[Bioclusters] mpiBLAST and NCBI Blast? BioseqFindFunc: couldn't uncache

Dan Bolser bioclusters@bioinformatics.org
Wed, 8 Sep 2004 13:10:54 +0100 (BST)


On Wed, 8 Sep 2004, Joe Landman wrote:

>Dan Bolser wrote:
>
>>
>>Basically I would like to know if I am or could be running mpiBLAST
>>without knowing it, and all these bugs traceback to mpiBLAST, or if
>>mpiBLAST is reporting (in the web page above) a bug which traces back to
>>BLAST on multi CPU archetecture.
>>  
>>
>
>It is highly unlikely you are running mpiBLAST without knowing it.  What 
>does your command line look like?  This looks like it is a translation 
>bug in the lower reaches of the blast code.  mpiBLAST hits those code 
>sections, so it sees the bug more often than not.  The notes indicate 
>that it is related to using one of the -m options (for output 
>translation).  What does the rest of your command line look like?

That is what I guessed (I didn't ever knowingly install mpiBLAST).

formatdb -i sequences.fa -o T
blastall -d sequences.fa -p blastp -i domain_sequences.fa  -m 8 -a4 > results



>
>>I get "blastall 2.2.8".
>>
>>
>>Anybody know if this 'bug' is serious? Can I ignore it?
>>  
>>
>
>Hmm.  Does it impact the output data you need (actual alignment 
>information), or are you after parameters and computational "side 
>effects" (bit scores, E-values,...)

I haven't looked in detail yet. I am interested in the similarity of these
sequences to themselves (hence blastall).

I want to make my own clusters of these sequences / compare clusterings.

I use -m 8 because I don't care about the alignment, but perhaps that
would give clues to the error.

If you like I can re-run the problem sequences with -m 7 and send along
some results.

Cheers,
Dan.

>
>
>