STAR-Fusion for finding fusion transcripts in cancer transcriptomes

[Brian Haas]

Greetings all,

For those of you that are interested in finding fusion transcripts in cancer transcriptomes, I'd like to draw your attention to STAR-Fusion

https://github.com/STAR-Fusion/STAR-Fusion

This script takes the aberrant junction and spanning reads output from the STAR aligner and maps them to reference gene annotations (by default, using gencode.v19).  Candidate fusion transcripts are reported along with counts of fusion reads and spanning fragments.

Sensitivity is quite good, rivaling or exceeding that of other popular fusion prediction tools. For example, running STAR-Fusion on a set of fusion-containing cancer cell lines demonstrates excellent sensitivity for identifying known fusion genes based on rna-seq data (heatmap attached showing binary yes/no finding of known fusions).

However, there is no annotation-based or meta-data based filtering done on the output, so currently there will be many many false positives.  Downstream filtering will be addressed in the future, along with visualization tools and additional fusion-related resources.

best wishes,

~brian

[Alexander Dobin]

Hi Brian,

thanks a lot for the great software, this is an important functionality many users were asking for.

I will include links to it in the documentation and GitHub (as submodule).

Cheers

Alex

[Dan]

Hi Brian,

would be possible to add to the heatmap also the (grand) total number of predicted fusion genes for each tool? That would give an idea of specificity!

Cheers,

Dan

[Jason Byars]

Sorry, if this is the wrong group to post in, I didn't see a separate group for STAR-Fusion questions. I'm working with single end IonProton data and having a little trouble getting the most STAR-Fusion to work. I'm running the latest release and using the example settings for STAR. I always end up with a decipher read name error.

-parsing gunzip -c /root/STAR/STAR-Fusion/resources/gencode.v19.annotation.gtf.exons.gz |
-building interval tree for fast searching of gene overlaps
-mapping junction reads to genes
Error, cannot decipher read name 5Y5EU:07487:06731 as /1 or /2  at /root/STAR/STAR-Fusion/STAR-Fusion line 665, <__ANONIO__> line 2.

Is there an easy way to tell STAR-Fusion not to look for paired end reads and simply work off single end reads where each half maps to a different gene?

[Brian Haas]

Hi Jason,

The current version won't allow this, but I'm in the process of putting a new version together, and I'll include this capability in the upcoming release.  I'm hoping to have it ready by this weekend.

best,

~brian

[Jason Byars]

Excellent, thank you for getting back to me. I will be happy to test.

 

Jason

--
You received this message because you are subscribed to a topic in the Google Groups "rna-star" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/rna-star/QBHa20rVwaU/unsubscribe.
To unsubscribe from this group and all its topics, send an email to rna-star+u...@googlegroups.com.
Visit this group at http://groups.google.com/group/rna-star.

[Brian Haas]

Hi Jason,

Attached is a slightly modified version that should work with single-end reads.  

I have an improved version scheduled for later this weekend, but this should suffice for the meantime.

best,

~brian

[Jason Byars]

Just wanted to say thanks. The fix works. I’m currently trying to work out what filters makes sense for IonTorrent data with the latest build.

 

Jason

[Vanessa Barker]

I also have a question specifically regarding STAR-Fusion - after running STAR-Fusion on output from STAR, I am not receiving full results. The "star-fusion.discordant_spans_to_genes.txt" and "star-fusion.spanning_read_names" output files are empty and the Spanning Fragments column in the "star-fusion.fusion_candidates.txt" file is always 0. This problem did not occur with the test data, but has occurred with every other sample I've run. I looked into the junction files and it appears that column 10 of the junctions file is a little bit different in the test data than the output file from STAR; the output file I have matches the description given in the STAR manual: column 10 contains the read name. Could this discrepancy be causing the problem? Is there possibly an issue with the version of STAR I am using? I am running STAR 2.3.1

[Brian Haas]

Hi Vanessa,

You might need to run the very latest release of STAR.  If that doesn't solve it, please let us know.  

Best,

-Brian

(by iPhone)

On Jul 2, 2015, at 2:58 PM, Vanessa Barker <nessa...@gmail.com> wrote:

I also have a question specifically regarding STAR-Fusion - after running STAR-Fusion on output from STAR, I am not receiving full results. The "star-fusion.discordant_spans_to_genes.txt" and "star-fusion.spanning_read_names" output files are empty and the Spanning Fragments column in the "star-fusion.fusion_candidates.txt" file is always 0. This problem did not occur with the test data, but has occurred with every other sample I've run. I looked into the junction files and it appears that column 10 of the junctions file is a little bit different in the test data than the output file from STAR; the output file I have matches the description given in the STAR manual: column 10 contains the read name. Could this discrepancy be causing the problem? Is there possibly an issue with the version of STAR I am using? I am running STAR 2.3.1

--

[Richard Harvey]

Dear Brian,

Has the improved version been released yet?  I am also trying to use STAR-Fusion with non-paired reads of Ion data.  The results have been a bit mixed with version 0.8.0.

Thanks.

Rick

[Brian Haas]

Not much has changed since v0.8.0 but you could always pull the latest code straight from github via  'git clone --recursive'

I haven't done any testing with single-end reads nor ion torrent data... everything has centered around Illumina PE data for now.  I'm not surprised to hear that it might not be working well w/ torrent SE data. Hopefully it finds the obvious fusions, but will likely require a lot of filtering to remove all the false positives.

~b

--
You received this message because you are subscribed to a topic in the Google Groups "rna-star" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/rna-star/QBHa20rVwaU/unsubscribe.

To unsubscribe from this group and all its topics, send an email to rna-star+unsubscribe@googlegroups.com.
Visit this group at https://groups.google.com/group/rna-star.

--

--
Brian J. Haas
The Broad Institute
http://broadinstitute.org/~bhaas

 

[Richard Harvey]

Thanks.  I have pulled the most current version and it works quite well for identifying many of the known fusions (both with Illumina data treated as single reads and with some Ion data).  There is quite a bit of weeding through the false positives to identify the actual fusions, but that is certainly manageable.  I'll tinker with the filtering of the output to reduce those and will also take a more detailed look at how to optimize STAR-Fusion for single reads.  I just didn't want to invest a bunch of time into that if it had already been done by the experts!

Rick

On Thursday, December 8, 2016 at 2:05:01 PM UTC-7, Brian Haas wrote:

Not much has changed since v0.8.0 but you could always pull the latest code straight from github via  'git clone --recursive'

I haven't done any testing with single-end reads nor ion torrent data... everything has centered around Illumina PE data for now.  I'm not surprised to hear that it might not be working well w/ torrent SE data. Hopefully it finds the obvious fusions, but will likely require a lot of filtering to remove all the false positives.

~b

On Wed, Dec 7, 2016 at 5:18 PM, Richard Harvey <harve...@gmail.com> wrote:

Dear Brian,

Has the improved version been released yet?  I am also trying to use STAR-Fusion with non-paired reads of Ion data.  The results have been a bit mixed with version 0.8.0.

Thanks.

Rick

On Wednesday, June 10, 2015 at 7:52:26 PM UTC-6, Brian Haas wrote:

Hi Jason,

Attached is a slightly modified version that should work with single-end reads.  

I have an improved version scheduled for later this weekend, but this should suffice for the meantime.

best,

~brian

--
You received this message because you are subscribed to a topic in the Google Groups "rna-star" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/rna-star/QBHa20rVwaU/unsubscribe.

To unsubscribe from this group and all its topics, send an email to rna-star+u...@googlegroups.com.

Visit this group at https://groups.google.com/group/rna-star.

[Brian Haas]

sounds good.  I've personally done little to no work w/ single-end data.  If you want to share insights on SE-based fusion prediction using this, I'm happy to add relevant documentation and credit you for it.

best,

~brian

To unsubscribe from this group and all its topics, send an email to rna-star+unsubscribe@googlegroups.com.

Visit this group at https://groups.google.com/group/rna-star.