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