How To Merge Gff Files

Question

[upendrakumar.devisetty]

Hi, I was wondering if anyone know how to merge 2 or more gff files to make a consensus gff file? I have gtf file from tophat/cufflinks pipeline gff file from velvet assembly and i would like to merge these two to the already annotated gff file. The idea is to have one gff file instead of three gff files so that i can load this as track on my Genome Browser. Thanks in advance

[Sean Davis]

Any reason to not simply load three separate files? Alternatively, you can simply concatenate them and sort by chromosome location.

[upendrakumar.devisetty]

No particular reason..I want to all three files separately along with a fourth file indicating the consensus annotation which users might feel more convenient dealing with.

[cmdcolin]

Programs such as MAKER combine evidence for gene models

Answer 1

[Alex Reynolds]

One way to do this is to start by converting them to sorted UCSC BED with BEDOPS gtf2bed (link) and sort-bed (link). Here's a quick way to convert a bunch of files if you use a bash shell:

$ for i in `ls *.gtf`; \ 
do gtf2bed < $i > $i.converted.bed; \ 
done;

Then do a multiset union set operation with BEDOPS bedops (link) to make a single BED file called answer.bed that can be loaded into your genome browser instance:

$ bedops --everything *.converted.bed | cut -f1-6 - > answer.bed

[Bioinfonext]

Hi Alex,

I have done genome guided assembly using StringTie and it also generated a gff file. I also have already reported CDS gff. I want to merge both gff and want to extract consensus sequences from genome.

Answer 2

[O.rka]

I know you probably figured this out since it was asked 6 years ago but I ended up using gffcompare in 2019 (for anyone else looking for this): https://ccb.jhu.edu/software/stringtie/gffcompare.shtml

It worked really well.

For installation:

conda create --name gffcompare_env -c bioconda gffcompare -y

(I usually create a separate environment for all my tools to keep things clean. )

The above tool takes in a file with paths to gtf files and then you just specify an output, Se the link for more information on the usage.

[Jan Röslein]

Why does it remove all features except "exon" and "transcript" from gff3 output? Doesn't make any sense.

[epaminonda]

I have also found that gffcompare removes everything except 'exon' and 'transcript'. Has anyone found out why? Any alternative tools for doing this?

[Juke34]

You can try agat_sp_merge_annotations.pl from AGAT

[Juke34]

How does it deal with overlapping loci? Merging them as a single locus?