BFAST indexing for 50+36 SOLiD paired end?

krobison · 10-27-2010, 07:49 PM

The BFAST manual gives one set of recommended seeds for reads <40 bp and one for greater than 40 bp

What if I have a SOLiD paired end dataset with one read above that threshold (50) and one below (35)? Suggestions?

Chipper · 10-27-2010, 10:09 PM

Tried bfast+bwa (latest release)?

drio · 10-28-2010, 05:30 AM

Quote:

Originally Posted by krobison

The BFAST manual gives one set of recommended seeds for reads <40 bp and one for greater than 40 bp

What if I have a SOLiD paired end dataset with one read above that threshold (50) and one below (35)? Suggestions?

Those keys will work fine for 35bp reads. You'll get a small decrease in the total number of alignments compared to 50bp reads. You could also explore
bfast+bwa branch and use bwaaln for the second tag.

krobison · 10-28-2010, 10:48 AM

Thanks -- I've just started trying to figure out bfast+bwa & the new options it gives

epigen · 11-02-2010, 03:58 AM

In this thread I posted my experience with this kind of PE reads:
http://seqanswers.com/forums/showthread.php?t=7100
BFAST postprocess seems to rescue a lot of the 35 bp ends that did not get a match with bwaaln. (The bwaaln step might profit from allowing more mismatches instead of the default, but as I found with BWA on SOliD data that slows it down considerably.)
Note that the git version of postprocess is an improvement over the one in the bfast+bwa package but still has a bug.
Maybe it's because of the rather low quality of my data that the localalign step is now the most time-consuming part: 50 Mio read pairs on 8 CPUs usually take 70 h, some almost double the time. Multithreading is not that efficient that I'd want to use 16 threads (other people also want to use the cluster). I'm currently experimenting with the parameters.

drio · 11-02-2010, 06:04 AM

@epigen: Take a look to this.

epigen · 11-02-2010, 06:25 AM

Thanks for the link, but I tried that before and it is not helping: far too many reads do not get paired. This patch is what I mean with the git version of postprocess, which does help (in terms of speedup and pairing, unfortunately with a bug):
http://sourceforge.net/mailarchive/m...mail.gmail.com

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
SOLiD paired-end read analysis with BFAST 0.7.0a	idonaldson	Bioinformatics	4	03-20-2012 09:31 PM
BFAST for SOLiD paired end reads	epigen	Bioinformatics	31	09-03-2011 06:20 AM
BFAST mapping paired end reads.	tanghz	Bioinformatics	10	04-29-2011 07:29 AM
How to map SOLiD paired end reads by Bfast	beliefbio	Bioinformatics	1	12-29-2010 01:55 AM
bfast paired end flag	guavajuice	Bioinformatics	0	11-02-2010 12:37 PM

10-27-2010, 07:49 PM	#1
krobison Senior Member Location: Boston area Join Date: Nov 2007 Posts: 747	BFAST indexing for 50+36 SOLiD paired end? The BFAST manual gives one set of recommended seeds for reads <40 bp and one for greater than 40 bp What if I have a SOLiD paired end dataset with one read above that threshold (50) and one below (35)? Suggestions?

11-02-2010, 06:04 AM	#6
drio Senior Member Location: 41°17'49"N / 2°4'42"E Join Date: Oct 2008 Posts: 323	@epigen: Take a look to this. __________________ -drd

10-27-2010, 10:09 PM	#2
Chipper Senior Member Location: Sweden Join Date: Mar 2008 Posts: 324	Tried bfast+bwa (latest release)?

10-28-2010, 10:48 AM	#4
krobison Senior Member Location: Boston area Join Date: Nov 2007 Posts: 747	Thanks -- I've just started trying to figure out bfast+bwa & the new options it gives

11-02-2010, 03:58 AM	#5
epigen Senior Member Location: Germany Join Date: May 2010 Posts: 101	In this thread I posted my experience with this kind of PE reads: http://seqanswers.com/forums/showthread.php?t=7100 BFAST postprocess seems to rescue a lot of the 35 bp ends that did not get a match with bwaaln. (The bwaaln step might profit from allowing more mismatches instead of the default, but as I found with BWA on SOliD data that slows it down considerably.) Note that the git version of postprocess is an improvement over the one in the bfast+bwa package but still has a bug. Maybe it's because of the rather low quality of my data that the localalign step is now the most time-consuming part: 50 Mio read pairs on 8 CPUs usually take 70 h, some almost double the time. Multithreading is not that efficient that I'd want to use 16 threads (other people also want to use the cluster). I'm currently experimenting with the parameters.

11-02-2010, 06:25 AM	#7
epigen Senior Member Location: Germany Join Date: May 2010 Posts: 101	Thanks for the link, but I tried that before and it is not helping: far too many reads do not get paired. This patch is what I mean with the git version of postprocess, which does help (in terms of speedup and pairing, unfortunately with a bug): http://sourceforge.net/mailarchive/m...mail.gmail.com