--mm command line flag dramatically increases runtime about hisat2

[infphilo]

Thanks, especially given the detailed information!, which should help us a lot to shoot the problem.

Did you use the official version of HISAT2 1.0.1-beta, available on the HISAT2 website? If so, that version didn't work with --mm option. I fixed this issue later, with the fix included in the master branch, here.

Thanks,
Daehwan

[infphilo]

I just confirmed this was fixed in the GitHub master version - I'll try to release the next release sometime this week, thanks!

[tantrev]

I also replicate this problem in the latest hisat2 build on Mac OS X Mavericks (v10.9.5). Additionally, a different error occurs on linux systems (hisat2 terminates, rather than running indefinitely like on Mac OS X):

(ERR): hisat2-align died with signal 11 (SEGV)

when administered with the following command:

~/tools/hisat2-master/hisat2 -x /mnt/ramdisk/Mus_musculus/grcm38_snp_tran/genome_snp_tran -1 input_1.fastq -2 input_2.fastq -S output_mm.sam --mm

Perhaps my command notation is incorrect?

[tantrev]

I think the problem may be with the current "grcm38_snp_tran" index on the hisat2 website. I just tried building my own index with transcripts (no SNP data) using the following command, but hisat2 now appears to execute much slower compared to the previous index on the hisat2 site:

../hisat2-2.0.2-beta/hisat2-build --ss ../splicesites.txt --exon ../exons.txt -p 28 ../Mus_musculus.GRCm38.dna.primary_assembly.fa mus_base

I used the following files from ensembl:
Primary genome - ftp://ftp.ensembl.org/pub/release-84/fasta/mus_musculus/dna/Mus_musculus.GRCm38.dna.primary_assembly.fa.gz
GTF file - ftp://ftp.ensembl.org/pub/release-84/gtf/mus_musculus/Mus_musculus.GRCm38.84.gtf.gz

The machine I used for the index building had 208GB of RAM - I suspect that should be enough to facilitate optimal '--dcv' values? Is there any other secret sauce to the reference indices?

[infphilo]

I just found out this is an issue in index building. There is nothing wrong with --mm option. I'll try to rebuild grcm38_snp_tran index. The other mouse indexes are fine.

[tantrev]

For what it's worth, I just tried running hisat2 with indices that I built using the parameters above and the "--mm" command, and the following error resulted on a Linux box:

Error while flushing and closing output
terminate called after throwing an instance of 'int'
(ERR): hisat2-align died with signal 6 (ABRT)

Parallel execution seems to work fine though.

UPDATE: Nevermind, sorry - I was just being stupid and had ran out of storage space. Everything works perfectly with the home-rolled index.

[infphilo]

I fixed this issue in the HISAT2 master branch and rebuilt genome_snp_tran indexes for the human and the mouse genomes.