# A multivariate approach to identify association between peripheral blood DNA methylation and cerebrospinal fluid biomarkers of Alzheimer disease

## TL;DR

This study uses a new multivariate method to find links between blood DNA methylation and Alzheimer's biomarkers in cerebrospinal fluid.

## Contribution

The paper introduces a multivariate penalized model to detect weak associations between DNA methylation and AD biomarkers.

## Key findings

- The multivariate approach improves detection of weak methylation-biomarker signals.
- The method was validated using simulations and data from the Canadian Longitudinal Study on Aging.
- Adjusting methylation for covariates enhances the accuracy of association detection.

## Abstract

DNA methylation has been shown to play a crucial role in many diseases, including Alzheimer’s disease (AD). Although many studies have correlated DNA methylation in blood samples with risk of clinical AD diagnosis, few have examined links with AD neuropathology. Using data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) study, we investigate the associations between peripheral blood DNA methylation and three AD-associated biomarkers in cerebrospinal fluid: amyloid-\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\beta$$\end{document}, phosphorylated tau-181, and total tau using an innovative multivariate approach. In our approach, we first adjusted the methylation values for covariates that have known wide-spread effects on methylation. We then developed and implemented a multivariate penalized model to find associations, jointly, between CSF biomarkers and sets of methylation residuals defined by regions around each gene. These penalized models then selected probes showing associations with one or more CSF biomarkers. We demonstrate, using both simulations and actual data, that our proposed multivariate approach is beneficial for detecting weak signals. We also provide complementary validation using data from the Canadian Longitudinal Study on Aging. Our multivariate strategy has the potential to increase feature selection accuracy among correlated predictors in epigenetic studies.