Problem with mpiBLAST output subject sequence names

[quokka]

Hi all,

I'm experiencing an issue when running mpiBLAST 1.6.0 on our cluster. It runs fine however the results (from a TBLASTX run) do not have the complete name of the subject sequence in the output. Here's an example of part of the output:

Sequences producing significant alignments:                  (bits) Value  N

13893_/work1/xxmwebb/mpi_blast/databas                             73   1e-14  1 

>13893_/work1/xxmwebb/mpi_blast/databas 
          Length = 9798

 Score = 73.0 bits (162), Expect = 1e-14
 Identities = 32/35 (91%), Positives = 33/35 (94%)
 Frame = -3 / -3

                                               
Query: 105  LACQTLKSGYTESSRGSRVYFLVAFSLFLCTILTF 1
            LACQTLKSGYTESSRGS VYFLVAFSLFLC+IL F
Sbjct: 8944 LACQTLKSGYTESSRGSSVYFLVAFSLFLCSILAF 8840

The problem is that the subject title is being truncated (and having the full path included in the subject name isn't helping the issue). Does anyone know how to get around this?

[maubp]

Given no updates since 2012, mpiBLAST is effectively dead. I guess that your best plan would be to rebuild your database using shorter names.

However, I would look at moving to using NCBI BLAST+ instead. This has built in multi-threading which works well on multi-core machines. In terms of exploiting a cluster (which is what mpiBLAST was for), splitting your input by query is the by far the easiest approach (assuming you are running searches with multiple-sequence FASTA files as input).

[quokka]

mpiBLAST is inserting the path into the subject name for some reason...so switching to a shorter path for the input might be required....hmmmn....I don't recall this being an issue before.....oh well,BLAST+ it is then....

Thanks for the advice maubp..