Hi Joel, On Sun, 2010-04-25 at 16:06 -0400, Joel McManus wrote: > Dear Mauve support, > > I'm trying to align two Drosophila genomes with mauve on a linux > computer, and got a "caught signal 11" message, before Mauve quit and > exited with no output. What caused this error? Is it something to do > with the input format? If it helps, one of the genome data files has > 14,730 contigs (fasta records) and the other has 15 chromosomes (fasta > records). Both files have gap sections with NNNNN in the sequence > (from scaffolding). The error occurred while loading the 14,730 > contig file. > > Will mauve work on fly genomes, or are they too large? For reference, > I've got 16 GB of RAM on the computer.
<snip> Short answer: yes, progressiveMauve will work on your fly genomes, but you probably need to run the 64-bit version that can be found in the linux-x64 folder of your Mauve 2.3.1 release. Long answer: Although I can't rule out a file format problem based on the information you've sent, the segmentation fault is probably happening because progressiveMauve is performing an operation that requires about 20 bytes of memory per base. Between the two fly genomes you've probably exhausted the ~4GB address space available to 32-bit linux programs. Mauve should have no problem with this task -- people here at UC Davis have used the contig reorderer module on plant genomes of comparable size, and I've been able to align pairs of draft genomes where each is >1GBp. That took a few days and used about 36GB of memory. In case it's not a 32bit/64bit problem, another thing to check is whether there's enough free disk space for the .sslist file that progressiveMauve is trying to construct -- it will be around 3GB for that genome. Also check that you have at least 1GB free in /tmp or whatever temporary file directory your system uses, since progressiveMauve will need about that much scratch disk space for drosophila size genomes. If /tmp isn't usable, you can point progressiveMauve to a different directory by setting the TMPDIR environment variable. In a bash shell you might do something like: export TMPDIR=$HOME/my_alignment_workspace In a compute cluster environment, TMPDIR would ideally be a disk that's local to the compute node rather than a networked storage device. That's often true of the default, which is /tmp. Hope that helps, -Aaron ------------------------------------------------------------------------------ _______________________________________________ Mauve-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/mauve-users
