Activity Number:
|
81
- Contributed Poster Presentations: Section on Statistics in Epidemiology
|
Type:
|
Contributed
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
Sponsor:
|
Section on Statistics in Genomics and Genetics
|
Abstract #312945
|
|
Title:
|
Leveraging Spatial Correlation to Improve Estimation of Cell Type Specific Methylation from Whole Blood
|
Author(s):
|
Richard Meier* and Jeffrey Thompson and Devin Koestler
|
Companies:
|
University of Kansas Medical Center and University of Kansas Medical Center and University of Kansas Medical Center
|
Keywords:
|
Bayesian;
hierarchical;
spatial;
methylation;
cell specific;
whole blood
|
Abstract:
|
Methylation of Cytosine-Guanine dinucleotides (CpGs) play an important role in the regulation of cellular processes and change in methylation has been linked to many human diseases. Cell type specific analysis of methylation levels in biological samples has shown promise to improve insight into the interplay of cellular processes and disease. Recent statistical modelling strategies now allow for these analyses to be performed based on bulk methylation data, without the necessity of isolating cells. Unfortunately, none of these approaches incorporate the knowledge that CpGs spatially correlate with base-pair distance. Here, we present a Bayesian hierarchical modelling strategy that leverages spatial correlation to improve estimation and statistical power when testing for differential methylation. Whole blood methylation data of isolated cell types is empirically evaluated to motivate the approach and further utilized in extensive simulation studies comparing benefits of candidate models. Our approach consistently increased prediction accuracy and statistical power compared to non-spatial models, suggesting it could improve efficiency of future cell specific methylation studies.
|
Authors who are presenting talks have a * after their name.