Discovery of mRNA Isoforms Using Paired-End RNA-Seq Data

• Activity Number: 59
• Type: Topic Contributed
• Date/Time: Sunday, July 31, 2011 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Haiyan Huang*+
• Companies: University of California at Berkeley
• Keywords: RNA-seq ; mRNA Isoforms ; transcription ; expression ; lasso

Abstract

In transcriptomic study, mRNA isoform discovery is an important yet challenging topic. The recently developed RNA-Seq technology, with deep coverage and base level resolution, provides unprecedentedly large amount of genome-wide data to study mRNA transcription and alternative splicing, and hence enables a de novo discovery of mRNA isoforms for multi-exon genes.

In a recent work, taking the known exon annotations in literature, we developed a statistical method to discern the set of mRNA isoforms that are most likely to present in an RNA-seq sample. This method is based on a linear model with a design matrix modeling the generating probability of RNA-Seq data from different possible mRNA isoforms. To tackle the unidentifiability issue in the model, we applied a modified Lasso procedure for parameter estimation. Based on the same linear model, our method can also estimate the expression abundance for each identified isoform. Performance of our method is investigated by both simulation and real data examples, in comparison to several popular methods in literature.