Comparisons of computational methods for differential alternative splicing detection using RNA-seq in plant systems

摘要: Alternative Splicing (AS) as a post-transcription regulation mechanism is an important application of RNA-seq studies in eukaryotes. A number software and computational methods have been developed for detecting AS. Most the methods, however, are designed tested on animal data, such human mouse. Plants genes differ from those animals many ways, e.g., average intron size preferred AS types. These differences may require different approaches raise questions about their effectiveness plant data. The goal this paper to benchmark existing differential splicing (or transcription) detection so that biologists can choose most suitable tools accomplish goals. This study compares eight popular public available packages analysis using both simulated real Arabidopsis thaliana All freely available. examines effect varying ratio, read depth, dispersion pattern, types, sample sizes influence annotation. Using looks at consistences between verifies subset detected events PCR studies. No single method performs best all situations. accuracy annotation has major impact which should be chosen analysis. DEXSeq well data when signal relative strong accurate. Cufflinks achieve better tradeoff precision recall turns out one incomplete provided. Some perform inconsistently Complex combine several simple impose problems especially MATS. MATS stands being evaluated events.