tophat-fusion-post 0 fusions are output in ./tophatfusion_out/potential_fusion.txt

[RocheKermit]

Dear All,

I would like to get tophat-fusion-post working, apparently as many users there... This thread aiming at the lucky ones that have managed to make tophat-fusion-post working. Could you please share the tips and tricks that appear here and there in the different threads in order to have to good pratices?
Ie: is it mandatory to avoid symlinks?
The blast db must in the Top_dir with the name "blast" or "blast_human"?
Which versions of Tophat/bowtie/samtools...should be used/avoid/locally compiled?
Is there a way to concatenate a positive fusion hit in the fusions.out file with synthetic data in order to distinguish biological absence from computational digressions?
Should fusions.out contain a syntaxe like chr1-chr10 or rather 1-10?
Have some of you modified the perl code?
...and so on...
Please feel free to stack any good advices to the community and make that interesting tool free to work as happily it could be.
Many thanks and nice fusioning!

[yingzhang]

I think tophat-fusion-post is written in python.

Here is my answers to part of your questions.

1. Symbolic link. Yes, I used symbolic links for blast database.
2. Since it is symbolic link, I made two directories, one for blast, one for blast_human.
3. I used tophat 2.0.6, bowtie 0.12.8, samtools 0.1.18. The only confusion is the ncbi_blast. I have both blast and blast+ installed. If you only have ncbi blast installed, I think if you make an alias for "blastn='blastall -p blastn'", it should work too.
4. I never took the effort to add a positive case in the fusion.out file.
5. In fusion.out, the format is chr1-chr1
6. I thought about modifying the codes, however, the script worked ...

My only lesson for tophat-fusion-post is that it used a lot of memory.

Here is some stats. I run tophat-fusion on 60G 2X100 bp reads. The successful job used 32G memory. Post-processing of the result (using tophat-fusion-post) requires 64G memory. To be more precise, the tophat-fusion job failed on 16G memory, and the tophat-fusion-post failed on 32G memory.

Again, I think tophat/2.0+ has a big issue of memory leaking. I emailed the developer earlier this year, and never heard from them on this issue.