Item

Abstract

As a fruit of the current revolution in sequencing technology, transcriptomes can now be analysed at an unprecedented level of detail. These technological advances have been exploited in diverse ways. Examples of such include the detection of differentially expressed genes across biological samples, and the quantification of the abundances of various RNA transcripts within single genes. A natural next step is now to extend the detection of differential abundance, focusing on individual transcripts within one gene. However, robust strategies to solve this problem have not yet been defined. In this work, we present two novel statistical tests to address this important methodological gap: a ‘gene-structure-sensitive’ Negative-Binomial (NB) test that can be used to detect differential transcript expression when the gene structure is known and a non-parametric kernel-based test, called Maximum Mean Discrepancy (MMD), for cases when the gene structure is incomplete or unknown. Both methods can account for biological variance and cope with replicates. Our analysis shows that both MMD and the NB-test identifies genes with differential transcript expression considerably better than approaches based on transcript quantification, such as rQuant and Cuffdiff. In this work we furthermore present some applications of the before mentioned methods and show some common pitfalls of RNA-Seq experiments. Here, we proposed two novel approaches to test for differential expression on the level of transcripts. While the NB-test exploits existing transcript annotation, the MMD test can be used in settings where details about transcripts are not available. On simulated reads and investigation of real data, we achieved promising results using these methods, highlighting their potential as discovery and statistical testing tool. Furthermore we show applications of those methods and point to some common problems of RNA-Seq experiments.