Blast On A Custom Galaxy Server Is Unable To Find Database

Question

[magvazaryarvtre]

Hello,

I am currently trying to use megablast on a Galaxy server. I use a custom db that I prepared with makeblastdb -in some_sequences.fasta -dbtype nucl -out some_sequences. I put the path to my database in galaxy-dist/tool-data/blastdb.loc, I restarted my Galaxy server, but when I want to perform a megablast, the following error is raised ;

Cannot execute megaablast. BLAST Database error: No alias or index file found for nucleotide database [/home/john/databases/database_name/some_sequences] in search path [/home/john/galaxy-dist/database/job_working_directory/000/119::]

what have I done wrong?

thanks

[Carlos Borroto]

This is odd. For some reason BLAST is not recognizing "/home/john/databases/database_name/some_sequences" as a full path to a database, but instead as a database string name. Could you dump here the line in your blastdb.loc? Could you check if you are correctly using TABs in that file?

[magvazaryarvtre]

here is the content of my blastdb.loc ;

SSUref_111_tax_silva SSUred_11_tax_silva /home/john/databases/SSUref_111_tax_silva/SSURef_111_tax_silva.fasta

I checked that the fields are tab-delimited in my blastdb.loc

the same error is raised even if I try blast from command-line

[Carlos Borroto]

OK, If this is a nucleotide database do you have these files in the same directory?: SSURef_111_tax_silva.fasta.nhr SSURef_111_tax_silva.fasta.nin SSURef_111_tax_silva.fasta.nsq

And, what error do you get when you run blast manually? Could share here the error and the command line invocation you are using.

[magvazaryarvtre]

In the directory of my database, I have files with the following extensions : .nhr, .nin, .nsd, .nsi, .nsq

In blastdb.loc, the file name I provided is the root of all these files.

When I run blastn -db databases/SSUref_111_tax_silva/SSURef_111_tax_silva.fasta -query some_random_sequences.fasta the output is ok (I mean, the blastn command obviously works)

[Carlos Borroto]

I though you said above the same error is raised in your environment if you tried to run blast from the command-line.

Sorry I can't be of much help.

[Peter]

In blastdb.loc you used an absolute path (which is sensible), /home/john/databases/SSUref_111_tax_silva/SSURef_111_tax_silva.fasta

Here you used a relative path, -db databases/SSUref_111_tax_silva/SSURef_111_tax_silva.fasta so presumably you were in the directory /home/john/ when you did this? Try using the exact full absolute path copied from blastdb.loc instead?

Answer 1

[Istvan Albert]

Tools are run in temporary directories. So you will need to specify a full path to your database.

Alternatively make sure the wrapper copies/links that database to the temporary location that contains the intermediary run files.

[magvazaryarvtre]

The paths I provided to Galaxy are full paths. I checked the location of the temporary files ; Galaxy looks for an index file in the directory /home/john/galaxy-dist/database/job_working_directory/000:: This directory is empty after the execution of the workflow (but I guess this is normal since it store temporary files)

I guess the problem is there, but I don't know how to fix it.

[Peter]

Yes, give absolute (full) paths to the database location in blastedb.loc