Predicting true SNPs from .vcf file

swbarnes2 · 04-05-2011, 01:22 PM

The .vcf file contains so many different measurements, and a couple of different quality scores, does anyone have a good empirical idea of which values are the best guidelines to picking out real SNPs from noise and artifacts? I've done about a hundred sanger sequencing reactions on a variety of predicted SNPs at a variety of quality levels, but the picture still isn't very clear.

For instance these SNPs looked real in the sanger:

GT:PL:GQ 0/1:20,11,149:13 gcacacacacacacacacacacacacacacacacacacacac gcacacacacacacacacacacacacacacacacacacac 211 DP=95 DP4=17,0,10,2 MQ=48 FQ=214

GT:PL:GQ 1/1:30,3,0:39 GAA GA 147 DP=10 DP4=0,0,1,6 MQ=48 FQ=-37.3

These did not confirm with sanger sequencing

GT:PL:GQ 0/1:24,0,114:27 A C 110 DP=130 DP4=14,35,2,22 MQ=45 FQ=113

GT:PL:GQ 0/1:40,0,47:42 T G 98.3 DP=71 DP4=17,1,14,0 MQ=44 FQ=101

Those don't look notably worse than the ones above them, so I'm not sure what I should have looked at to predict that the bottom two were false positives.

(My a priori assumption was that these variants were all real, because I made a multi-vcf with mpileup with this samples and many sibling animals, and these variants were common to all the animals)

elisadouzi · 04-06-2011, 03:29 PM

maybe the depth of the bottom two is too high. Have you try the -D options?

Quote:

Originally Posted by swbarnes2

The .vcf file contains so many different measurements, and a couple of different quality scores, does anyone have a good empirical idea of which values are the best guidelines to picking out real SNPs from noise and artifacts? I've done about a hundred sanger sequencing reactions on a variety of predicted SNPs at a variety of quality levels, but the picture still isn't very clear.

For instance these SNPs looked real in the sanger:

GT:PL:GQ 0/1:20,11,149:13 gcacacacacacacacacacacacacacacacacacacacac gcacacacacacacacacacacacacacacacacacacac 211 DP=95 DP4=17,0,10,2 MQ=48 FQ=214

GT:PL:GQ 1/1:30,3,0:39 GAA GA 147 DP=10 DP4=0,0,1,6 MQ=48 FQ=-37.3

These did not confirm with sanger sequencing

GT:PL:GQ 0/1:24,0,114:27 A C 110 DP=130 DP4=14,35,2,22 MQ=45 FQ=113

GT:PL:GQ 0/1:40,0,47:42 T G 98.3 DP=71 DP4=17,1,14,0 MQ=44 FQ=101

Those don't look notably worse than the ones above them, so I'm not sure what I should have looked at to predict that the bottom two were false positives.

(My a priori assumption was that these variants were all real, because I made a multi-vcf with mpileup with this samples and many sibling animals, and these variants were common to all the animals)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
How to get list of column in vcf file using Vcf.pm?	jessada	Bioinformatics	0	01-20-2012 07:22 AM
GATK UnifiedGenotyper calling way too many SNPs in vcf	swbarnes2	Bioinformatics	0	08-17-2011 01:33 PM
VCF formated bovine SNPs	Moo	Bioinformatics	3	05-23-2011 05:01 AM
Bovine SNPs in VCF format????? HELP!	AKilleen	Bioinformatics	1	05-10-2011 01:54 PM
Calling multiple BAM files for SNPs and vcf	newbietonextgen	Bioinformatics	3	04-19-2011 11:29 AM

04-05-2011, 01:22 PM	#1
swbarnes2 Senior Member Location: San Diego Join Date: May 2008 Posts: 912	Predicting true SNPs from .vcf file The .vcf file contains so many different measurements, and a couple of different quality scores, does anyone have a good empirical idea of which values are the best guidelines to picking out real SNPs from noise and artifacts? I've done about a hundred sanger sequencing reactions on a variety of predicted SNPs at a variety of quality levels, but the picture still isn't very clear. For instance these SNPs looked real in the sanger: GT:PL:GQ 0/1:20,11,149:13 gcacacacacacacacacacacacacacacacacacacacac gcacacacacacacacacacacacacacacacacacacac 211 DP=95 DP4=17,0,10,2 MQ=48 FQ=214 GT:PL:GQ 1/1:30,3,0:39 GAA GA 147 DP=10 DP4=0,0,1,6 MQ=48 FQ=-37.3 These did not confirm with sanger sequencing GT:PL:GQ 0/1:24,0,114:27 A C 110 DP=130 DP4=14,35,2,22 MQ=45 FQ=113 GT:PL:GQ 0/1:40,0,47:42 T G 98.3 DP=71 DP4=17,1,14,0 MQ=44 FQ=101 Those don't look notably worse than the ones above them, so I'm not sure what I should have looked at to predict that the bottom two were false positives. (My a priori assumption was that these variants were all real, because I made a multi-vcf with mpileup with this samples and many sibling animals, and these variants were common to all the animals)