mapping against concatenated fa files

[Alan Bilsland]

Hi,

First, thanks very much for the excellent programme. It feels a bit
odd posting on a forum as I have never done so before. Essentially,
I'm a wet lab scientist, but I know some C and I'm using these tools
to try to develop a workflow using SRA datasets with the current
objective of ultimately providing some data in support of a
justification for a benchtop NGSequencer. So, I don't do NGS regularly
at the moment.

It's possible my problem might be solved very easily before the
technical details. Basically, bsmap works briallintly against a single
human chromosme .fa file, but when I concatenate more than one chr.fa
file and use this as the reference (-d concat.fa) I get the [stack
errors?]:

0 [main] bsmap 6908 exception::handle:Exception:
STATUS_ACCESS_VIOLATION
905 [main] bsmap 6908 open_stackdumpfile: Dumping stack trace to
bsmap.exe.stacktrace

Maybe I'm just assembling the genome_concat.fa the wrong way, but I
read it should be OK just to concatenate the chr.fa files. If that's
wrong, any help would be appreciated. I'm currently downloading
individual chr.fa files from

http://hgdownload.cse.ucsc.edu/goldenPath/hg19/chromosomes

So, the software nuts and bolts are:

I use cygwin-1.7.9-1 and built bsmap-2.1 under this. To make the build
I deleted the supplied samtools folder and replaced it with
samtools-0.1.16 which I knew builds under cygwin with the makefile
modification suggested at

http://seqanswers.com/forums/showthread.php?t=7391

change top of file to

CC= gcc
CFLAGS= -g -Wall -O2 #-m64 #-arch ppc
CFLAGS= -I/usr/include/ncurses
DFLAGS= -D_FILE_OFFSET_BITS=64 -D_USE_KNETFILE -D_CURSES_LIB=1 -
Dexpl=exp -Dlogl=log
KNETFILE_O= knetfile.o:

Also, I had to make the single modification to the bsmap makefile

Change:

$@ $(THREAD) -lz -lbam

to:

$@ $(THREAD) -lbam -lz

As I say, this ran very well as tested against SRR222421.sra and
chr1.fa from the website above. So, I downloaded a few chr.fa files
and concatenated them in cygwin

$ cat chr1.fa chr2.fa etc.fa > concat.fa

I cannot map the bsreads against the concatenated file. I don't know
if needs indexed or something. Help!

Thanks,
Alan

[yxi]

Hi Alan,

Thanks for the feedback. bsmap should work with concatenated
human .fa files, I always do it.

But I haven't tested it on cygwin. One possible reason may be the
stack size of cygwin is not big enough. We experienced similiar
problem with some linux systems too, and it could be solved by
increasing the stack size using ulimit. I'm not familiar with cygwin
stack setting though. Since shot gun whole human genome take ~16GB
RAM, I recommend you run it on a linux64 server.

Thanks,

Yuanxin