StringTie or FeatureCounts

Hi! Would it be better to use FeatureCounts or StingTie for generating count tables in a RNA-Seq analysis?
Galaxy tutorials recommend FeatureCounts, but I’ve also read that StringTie is very efficient and widely used.
I’m working with Arabidopsis thaliana so I’m not really interested in novel transcripts. I’d just like to know what is the difference between these tools and which one should I use, since I get better variances in the PCA plot using tables from FeatureCounts than using tables from StringTie from the exact same data.

1 Like

Hi @MarioG

If you don’t care about novel transcripts and have a reference genome, go straight to Featurecounts and include a known annotation GTF.

Stringtie would include novel transcripts/genes plus known annotation – but only if the known annotation was included. You can also adjust the settings in Stringtie to only output results based known annotation. If you don’t include known annotation, only counts based on the regions your read data represent (transcripts/genes) will generated – and that may be where the biggest differences are coming from.

So, there are a few different ways to do this and the results would not be expected to be exact between any two DE methods/pipelines based on distinct algorithms. Using differently created transcript/gene boundaries for the “annotation” to determine expression levels is also a factor.

Extended Help for Differential Expression Analysis Tools


1 Like

Thank you so much!
This really helps with my analysis!

1 Like