Dear All,
I am trying to check for isoform diversity inside each transcriptome from RNA-Seq data.
One way could be to use junction.bed output from TpHat2 in DEXseq. Did anyone try this before. I need a way to convert junction.bed file to GTF that can be used in DEXSeq.
Many thanks in advance for your suggestion,
Rahel
There are quite a few posts around to convert bed to gtf. If you convert the junctions file to gtf, that means that the width of your junctions would be classed as exons? So when you run it through DEXSeq, you'll just be looking at reads that are mapping to the junction gap, which doesn't really make sense.
Dear Andrew,
Thanks for your reply. You are right, I could find some script to convert bed to gtf. But as I said I am looking for isoform diversity and I want to have the intron, exon ratio for each transcript. I did get exon and also intron differentially quantification using DEXSeq. My idea by using junction.bed in DEXSeq was to collect as much as information I can get from RNAseq data. Do you know any other packages in R that can help me? I am also currently running QuasR on my dataset to see if I can have reasonable result.
Kind Regards,
Rahel
If you want to look at intron usage, then essentially what you're doing will work. - Beware that the junctions.bed file from tophat does identify novel junctions, but there's a parameter to make it return only known junctions - --no-novel-juncs
Another option to look at is bedtools complement
Dear Mat,
I am also looking for intron retention. For that I used a script to have only the intronic-part in a gtf file then I used DEXSeq on this parts to get the differentially expressed intronic parts in my data set. Do you know any other way for this if I am doing wrong?
Kind Regards,
Rahel
Hello Rahel,
regarding the intron retention. I did the same. I created the skip or aggregated gff file with the DEXSeq script dexseq_prepare_annotation which I used afterwards to create an intron gff file. But here I'm using only unique intron side which means I don't look at chromosomal position that are overlapping with exons because I simply wouldn't know if reads are in this area intron retention or exonic. There was a thread on seqanswer (http://seqanswers.com/forums/showthread.php?t=42420) and they are doing the same.
Now back to the junction.bed. I'll start playing around with it tomorrow and I think I have an idea. At first categorise the junction by transcripts and by gene. Then I remove unknown junction and just look at known one. With that information, it should be possible to build a gff file. What I don't take into account at the moment, are overlapping genes which are on the same strand. Right know, junction which are overlapping with those genes are counted in both of them.
best
Mathias
Hi Mat,
Thanks for your reply. I did use the same thread on seqanswer and it was so helpful. I did DEXseq once only with intron bins and once with intron+exons. At the moment I am looking at the result in IGV to make sure the results is giving the exact intron retention.
As I did DEXseq with 14 different condition, at the moment I am trying to build a nice heatmap with the current DEXseq result as DESEQ2 does (hopefully!) and afterwards I will start with Junctions.bed conversion.
Please let me know if you get to work with gff file. I am biginner in this field and everything take more time for me to get through.
Many thanks in advance,
Rahel
hi Rahel,
can you see a difference in intron bins vs intron+exons. I assume you get less introns which show different usage in the intron+exons. If I remember correctly dexseq uses the same independent filtering like deseq and therefore in intron+exons, introns should fall more into the background noise due to low number of reads. What do you see?
best
mathias
Hi Mathias,
Yes, You are completely right. When I check the results from intron+exon. I loose the highly differentiated intron out of the analysis only with interon. I am going to check them in IGV and if those introns are located exactly in the middle of intron position, I am going to continue my work with only intron analysis.
At the moment the server from my work is on maintenance. I have no access to the .bam files to check the result on IGV (That is why also I could not work on junction.bed file). When I check the result in detail I will write you back.
How is the .gff file progress for junction.bed? Could you manage it?
Kind Regards,
Rahel
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
Hey rahel,
I'm doing more or less the same at the moment. I created unique IDs for the junctions and changed them in all the bed files. Converting the bed/splice site to gff is not the problem what I'm wondering is how you relate the junctions to each of their transcripts. That's an information which DEXSeq needs. Do you do bedtools intersect with the junction bed and an annotation file?
Hi Mat,
I also found some way to convert .bed to gtf but DEXSEq need to have a gtf which it is linked between junctions to parent genes.... This is what I am also looking for. I know some people did it. but so far, I could not find any thing ...
Hi Rahel,
Sorry, I was away for while. Have you any new on the linking of junction. I'll start working on it tomorrow again.