Comparative evaluation of full-length isoform quantification from RNA-Seq

Full-length isoform quantification from RNA-Seq is a key goal in transcriptomics analyses and an area of active development. The fundamental difficulty stems from the fact that RNA transcripts are long, while RNA-Seq reads are typically short. We have generated realistic benchmarking data, and have performed a comprehensive comparative analysis of isoform quantification, including evaluating them on the level of differential expression analysis. Genome, transcriptome and pseudo alignment-based methods are included; and a naive approach is included to establish a baseline. Kallisto, RSEM, and Cufflinks exhibit the highest accuracy on idealized data, while on more realistic data they do not perform considerably better than the naive approach. We determine the effect of structural parameters, such as number of exons or number of isoforms, on accuracy. Overall, the tested methods show sufficient divergence from the truth to suggest that full-length isoform quantification should be employed selectively.