search

pachterlab / kallisto

Near-optimal RNA-Seq quantification
https://pachterlab.github.io/kallisto
BSD 2-Clause "Simplified" License
654 stars 172 forks source link

RNA-Seq gene fusion calling #122

Open schelhorn opened 8 years ago

schelhorn commented 8 years ago

Hello @lakigigar and co-developers,

we're currently running @chapmanb's bcbio in production for quantitating RNA-Seq expression and calling gene fusion events; thanks to @roryk's work bcbio supports the triad of hisat2, salmon, and kallisto for the former (to different degrees), and star in combination with oncofuse for the latter.

The issue with star, while being a great approach in itself, is it's huge appetite for memory due to the reliance on full genomic alignments, as well as its incompatibility with hg38, a fact that does not seem to change anytime soon. We would therefore like to have a look at the newer three approaches for both expression quantitation and RNA-Seq gene fusion calling. Unfortunately, none of hisat2, salmon, and kallisto seem to support calling gene fusions explicitly at the moment. While some of these methods support generating BAM files with pseudo-alignments that may be usable by dedicated gene fusion calling pipelines, no such approach has been documented where it can be found easily (i.e., by googling the relevant terms).

To accelerate usage of fast hg38 gene fusion pipelines we would be willing to spend developer time on integrating an existing approach into bcbio. In order to better determine which approach to use I have already raised the issue with salmon (discussion), and hisat2 (discussion) to get the conversation going. kallisto seems to support gene fusion calling as well, but evidence is anecdotal. May I ask you to elaborate on your view of the issue, both in general and specifically as regards kallisto?

schelhorn commented 7 years ago

Sorry to bother, but would anyone be able to answer this question? Thank you.

DarioS commented 7 years ago

It appears that the software is no longer being maintained. There haven't been any bug fixes in almost a year, nor any replies to Issues tracker, either.

I'm sticking with STAR for now.

maximilianh commented 7 years ago

How about heading over to Salmon? The Salmon developers are extremely responsive.

maximilianh commented 7 years ago

You can see lots of active discussions here: https://github.com/COMBINE-lab/salmon/issues

pmelsted commented 7 years ago

Hi @schelhorn, sorry for the late reply. We are working on fusion calling directly in kallisto and will be releasing the code in the coming weeks.

As far as bug fixes go, most of them have been addressed and will go into the next release. The lastest one fixed was upstream in the kseq library https://github.com/attractivechaos/klib/issues/78

schelhorn commented 7 years ago

Excellent news, thank you. Will kallisto stick with the closed commercial license, or will you open that up at some point? I'd love to use you tool for pharma research, but I don't see us purchasing a license for it anytime soon.

scchess commented 7 years ago

@schelhorn Kallisto is free for academics research as long as you're not making money. You don't need to ask. https://github.com/pachterlab/kallisto/blob/master/license.txt

schelhorn commented 7 years ago

So, things have changed. Lior has changed the license, and very gracefully. And you guys have made and released pizzly. Thanks for both.