Three papers on new RNA-seq methods offer a new way to do RNA-seq analysis

It is fair to say that the new year has began with a bang for methods for RNA-seq analysis. Three new papers on new methods analyzing RNA-seq data have been published in the last month or so. And guess what, not so surprisingly all of the papers have Johns Hopkins University connection :-). yes, Steven Salzberg is part of all the papers with members from Ben Langsmead and Jeff Leek’s group.

The three papers kind of replaces earlier tools from Salzberg’s group (Bowtie/TopHat,Cufflinks, and Cuffmerge) the offer a totally new way to go from raw RNA-seq reads to differential expression analysis

  • align RNA-seq reads to genome (HISAT instead of Bowtie/TopHat, STAR),
  • assemble transcripts and estimate expression (StringTie instead of Cufflinks), and
  • perform differential expression analysis (Ballgown instead of Cuffmerge).

HISAT (Hierarchical Indexing for Spliced Alignment of Transcripts), is a fast splice-aware aligner with low memory footprint just came out in Nature Methods. An earlier version of manuscript has been on bioRxiv since December. The HISAT software is available here at

StringTie is a new fast and efficient reference based transcriptome assembler using RNA-Seq reads. StringTie uses network flow algorithm to assemble transcripts and it also estimates transcript expression. StringTie software is available at

Ballgown allows to do estimation of differential expression of genes, transcripts, or exons from RNA seq experiments and more. Ballgown has out been out there for a while and its preprint was first available in bioRxiv last year. Ballgown software is available at


  1. Elin Videvall says:

    What would you say is the advantage of using these three in comparison to the old workflow?

    • Elin – there are many differences, but one of the first things you’ll notice is that each of the programs is many times faster than the one it replaces/updates. So your workflow should run far faster – maybe 10 times faster or even more for large data sets. Plus we are finding that the new methods are more accurate – significantly more so, in some cases.

  2. Mikael Huss says:

    If I may comment on that, I like Ballgown because it works on assembled transcripts (so it includes unannotated splice variants and novel transcripts in general), includes functionality for testing differential expression of these transcripts *with covariates* (which CuffDiff doesn’t) and on time series, and also it provides a nice “back end” to a set of assemblies where you don’t have to fiddle around with your own scripts to match contigs in different samples to each other. (Haven’t tried HISAT or StringTie)

Speak Your Mind