Blast 2.25+ Segmentation Fault When -Outfmt Set To 6,7 Or 10

Question

[Biomonika (Noolean)]

Hi all!

I am running blast 2.25+ and willing to get output in format 7, setting -outfmt 7. However, with this setting I always get Segmentation fault. Same happens with -outfmt 6 and 10. Other settings of -outfmt work as expected. Is there any solution to this?

Thanks a lot.

Monika

[Amr]

are you running directly through the command line or via a script of some kind?

also have you tried to add the custom format specifiers? I know that leaving it blank should just give you efault but try specifying something after the option - eg -outfmt "6 qacc"

[Biomonika (Noolean)]

your setting -outfmt "6 qacc" gives no Segmentation fault. However, it just writes "Query_1" on as many lines, as many hits it found.

[Michael]

Maybe we should make a nice collection of sequences that can crash BLAST, or other cases where normal input crashes bioinformatics software xD

Answer 1

[Michael]

Similar problem here, but my problem shows up as a a null-pointer exception when using tblastx with outfmt 6,7,10 and when the input contains a certain sequence, but occurs independent of the database. I guess there are more sequences that can crash blast. It is a bit funny, but until 2.26 is out there is not much you can do. My solution was to use standard blast format. But just in case you should try to send a reproducible example to NCBI.

---

Excerpt from my email:

I have now boiled down the case and have isolate a single input sequence that crashes tblastx with format 6 and 7 with each database I have tried:

>crashblast
TGTGTTGGTGTGTTGGTGGTGTGTGTGTGTGTGTGTGTGTGT

$ tblastx -query crashblast.fa -db testdb.fasta  -outfmt 6
Error: NCBI C++ Exception:
   "/am/ncbiapdata/release/blast/src/2.2.25/Linux64-Suse-icc/c++/ICC1010-ReleaseMT64--  
Linux64-Suse-icc/../src/corelib/ncbiobj.cpp", line 688: Critical:  
ncbi::CObject::ThrowNullPointerException() - Attempt to access NULL pointer.

I have made sure that there are no funny character or anything by re-typing the sequence in an editor. The error does not depend on the fasta header nor the exact length of the sequence. While adding e.g. an A to the beginning 'cures' it, deleting one GT from the end retains the error.

The error is raised using formats 6,7,10,11, other formats work correctly and show no hit.

This seems to be independent of the database, anyhow for your reference I post the steps to create my database:

makeblastdb -in testdb.fasta  -dbtype nucl

I have been in contact with the blast support and they said they possibly have a fix for this in the next version.

[Daniel Standage]

The world would be a happier place if all bug reports were this good!

[Michael]

Thanx :) Btw, extracting the 'offending' sequence is not that hard, I looked at the last line of output, and the sequence immediately following it in the input fasta was the offender.

Answer 2

[Biomonika (Noolean)]

I have had conversation with Peter Cooper from NCBI and I have already got answer to this problem. Thing is that my database missed first line with description. It mystically worked with some formats and with some not. To sum it up, after adding

>database_description_line

and making database from this fasta file it worked out. Have a nice day!