|
Activity Number:
|
484
- Methods for High-Dimensional Data in Genetics and Genomics
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Biometrics Section
|
|
Abstract #312845
|
|
|
Title:
|
A Bayesian Model Averaging Approach for High-Throughput RNA-Seq Count Data (BMAseq) and Its Applications
|
|
Author(s):
|
Lingsong Meng* and Xiaoting Xing and Xi Kathy Zhou
|
|
Companies:
|
University of Florida and Sanofi and Weill Medical College of Cornell University
|
|
Keywords:
|
RNA-seq counts;
Bayesian model averaging;
voom-limma;
heterogeneous samples
|
|
Abstract:
|
Differential gene expression analysis methods are typically single model based where the detection power can be negatively affected by having heterogeneous samples due to increased bias when there is insufficient covariate adjustment and/or decreased efficiency with unnecessary adjustment. We developed a Bayesian model averaging approach, BMAseq, within the voom-limma framework to improve the analysis of RNA-seq count data obtained from heterogeneous observational samples or complex experiments. Extensive simulation studies were carried out to examine its performance versus voom-limma and edgeR based single model approaches. Our results demonstrated that BMAseq outperformed the single model approaches with higher sensitivity and lower FDR. The improvement was especially pronounced in the setting with correlated covariates and unbalanced study groups. We also applied BMAseq to the analysis of RNA-seq data from GTEx. Our analysis results showed that the identification of differentially expressed genes could be improved by more elaborate model space specification and the comprehensive investigation of phenotype and genotype relationship could be achieved within a coherent framework.
|
Authors who are presenting talks have a * after their name.
