Comment on “TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions” by Kim et al.

Abstract

In the recent paper [1] (thereafter referred to as \"TopHat2paper\") the accuracy of TopHat2 was compared to other RNA-seq aligners. In this comment we re-examine most important analyses from the TopHat2paper and identify several deficiencies that significantly diminished performance of some of the aligners, including incorrect choice of mapping parameters, unfair comparison metrics, and unrealistic simulated data. Using STAR [2] as an exemplar, we demonstrate that correcting these deficiencies makes its accuracy equal or better than that of TopHat2. Furthermore, this exercise highlighted some serious issues with the TopHat2 algorithms, such as poor recall of alignments with a moderate (>3) number of mismatches, low sensitivity and high false discovery rate for splice junction detection, loss of precision for the realignment algorithm, and large number of false chimeric alignments. ...