Hi Galaxy, I am trying to install HISAT2 from the toolshed to my Cloudman instance (revision da2cc588e5) and not succeeding.
When I try to install HISAT2 from the toolshed one of the packages gets hung up (I think 'ncurses'), but if I go back one page and re-click the install button enough times I can get HISAT2 to the 'installed' state and it shows up in my tool panel. I can then install data managers and get the appropriate HISAT2 index installed. When I try running HISAT2 on some small sample data I get the following error:
Fatal error: Matched on hisat2-align exited with value 1
Error reading _plen[] array: 0, 17179869180
Error: Encountered internal HISAT exception (#1)
Command: /mnt/galaxy/tools/hisat/2.0/iuc/package_hisat_2_0/c65f00072e57/bin/hisat2-align-s --wrapper basic-0 -p 1 -x /mnt/galaxyIndices/hg19/hisat2_index/hg19ID/hg19ID -k 5 --n-ceil C,0,0 --pen-cansplice 0 --pen-noncansplice 3 --pen-canintronlen C,0,0 --pen-noncanintronlen C,-8,1 --min-intronlen 20 --max-intronlen 500000 --rna-strandness R -1 /mnt/galaxy/files/003/dataset_3659.dat -2 /mnt/galaxy/files/003/dataset_3657.dat
(ERR): hisat2-align exited with value 1
[samopen] no @SQ lines in the header.
[sam_read1] missing header? Abort!
[bam_header_read] EOF marker is absent. The input is probably truncated.
I understand the way I am installing HISAT2 from the toolshed is not optimal. If anyone has had success installing and using HISAT2 on Galaxy Cloumdman could you share your protocol?
Mo do you have the lncurses error for me. I do not have cloudman :(
Hi Bjoern, I don't see an error upon loading HISAT2 from the toolshed, but I do see that I have three version of 'ncurses packages' in my repositories on Cloudman (one at version 6.0 and two at version 5.9).
The latest version of Cloudman (revision: 5a50a4691c) has HISAT2 pre-installed. One still has to create the HISAT2 index using Data Managers.
When I try running a small test data set with HISAT2, the job goes green and returns an empty file with the message: "Could not open index file /mnt/galaxyIndices/mm9/hisat2_index/mm9/mm9.2.ht2 (ERR): hisat2-align died with signal 11 (SEGV) (core dumped)"
The same data runs successfully with Tophat (or even HISAT2 on Galaxy-test).
Mo - just checking - SAM indexes and Picard indexes already exist (created by DM). Jobs using these indexes are successful? As well as Tophat2/Bowtie indexes (creation) & jobs? All used the exact same dbkey and it exists in the builds list? You can check in the Admin loc files (view mode, under where the DMs are listed out in admin area pre-job launch).
Also - just because so many issues come up because of this - does the source fasta have long identifiers or contain non-alphanumeric chars (most should not be a problem but worth a test), duplicate identifiers (unlikely, but ..), and does not contain description content (no content after the first white-space in the index line ">"). Tools ideally would strip this but all do not and it is known issue when Custom genomes contain description content, odd errors can pop out, mostly seeming to be unrelated.
If want to share the target genome (load into Test, just in a history), I can test it out here too to see if can reproduce. Can load whole thing and also your test case. Share with me (jen@bx.psu.edu) directly and email or ping here so I know when to look for it. Can expand to test on cloudman test instance if needed.
Thanks! Between Bjoern, others, and I can hopefull sort out and get this going for you. My tests may turn up nothing, but that is info too. :)
Hi Jen, Yes, SAM and Picard indexes are present in the Local data (Data Manager) section of the admin page for the reference genome I am using (mm9).
I'm working with mouse data, trying to align to mm9, not a custom genome. I am generating the HISAT2 index for mm9, which is in the drop-down menu of the Data Manager for generating HISAT2 indexes.
I shared the sample data I've been using with you on the Test Galaxy site, am also happy to share my Cloudman instance (offline) if helpful.
Thanks Mo, I will get back to you tomorrow or Fri. Expect that input from John will be needed if a bug in collections themselves. But, let me review first. Thanks for the details, data share, and reporting the problem! Jen