Joint Sparse Modeling of Multiple RNA-Seq Samples for mRNA Isoform Discovery and Abundance Estimation

• Activity Number: 453
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: WNAR
• Author(s): Jingyi Jessica Li*+ and Peter J. Bickel and Haiyan Huang and Shihua Zhang
• Companies: University of California, Los Angeles and University of California, Berkeley and and Chinese Academy of Science
• Keywords: RNA-Seq ; joint modeling ; sparse estimation ; hierarchical modeling ; lasso ; isoform discovery

Abstract

Since the inception of next-generation mRNA sequencing (RNA-Seq) technology, various attempts have been made to utilize RNA-Seq data in assembling full-length mRNA isoforms and estimating the isoform abundance. The problem is challenging and often involves identifiability issues in statistical modeling. We have developed a statistical method called "Sparse Linear modeling of RNA-Seq data for Isoform Discovery and abundance Estimation" (SLIDE) that takes exon boundaries and RNA-Seq data as input to discern the set of mRNA isoforms most likely presenting in an RNA-Seq sample (Li et al. (2011) PNAS 108(50), pp.19867-72). The published version of SLIDE takes one RNA-Seq dataset as input. As the cost of RNA-Seq decreases, biologists tend to produce more biological or technical replicates, in order to reduce possible noises in a single RNA-Seq experiment. In this talk, we propose a hierarchical model that extends our SLIDE work to effectively use publicly available RNA-Seq replicates, so that the discovery and quantification of mRNA isoforms can achieve better accuracy than using a single RNA-Seq dataset.