediation analysis with unmeasured confounders for survival data in the observational epigenetic study

“…Recently, several methodologies have been proposed for the analysis of high-dimensional mediation analysis ( Dai et al, 2020 ; Zhang et al, 2021a ; Yang et al, 2021 ; Perera et al, 2022 ; Wang et al, 2022 ; Zhao and Li, 2022 ; Zhao and Luo, 2022 ). Zhang et al (2016) raised the issue of estimating the high-dimensional mediating effect in survival analysis ( Zhang et al, 2016 ).…”

DNA methylation and aeroallergen sensitization: The chicken or the egg?

“…It could also be speculated that single CpGs might be more relevant in relation to allergic sensitization than methylation across a whole gene, as this is the biggest difference between gHMA as a gene-based approach and the others (HIMA and DACT) as CpG-based approaches. Further, applying the PRS as an exposure, we did not check whether there is significant mediation between single SNPs and CpGs, but with the development of relevant methodology [ 52 ] this is of great interest for future studies. MRS were further determined according to their cross-sectional prediction accuracy and not optimized according to their performance in a prospective or mediation setting as applied here.…”

HILAMA: High-dimensional multi-omic mediation analysis with latent confounding

“…Specifically, for the method “mvregmed” [Schaid et al, 2022], we apply the R package regmed . While for the method developed by Zhao et al [2022] (abbreviated as ZY), we implement their penalized regression algorithm and omit the dimension reduction step for comparison. Here, we only compare the two penalized methods in the low dimensional setup in simulation 2 introduced below due to their slow running time (See Figure 4d).…”

“…Zhang [2022] consider high-dimensional exposures and mediators through two different procedures; however, they require the mediators to be independent and mainly focus on mediator selection. Meanwhile, Zhao et al [2022] develop a novel penalized principal component regression method that replaces the exposures with their principal components in a lower dimension. This approach, however, lacks causal interpretation.…”

“…HILAMA addresses two critical challenges in applying mediation analysis (or any causal inference method) to multi-omics studies: (1) accommodating both high-dimensional exposures and mediators, and (2) handling latent confounding. In contrast to competing methods [Baron and Kenny, 1986, Zhang et al, 2016, Gao et al, 2019, Schaid et al, 2022, Zhao et al, 2022], our method maintains control over FDR at the nominal level for multiple mediation effect testing, even in the presence of latent confounders. Now, we briefly sketch the essential components of our method.…”