New: empirical splice junction finder from RNA-seq data

[DougB]

We have recently published and released source code for supersplat, a tool which empirically identifies all potential splice junctions in a genomic reference given RNA-seq data. Supersplat does not include bias for canonical intron-terminal dinucleotides and does not build exon islands on which to base its conclusions. Supersplat searches over the input short reads for those which can in any way be partitioned to indicate potential introns, thereby identifying every potential splice junction supported by empirical evidence. While this exhaustive approach can yield a massive amount of data, with careful filtering many splice junctions may be identified which would otherwise be missed.

Currently advanced published in bioinformatics (http://bioinformatics.oxfordjournals...reprint/btq206), supersplat source code and binaries can be downloaded from http://supersplat.cgrb.oregonstate.edu/.

[jeny]

Hi DougB,

I am interested in supersplat.
I run it with parameters -c 4 and -i 15, on a dataset of 21461960 unique reads and 3584 reference sequences (there are scaffolds).
superslpat processed only 6 references in ~3 days, it's to long.
What are your recomandations to speed up the job ?
Is it possible to run supersplat with multithread ?

Thanks for your reply.
Regards,

Jeny

[Lee Sam]

Hi Doug,

I'm playing with supersplat with some prostate cell line RNA-Seq data and I've got some questions like Jeny's. Do you have any suggestions as to how get good performance out of the application (which parameters outside of MICS affect performance the most)? Also, the program binary is provided as a ".bin" file - do I assume correctly that this is compiled as a x64 Linux binary?

[paulorapazote]

Hi everyone
I'm working with SuperSplat, and I want to convert the final output (.supersplat) to Sam tools format? Any idea how can I do that?

Cheers,

P