Figure 4
From: Assessing the impact of human genome annotation choice on RNA-seq expression estimates
The average coefficient of variation varies with different annotations and gene or isoform subgroups. Sub-figures (a)-(d) represent different combinations of samples (top: MAQC samples with SRA accession number SRP000727; bottom: thrombin study samples with SRA accession number SRP008482) and expression estimates (left: TPM estimates from OSA package; right: FPKM estimates from TopHat alignment with Cufflinks quantification). The RefSeq annotation always has the smallest average coefficient of variation, while the AceView annotation has the highest average coefficient of variation for most of the cases. The variation is small when focusing on only common genes.