BLAST segmentation fault

1,227 views
Skip to first unread message

pelon

unread,
Mar 2, 2007, 1:41:38 PM3/2/07
to Group-4-Bioinformatics
Hi:

I am having a very strage behaviour on my standalone BLAST.... Every
time that add the XML output parameter (-m 7) I get a "segmentation
fault"... If I do the same search without that type of output I do not
get the error....

I am currently using version 2.2.15, on fedora core 5 (kernel
2.6.18-1.2257.fc5)... I have observed the same error on blast version
2.2.14 with fedora core 5 (2.6.16-1.2133_FC5)

Any comments would be greatly apreciated....


Jose

Fred

unread,
Mar 5, 2007, 4:15:51 PM3/5/07
to Group-4-Bioinformatics
Hello Jose,

My guess would be that the resulting XML output from your blast is
very large (due to all the XML tags) and that it causes a memory
problem.
So could you add the following options to your blastall command line
in order to reduce the size of your blast result file in XML format ?
-v 2 -b 2
(-v => Number of one-line descriptions in the output)
(-b => Number of alignments to show)

So do you still get the "segmentation fault" with these new settings ?

Hope this help,,

Fred

pelon

unread,
Mar 6, 2007, 2:10:07 PM3/6/07
to Group-4-Bioinformatics
Fred

Thank you for the prompt response... Adding those parameters did
actually help, now Im curious, If you limit the number of alignments
to show in the output (i.e.: -b 2) is there any way to access those
alingments that are not shown? Is it possible to iterate amongst those
that were not in the final output?

And another thing... I still havent found a good document explaining
these issues out there... a more advanced, but properly formated
documentation of blast... any suggestions for the latter?

Cheers!

Jose

Fred

unread,
Mar 7, 2007, 3:20:21 PM3/7/07
to Group-4-Bioinformatics
Hi Jose,


> Thank you for the prompt response... Adding those parameters did
> actually help, now Im curious, If you limit the number of alignments
> to show in the output (i.e.: -b 2) is there any way to access those
> alingments that are not shown? Is it possible to iterate amongst those
> that were not in the final output?

Actually I advised you to limit the number of alignments just to check
if it works when you Blast result file is small.
So it seems to work. Now you should check at which number N of
alignments (i.e. : -b N) it starts to fail and gives you a
"segmentation fault". Then check the size of your Blast result file
with the option -b (N-1).
Because it seems that it is the size of the generated file that is
critical. Hopefully there will be an option to change in the set up of
Blast or of your computer that can be modify in order that you get
your complete Blast result file.

>
> And another thing... I still havent found a good document explaining
> these issues out there... a more advanced, but properly formated
> documentation of blast... any suggestions for the latter?

For the Documentation related to Blast the following one is quite nice
and gives you access to several kinds of information through numerous
links:
http://www.ncbi.nlm.nih.gov/BLAST/

This one is good for the blast parameters explanation :
http://cbi.labri.fr/Tools/blast/docs/full_options.html


Regards,

Fred

Reply all
Reply to author
Forward
0 new messages