Change in the way exon features from GTFs are displayed?

[John B.]

I'm trying to figure out why certain exons are no longer being connected to one another in the current version of IGV. Previously, exon features that had the same transcript ID would be connected together, but this does not seem to be the case any longer. It looks like it leads to loading of the exons in an non-intuitive way as well. Here is a screenshot highlighting some issues.

I'm showing a subset of a custom gtf file made utilizing cufflinks. In older screenshots I took of IGV sessions a few months back, the exons would be connected as they were from the same transcript and had the same transcript id and gene id. However, that does not appear to be the case. This also leads to unintuitive listing of the exons, as I then show in the next track of the UCSC gtf file I have. The three regions highlighted in yellow come from the same transcript, but are not aligned to each other. The green box shows a case where the exons were correctly connected. This transcript also has listed CDS features in the gtf file. That appears to be what tells IGV to link the exons together. In previous versions though, even without a CDS feature, the exons from the same transcript would be linked together. Is this something new with GTF2.2 standard or is there something else that might be going on? I'm using IGV 2.3.14. 

Thanks for any help you can provide.

[Jim Robinson]

Thanks for the report.  Could you attach or send us a small sample gtf illustrating the poblem?   You can email it to igv-tream (at) broadinstitute.org if you prefer.

[John B.]

I've attached a small gtf of the 5 exons that are from the same transcript but do not show up as connected. I noticed that if I change their feature column to "CDS" from "exon" that they become connected in IGV. However, in the past when I've used IGV, exon features were connected when their transcript ids matched up. 

[Jim Robinson]

I see the problem, its a bug in IGV. Thanks for the test file.

Recently we made changes to the GFF/GTF parser to improve compliance
with the GFF3 spec. If this were a GFF file the exons would be orphaned
because not transcript is defined in the file. We overlooked the fact
that in GTF files the transcript is implied, and never explicitly
defined as a record. I will correct it for the next bug-fix release,
probably the end of this week. The fix will also be included in
tonight's nightly build if you need it sooner, see the links on the
downloads page to use the nightly build.

Jim

[John B.]

Thank you for the quick response! I'm glad I could help as this definitely had me scratching my head.

[John B.]

Has this been updated? I've downloaded the nightly build zip file and ran the .bat file and the problem still seems to be there. 

Thanks for any help!

[Jim Robinson]

Yes it has been fixed, and the test file you provided looks normal to me (see screenshot). 

[John B.]

I see that when I launch it from the website using the .jnlp file, that it appears fixed, however when I run IGV (on either a linux cluster or on my local windows machine) using the following: http://www.broadinstitute.org/igv/projects/downloads/snapshot/IGV_snapshot.zip, it does not appear fixed. Is there a difference in the Binary distribution archive? 

Thanks,

John

[Jim Robinson]

John, apologies, the zip had not been updated. I've corrected it, now
they should be in sync.

Jim