Genomic and epigenomic prediction of cardiovascular risk and modulation by diet.
Westerman, Kenneth.
2019
-
Personalized dietary
approaches to cardiovascular disease (CVD) prevention will require sensitive biomarkers
of multiple types. First, direct CVD risk biomarkers will enable the prioritization of
dietary interventions for specific individuals. Second, biomarkers of diet response will
allow more effective dietary recommendations for maintaining optimal cardiovascular risk
factor (CRF) levels. ... read moreBoth genomic variants and epigenomic factors such as DNA
methylation have the potential to enable improvements in both of these biomarker types.
Aim 1 of this work involved an investigation of mechanistic links between DNA
methylation and incident CVD events. Differentially-methylated regions (DMRs) and the
WGCNA module-based clustering approach were used to derive robust and interpretable
epigenomic features before testing associations with CVD in the Women's Health
Initiative and Framingham Heart Study Offspring Cohort. Multiple novel signals were
uncovered, including differential methylation near immune function-related genes
(SLC1A5, SLC9A1, and TNRC6C) and of development- and immune-related epigenomic modules.
Module methylation was also linked to cumulative CRF exposure, providing
proof-of-concept for the use of DNA methylation as a molecular recorder of cumulative
risk factors over the life course. Aim 2 expanded on these findings in seeking to
directly optimize a predictor of CVD risk. Using time-to-CVD-event elastic net
regression models, a methylation-based CVD risk score (MRS) was able to add predictive
power after adjustment for traditional risk factors. A recently-introduced ensemble
method for training predictive models in omic datasets showed comparable performance to
a standard analysis pipeline. A preliminary analysis of interactions between the MRS and
other CVD risk metrics (based on traditional risk factors or genomics) found that the
MRS appears to perform better in lower-risk regions of these alternative metrics,
suggesting its value in identifying individuals with "hidden" risk. In Aim 3,
genome-wide interaction study (GWIS) results were aggregated into genome-wide genetic
scores for response to dietary fat for six CRFs. When testing these scores in the fat
reduction-focused Dietary Modification trial from the Women's Health Initiative, only
one of six scores (corresponding to LDL-cholesterol) showed meaningful prediction of
one-year changes in the corresponding CRF, explaining a small proportion of the
associated variance. Further longitudinal investigation of the LDL-FRS revealed
suggestive interaction patterns between control and fat reduction arms for chronic
disease outcomes, including support for a known divergence in LDL-C impact on different
stroke subtypes. An additional preliminary study from Aim 3 explored the epigenomic
signature of response to vitamin K intake. After categorizing vitamin K responders and
non-responders based on their changes in plasma phylloquinone during a supplementation
trial, an epigenome-wide association study was performed to compare these groups.
Differential DNA methylation was identified at known and novel loci including the NPC1L1
gene, and was supported by a second EWAS for baseline plasma phylloquinone
concentrations. Taken together, these results build on and provide a more robust
foundation for the development of improved omics-based biomarkers for CVD risk
prediction and diet-centered risk reduction.
Thesis (Ph.D.)--Tufts University, 2019.
Submitted to the Dept. of Biochemical and Molecular Nutrition.
Advisor: José Ordovás.
Committee: Dawn DeMeo, Paola Sebastiani, and Paul Jacques.
Keywords: Bioinformatics, Nutrition, and Epidemiology.read less - ID:
- f7623r792
- To Cite:
- TARC Citation Guide EndNote
