Abstract:
|
RNA-Seq has revolutionized the field of molecular biology thanks to its superiority in dynamic range and improved resolution. A series of new research possibilities have been enabled by the application of this technology. One of them is detecting compositional differences in transcripts generated from the same genetic locus under different conditions, i.e., differential alternative splicing (DAS). Available statistical models for discovering and quantifying DAS events from RNA-Seq data suffer from some known issues. The MCMC-based Bayesian method is inefficient computationally, while lacking power as a result of considering only junction reads. Furthermore, the likelihood-based methods rely on the assumption of indepdence, which limits their applicability in experiments with matched samples or repeated measurements. A novel approach is proposed to model the difference in the exon-inclusion levels based on Hotelling's T-squared distribution, with consideration of possible correlations between the paired samples on the transcript level. In addition, a p-value aggregation algorithm inspired by Fisher's method is also implemented to generate p-values on the gene level.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.