1. How to cluster transcripts from multiple query gtf files? 2. Something wired about tss_id

[dudududu12138]

Hi, I used gffcompare to merge multiple assemble results from different samples. The code is listed below:

gffcompare -T -S $gffcompare_inputs -o $output 

I didn't offer a reference annotation cause I just want to merge assemble results from different samples.But I got two questions.

Question 1.

How does gffcompare cluster multiple samples without a reference annotation? There is only description of duplication removing in your manual[https://ccb.jhu.edu/software/stringtie/gffcompare.shtml] and your paper. In my results, I saw some transcripts with contained_in tag. You said you want to remain alternative start site. But there are some transcripts with the same tss_id and one of them with contained_in tag. It is so wired. Below is an example:

Question 2.

After I found the problem above, I checked the tss_id in my results. I followed 3 steps:

• step1: Group transcripts with tss_id

• step2: Calculate the max distance of start site among transcripts with the same tss_id

• step3: Get the distribution of distance

Below are 2 figures of my checking results. Although your default setting of parameter -d is 100, there are still transcripts using the same tss_id dist more than 100bp.
Below is a pie plot and Category means max distance=0, >0 & <=100 and >100

Below is a histogram of distance, I removed distance=0 firstly.

Thank you so much to solve my question and looking forward to your reply!

[dudududu12138]

Hi, I have figured out why the two transcripts have the same tssid, but the transcription start sites are far apart. You set a cutoff to cluster transcript start site(-d with default 100). But that only applies to positive strand genes, negative strand genes don't adhere to that standard.