On Wed, Apr 24, 2013 at 11:40 PM, Ganote, Carrie L cganote@iu.edu wrote:

Hi list,

I have a newly installed Galaxy instance (security_2013.04.08) and I am testing out NCBI Blast+ on it. I'm having trouble getting it to run.

I'm using the following tools from the Galaxy Main Toolshed: blast_datatypes b3a3ba0c1d47 ncbi_blast_plus 1f546099212f

Those are the latest releases on the main ToolShed, which is good.

http://toolshed.g2.bx.psu.edu/view/devteam/blast_datatypes Revision: 5:b3a3ba0c1d47 Version 0.0.15

http://toolshed.g2.bx.psu.edu/view/devteam/ncbi_blast_plus Revision: 8:1f546099212f Version 0.0.17

How did you install them? Via the automated Tool Shed install, or manually?

How did you setup the databases?

Looks like blast is being called fine, but I get this:

Fatal error: Matched on Error: Error: NCBI C++ Exception:

"/N/soft/rhel6/ncbi-blast+/ncbi-blast-2.2.28+-src/c++/src/corelib/ncbiobj.cpp", line 925: Critical: ncbi::CObject::ThrowNullPointerException() - Attempt to access NULL pointer.

Searching for that error suggests it can be triggered by a bad database: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=644962

If this is an NCBI database, check the MD5 checksums and/or redownload it.

If this is your own database, try deleting it and rebuilding it from the FASTA file.

WARNING:galaxy.datatypes.registry:Error loading datatype with extension 'asn1': 'module' object has no attribute 'GenericAsn1' WARNING:galaxy.datatypes.registry:Error loading datatype with extension 'asn1-binary': 'module' object has no attribute 'GenericAsn1Binary'

Anyone seen this?

I think this might be a fault in Galaxy with some recently added ASN1 formats (which we will later be subclassing for using with BLAST or other NCBI tools). The commit was here:

https://bitbucket.org/galaxy/galaxy-central/commits/d6f866cbf12d0242089fbbbf... https://bitbucket.org/galaxy/galaxy-dist/commits/d6f866cbf12d0242089fbbbfb79...

Curious,

Peter

On Mon, Apr 29, 2013 at 2:53 PM, Ganote, Carrie L cganote@iu.edu wrote:

Hi Peter,

...

How did you install them? Via the automated Tool Shed install, or manually?

I installed the tool wrappers via the Tool Shed, and Blast 2.2.28 is already installed on my cluster so I put the path in my environment setup file.

I'm not 100% sure but I think the BLAST 2.2.26+ which Galaxy installed via the Tool Shed will take priority over the system installed BLAST 2.2.28+ already on the cluster.

(I am intending to update this as BLAST 2.2.26+ is quite old, but I am currently struggling with the Galaxy unit test framework.)

...

How did you setup the databases?

I added the existing databases to my .loc files. They all show up in the menu for Blast so I think that checks out. Sometimes I've noticed that emacs has a problem with adding tabs so I used pico. Our HPC team updates the blast databases monthly, but we run with them from the bare iron to so I don't think they're the culprits.

I still use emacs for tabs, but curse regularly when I accidentally do Apple+Q and quite the terminal (rather than Ctrl+Q, tab).

...

...

"/N/soft/rhel6/ncbi-blast+/ncbi-blast-2.2.28+-src/c++/src/corelib/ncbiobj.cpp", line 925: Critical: ncbi::CObject::ThrowNullPointerException() - Attempt to access NULL pointer.

We were getting the null pointer exception even from the command line when we ran it with more than one thread. Funny, we reinstalled Blast 2.2.28 and this error went away. Who knows?

My guess is that your BLAST databases were updated around the same time (or maybe one cluster node had a bad database copy or something). But it could be down to the tool binaries themselves. Strange.

...

...

WARNING:galaxy.datatypes.registry:Error loading datatype with extension 'asn1': 'module' object has no attribute 'GenericAsn1' WARNING:galaxy.datatypes.registry:Error loading datatype with extension 'asn1-binary': 'module' object has no attribute 'GenericAsn1Binary'

...

I think this might be a fault in Galaxy with some recently added ASN1 formats (which we will later be subclassing for using with BLAST or other NCBI tools). The commit was here:

...

...

https://bitbucket.org/galaxy/galaxy-central/commits/d6f866cbf12d0242089fbbbf... https://bitbucket.org/galaxy/galaxy-dist/commits/d6f866cbf12d0242089fbbbfb79...

Nothing in those commits looks alarming, so I'm not sure where the bug is. Now blast works, but I'm still getting this error with every tool I run. I think I'll undo the commits for now.

Thanks!

Carrie Ganote

Have you checked the data.py and binary.py files do include the two classes GenericAsn1 and GenericAsn1Binary? It could be an hg merge went wrong and your local copy of Galaxy is missing them.

Peter

On Mon, Apr 29, 2013 at 3:38 PM, Ganote, Carrie L cganote@iu.edu wrote:

Hi Peter,

...

I'm not 100% sure but I think the BLAST 2.2.26+ which Galaxy installed via the Tool Shed will take priority over the system installed BLAST 2.2.28+ already on the cluster.

Blast 2.2.26 probably would take priority, if I installed it =D but I didn't set up the missing tool dependencies for this reason.

...

(I am intending to update this as BLAST 2.2.26+ is quite old, but I am currently struggling with the Galaxy unit test framework.)

Anything that can be helped with? From my testing, 2.2.28 works fine with the tool repository 1f546099212f.

There's a regression with makeblastdb and the -parse_seqids option, but otherwise BLAST 2.2.28+ fixes a lot of niggles I had with BLAST+. http://blastedbio.blogspot.co.uk/2012/10/my-ids-not-good-enough-for-ncbi-bla...

On the Galaxy side there are still limitations in the test framework, for example I'm stuck at the moment writing tests for makeblastdb: http://lists.bx.psu.edu/pipermail/galaxy-dev/2013-April/014302.html

I may compromise and update the Tool Shed to use BLAST 2.2.27+ in the short term. That should just require regenerating all the test output as things like e-values fluctuate between minor releases.

...

My guess is that your BLAST databases were updated around the same time (or maybe one cluster node had a bad database copy or something). But it could be down to the tool binaries themselves. Strange.

It wouldn't be the first time our blast databases were uploaded with error, so that is quite possible.

...

Have you checked the data.py and binary.py files do include the two classes GenericAsn1 and GenericAsn1Binary? It could be an hg merge went wrong and your local copy of Galaxy is missing them.

That was it. I'm not sure why those weren't merged, but I went ahead and made those changes!

Oh good - nothing me or Nicola can be blamed for ;)

Thanks for the update,

Peter