Simulation Study for Comparison and Evaluation of Differentially Expressed Gene-Detection Methods in mRNA-Seq

• Activity Number: 623
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Sunghee Oh*+ and Hongyu Zhao and James P. Noonan
• Companies: Yale University and Yale University and Yale University
• Keywords: mRNA-Seq ; differentially expressed(DE) gene detection ; count data ; simulation study ; poisson, negative binomial, binomial distribution ; small sample size, low read count, noisy set

Abstract

Even considerable technology advances in next generation sequencing, there are newly changeable statistical issues for count data. Biomarker gene lists from differentially expressed (DE) methods are still a crucial part in mRNA-seq. Some people have proposed DE gene detection methods for mapped read counts. However, there are no studies on the comparison and evaluation between DE methods in depth. Here I examine which DE method is more robust, especially under low read counts and small sample size by a simulation study in poisson, negative binomial, and binomial distribution. Two groups are compared with p=samples (p=2,3,5, 10) in each group using 8 DE methods. Overlap between different DE methods is examined by comparing the content of gene lists of top-ranking candidates. Besides, behaviors of DE methods for weakness such as sample size, low read counts, and noisy set are explored.