SNPs at Interval Boundaries in GATK HaplotypeCaller

peterlchang · 10-24-2017, 02:53 PM

I'm using HaplotypeCaller with the -L option, where I explicitly state the interval. I am trying to break my task into pieces and assign it to different jobs for it to complete faster. In one run, I am running it, from say,

Chr1:1-1000000

and another run, I am running it twice, from

Chr1:1-500000 on one jobs and
Chr1:500001-1000000 on another

essentially splitting the interval into 2 different jobs. What I am seeing in this second run is that there are SNPs identified +/- 100 bases from 500000 that are not found in the first run.

My guess is that asking HC to focus only on a region (ie -L 1-500000) does not allow the local aligner to reassemble properly the reads, and hence results in spurious reads and SNPs. I was hoping that by specifying -L, it does the local aligner in a larger region and just report the SNPs in the -L region.

Has anybody heard of this or have a way around?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
GATK HaplotypeCaller READS?	hr3y	Bioinformatics	0	06-29-2017 02:31 PM
Can't run GATK HaplotypeCaller set ploidy = 1	Genomics101	Bioinformatics	1	09-29-2015 06:37 AM
Too many Variants called by HaplotypeCaller GATK	drmaly	Bioinformatics	2	12-05-2013 09:56 PM
GATK HaplotypeCaller memory problem	Robby	Bioinformatics	1	04-03-2013 11:18 AM
GATK- haplotypecaller or unifiedgenotyper?	lre1234	Bioinformatics	2	03-28-2013 09:48 AM

10-24-2017, 02:53 PM	#1
peterlchang Junior Member Location: Los Angeles Join Date: Mar 2010 Posts: 5	SNPs at Interval Boundaries in GATK HaplotypeCaller I'm using HaplotypeCaller with the -L option, where I explicitly state the interval. I am trying to break my task into pieces and assign it to different jobs for it to complete faster. In one run, I am running it, from say, Chr1:1-1000000 and another run, I am running it twice, from Chr1:1-500000 on one jobs and Chr1:500001-1000000 on another essentially splitting the interval into 2 different jobs. What I am seeing in this second run is that there are SNPs identified +/- 100 bases from 500000 that are not found in the first run. My guess is that asking HC to focus only on a region (ie -L 1-500000) does not allow the local aligner to reassemble properly the reads, and hence results in spurious reads and SNPs. I was hoping that by specifying -L, it does the local aligner in a larger region and just report the SNPs in the -L region. Has anybody heard of this or have a way around?