Assessing Models of RNA-Sequencing Data

• Activity Number: 194
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Yanming Di*+ and Gu Mi and Sarah Emerson and Daniel Schafer
• Companies: Oregon State University and Oregon State University and Oregon State University and Oregon State University
• Keywords: RNA-Seq ; power ; robustness ; goodness-of-fit ; sample size

Abstract

The negative binomial (NB) distribution and NB regression models have been shown to be useful for modeling frequencies of mapped RNA-Seq reads from complex experiments. The NB distribution uses a dispersion parameter to capture the extra-Poisson variation commonly-observed in RNA-Seq experiments. Due to the typically small sample sizes in RNA-Seq experiments, significant statistical power can be saved by modeling the dispersion parameter as a smooth function of the expression level and thus effectively pooling information across genes/transcripts. We clarify various modeling and inferential choices for maximizing the information obtained from a given RNA-Seq experiment, discuss methods for assessing model adequacy, clarify the tradeoffs between power and robustness, and provide guidelines on the sample size needed to adequately address scientific questions.