|
Activity Number:
|
209
- Statistical methods for genomic and epigenetic data analysis
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2021 : 1:30 PM to 3:20 PM
|
|
Sponsor:
|
Section on Statistics in Genomics and Genetics
|
|
Abstract #318131
|
|
|
Title:
|
A Novel Framework for the Identification of Reference DNA Methylation Libraries for Reference-Based Deconvolution of Cellular Mixtures
|
|
Author(s):
|
Shelby Bell-Glenn* and Devin C Koestler
|
|
Companies:
|
Univerisity of Kansas Medical Center Department of Biostatistics and Data Science and University of Kansas Medical Center
|
|
Keywords:
|
DNA Methylation;
Deconvolution;
Computational Algorithms
|
|
Abstract:
|
Reference-based deconvolution approaches provide accurate estimates of cellular mixtures. These methods leverage the cell-specificity of DNAm and a reference library of cell-specific DNAm measurements. Their accuracy depends highly on the specific CpGs comprising the reference library. Existing approaches for optimized library selection require a training data set with two components: DNAm profiles over a heterogenous cell population and gold-standard measurements of cell composition. The primary purpose of this project is to utilize a modified Dispersion Separability Criteria (DSC) to optimize library selection in the absence of training data sets. We repeat the following steps many times: (1) get a candidate set of cell-specific differentially methylated loci (DMLs), (2) sample DMLs from the candidate set, and (3) compute the modified-DSC. The candidate library with the largest modified-DSC is selected for subsequent deconvolution. The proposed method was compared to the existing deconvolution legacy method by obtaining predictions using publicly available datasets and then computing the RMSE and R^2. Our proposed method outperformed the legacy approach for fixed library sizes.
|
Authors who are presenting talks have a * after their name.
