Stringtie de novo VS reference guided (-G)

Question

[vasanthmandla]

Hi Community,

I am currently preforming RNA-seq analysis of human dataset and my aim is to find novel transcripts and isoforms. I have aligned the sequences to the reference genome using Hisat2 and assembled the transcripts using Stringtie in both reference guided and de novo methods. When I looked at the number of assembled transcripts I see reference guided mode has assembled twice number of transcripts than de novo method:

stringtie denovo: exon 400375; transcript 59007

stringtie reference-guided (-G): exon 708407; transcript 121080

My question is: Is it normal to see this difference and if so could you please, help me in understanding the reason or refer to any article.

Thanks in advance.

Answer 1

[Istvan Albert]

The Stringtie manual states:

Although StringTie is primarily a genome-guided approach, it can borrow algorithmic techniques from de novo genome assembly to help with transcript assembly.

https://ccb.jhu.edu/software/stringtie/index.shtml?t=manual

The above implies that it will perform more effectively in guided mode, which is what you also observe. That being said what the difference should be will depend on a whole slew of properties of both your sequence data and the reference annotations.

I would recommend to investigate and compare the regions where one approach produces more transcripts than the other.

• What kind of commonalities can you observe.
• What kinds of tradeoffs can you observe
• Which assembly appears to be more trustworthy