RNA-Seq Gene Profiling - A Systematic Empirical Comparison

• DOI: 10.1371/journal.pone.0107026
• Publication Date: 2014-09-30T14:10:19Z
• Authors: Nuno A. Fonseca; John Marioni; Alvis Brazma
• Keywords: RNA-Seq; Gene prediction; Expression (computer science)

Accurately quantifying gene expression levels is a key goal of experiments using RNA-sequencing to assay the transcriptome. This typically requires aligning short reads generated genome or transcriptome before pre-defined sets genes. Differences in alignment/quantification tools can have major effect upon found with important consequences for biological interpretation. Here we address two main issues: do different analysis pipelines affect inferred from RNA-seq data? And, how close are "true" levels? We evaluate fifty profiling experimental and simulated data characteristics (e.g, read length sequencing depth). In absence knowledge 'ground truth' real RNAseq sets, used assess differences between those reconstructed by pipelines. Even though this approach does not take into account all known biases present data, it still allows estimate accuracy values The results show that i) overall there high correlation best true quantification values; ii) error estimated vary considerably across genes; iii) small set genes estimates consistently (across methods). Finally, although mapping software important, method makes greater difference results.