I've been holding back because the stats software to interpret differential gene expression from the bootstrapped values has been slowly evolving. In June a preprint about this analysis software, called Sleuth, was put up on bioRXiv. So it was probably time I checked it out more thoroughly.
In particular I was drawn by microdissected tissue RNASeq experiments where DESeq2 was only showing very modest differential expression of genes, but a lot of it (1000's of genes). In theory, Sleuth should work better at resolving splicing variation in genes which might be going on, or even maybe compensating for tissue impurity insofar as it causes systematic underreporting of differential expression ratios.
Here's a brief rundown on running a combined Kallisto + Sleuth analysis.
Create a Kallisto searchable database of transcripts for your species, if you don't already have one. Personally I like RefSeq, but you could use Ensembl transcripts, etc. Here's how to get the RefSeq fastas, while stripping the GIs and leaving the NM_ or NR_ ID as the one Kallisto will report.
Now let's run some differential expression analysis. First, create a tab delimited metadata file for the experiment so we have the factors for each sample ID (header "sample"), and the file path (header "path") to the kallisto output for each sample. The file should look something like this:
sample path time
mysample1 mysample1.kallisto D0
mysample2 mysample2.kallisto D0
Note that the tutorial on the Sleuth Web site uses a somewhat convoluted method to get the right metadata table together. Here, I've simplified it, assuming you are running R from the directory where all the kallisto quant output directories reside.
If you've had the patience to read this far and understand Beta, why don't we look at some pretty graphs to see what our data look like (e.g. segregation on a Principle Component Analysis plot, volcano plots, etc.). Sleuth uses a Web interface called Shiny to allow interactive plotting, and you went through the installation instructions for both Kallisto and Sleuth you're set to go, right? The following will launch the Web interface, or at least give you a URL to paste into your browser.