BA23 - Integrative genomics for risk of CHD and related phenotypes in the WHI

Investigator Names and Contact Information

Phil Tsao (ptsao@stanford.edu)

Introduction/Intent

This page provides study documentation for BA23. For description of the specimen results, see Specimen Results Description (open to public). Data sets of the specimen results are included in the existing WHI datasets located on the WHI Data on this site (sign in and a completed Data Distribution Agreement are required; see details on the Data site).

PIs (multi-PI model): Themistocles (Tim) Assimes & Phil Tsao (Stanford University School of Medicine); Devin Absher (HudsonAlpha Institute of Biotechnology); Steve Horvath (University of California Los Angeles)

Investigative Team: Stephen Pan, Matthew Wheeler, Mike Snyder (Stanford University), Peter Langfelder (University of California Los Angeles); Lesley Tinker (Fred Hutchinson Cancer Research Center)

Study Objectives

• to profile ~1070 cases and ~1070 controls for miRNA and methylation
• to identify miRNA and methylation genomic biomarkers of CHD events
• to integrate biomarkers into diagnostic and prognostic predictors of CHD and related phenotypes
• to test whether aging related CpGs found in our preliminary studies mediate the risk of age on CHD
• to elucidate the biology of CHD and related phenotypes through a method referred to as weighted gene co‐expression network analysis (WGCNA)
• to revisit the GWAS studies of CHD using new SNP set analysis informed by our new genomic data

Specific Aim 1

To apply validated high throughput genomic procedures developed by Applied Biosystems and Illumina to the plasma and DNA samples of a subset of WHI participants with existing GWAS and CVD biomarker data. Using these procedures, we will precisely measure the levels of ~750 miRNAs in plasma as well as the methylation status of over 480k genomic loci in circulating WBCs. We will obtain these measures in a multi-ethnic case-cohort sample from baseline that includes ~1070 cases occurring between baseline and 2011 and ~1070 controls.

Specific Aim 2

We will apply standard as well as state-of the art statistical procedures to the data generated in Aim 1 including machine learning methods as well as a recently developed highly accurate and sparse predictor (Random Generalized Linear Model) to test promising CpG and miRNA biomarkers for association with the presence of or the development of clinical CHD and traditional risk factors (TRFs). We will train our prediction algorithms in half the sample and validated them in the other half (~535 cases+535 controls each for CHD outcome). We will explore whether the resulting prediction algorithms differ for subtypes of CHD presentation and timing of presentation in relation to blood draw. Specifically, we will test the following hypotheses based on published work by others and ourselves:

a) Single time point global methylation status will be associated with incident CHD outcomes independent of TRFs.

b) Single time point methylation levels at age-related CpGs will be associated with the presence or the development of TRFs and with incident CHD.

c) Single time point levels of circulating miRNAs that have been previously associated with lipid metabolism, glucose homeostasis, and Type 2 diabetes will be associated with the presence or the development of either dyslipidemia, insulin resistance, β-cell dysfunction, impaired fasting glucose, and Type 2 diabetes, or a combination of these cardio metabolic risk factors.

d) Single time point levels of circulating miRNAs that have been previously associated with vascular smooth muscle cells and hypertensions will be associated with the presence or the development of either prehypertension or hypertension.

e) Single time point levels of circulating miRNAs that are cardiac muscle, endothelial, and vascular smooth muscle cell specific will be associated with incident CHD independent of traditional risk factors.

f) A single time point signature of methylation patterns at age-related CpG sites combined with circulating miRNA levels will provide incremental predictive value over standard CHD prediction models (e.g. Framingham risk score).

g) Certain elements of the genomic signature in (e) will perform better in the prediction of near term incident CHD events compared to longer-term events. For example, we hypothesize that miRNAs that rise in the setting of ACS will perform better in the prediction of near term events compared to longer-term events.

Specific Aim 3

In addition to standard marginal association methods that focus on one variable (e.g. CpG or miRNA) at a time, we will apply systems biologic (network) approaches to the data generated in Aim 1. Specifically, we will adapt a popular method we have developed named weighted correlation network analysis (WCGNA) to the data to identify CHD related co-methylation modules (pathways), miRNA modules, and key regulators of these modules. We hypothesize that both miRNA and methylation modules specific to CHD and related phenotypes exist and that some of these modules will show cross-omic correlations at the level of module eigengenes. These modules and their key drivers will be top candidates for experimental validation and further development with respect to therapeutics and biomarkers for CHD.

Specific Aim 4

We will revisit genome wide association studies (GWAS) of CHD and related phenotypes based on insights gained from our methylation and miRNA data. We hypothesize that hundreds of CHD and related phenotype CHD susceptibility polymorphisms remain undiscovered because of uncertainty created by testing a very large number of SNPs concurrently. We further hypothesize that selecting and testing small subsets of SNPs informed by our methylation and miRNA association and network analyses will robustly uncover at least some of these CHD susceptibility polymorphisms, genes, and/or pathways and possibly molecular sub phenotypes of CHD. We will test these hypotheses not only within the WHI GWAS but also in external independent datasets including CARDIoGRAM+C4D, GLGC, GIANT, and ICPB consortia, which respectively represent the largest GWAS for CHD, lipids, BMI, and blood pressure in the world. ​