What's the difference of FPKM generated by htseq and stringtie?

Question

[hxlei613]

Hi I get FPKM for gene (not transcript) from stringtie by these parameter: -e -G -A. I also use htseq (default parameter )to generate counts matrix for the same gene sets and calculate fpkm using R. The gtf I used in this two method is the same. However the result fpkm is not the same. It seems that the overall fpkm generated by htseq is lower than stingtie. I try to use stringtie to get counts matrix for each gene but stringtie seems not generate count for gene level. I have searched the difference between them but I didn't find any answers. Thank you very much for helping me!

Answer 1

[Devon Ryan]

StringTie is able to use multimappers, which htseq will ignore them. That is the most probably cause for the difference you're seeing.

[hxlei613]

That is helpful. However I also find some gene whose counts are not zero in all samples and the fpkm I calculate is not zero (for example the average fpkm of such genes can be 300 or 3000) but fpkm generated by stringtie is zero. What's possible reason for that?

[Devon Ryan]

Without further information it's impossible for anyone to know.