Assessment of RNA-Seq Differential Expression Methods Based on TCGA Data

• Activity Number: 336 - Next- Generation Sequencing
• Type: Contributed
• Date/Time: Tuesday, August 1, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Guy Brock* and Meredith McCormack-Mager and Yu Lianbo and Soledad Fernandez
• Companies: Center for Biostatistics, The Ohio State University College of Medicine and Center for Biostatistics, The Ohio State University College of Medicine and Center for Biostatistics, The Ohio State University College of Medicine and Center for Biostatistics, The Ohio State University College of Medicine
• Keywords: RNA-seq ; TCGA ; false positives ; power ; differential expression

Abstract

Recent evidence has suggested that negative binomial based RNA-seq analysis methods may be too liberal and produce excessive false-positives, particularly for small sample sizes. In this work, we conducted novel simulations based on real data from The Cancer Genome Atlas (TCGA) to evaluate whether this occurrence resolves asymptotically. Negative binomial methods, including edgeR, DESeq, and baySeq, were compared to several normal distribution based methods including limma-voom, limma-trend, and the t-test. In addition to sample size, we also determined the extent to which mean gene expression was related to false positive rates and power. We believe our results will provide important guidance to practitioners regarding analysis of this data in the future.