SEQanswers

Go Back   SEQanswers > Bioinformatics > Bioinformatics



Similar Threads
Thread Thread Starter Forum Replies Last Post
How to collapse sequences excluding sequencing artifacts? (454) Yevaud Bioinformatics 3 01-30-2013 06:59 AM
Good variant filters christophpale Bioinformatics 0 11-21-2011 02:20 PM
SV emPCR Filters Anthony.287 454 Pyrosequencing 0 11-07-2011 06:41 AM
Non uniform gene coverage and Cufflinks predictions : artifacts ? hugo D RNA Sequencing 0 03-02-2011 07:04 AM
PubMed: Systematic artifacts in metagenomes from complex microbial communities. Newsbot! Literature Watch 0 07-10-2009 05:00 AM

Reply
 
Thread Tools
Old 09-20-2011, 12:48 PM   #1
efoss
Member
 
Location: Seattle

Join Date: Jul 2011
Posts: 98
Default sequencing artifacts passing GATK's ApplyRecalibration filters

I have five human exome data sets that I am putting through the GATK "Best Practices" pipeline. After running things through UnifiedGenotyper, I separated the SNPs from the indels and then ran VariantRecalibrator on the SNPs:

java -Xmx36g -jar /home/efoss/sequencing/GenomeAnalysisTK-1.1-31-gdc8398e/GenomeAnalysisTK.jar -T VariantRecalibrator -R /mnt/fred/home/efoss/gene_databases/human_genome/human_g1k_v37/human_g1k_v37.fasta --percentBadVariants 0.05 --maxGaussians 6 -B:input,VCF /mnt/fred/home/efoss/sequencing/MDS/all_five_patients_UnifiedGenotyper_output_SNPs_091711_1.raw.vcf -B:hapmap,VCF,known=false,training=true,truth=true,prior=15.0 /mnt/fred/home/efoss/gene_databases/GATK_resource_files/hapmap_3.3.b37.sites.vcf -Bmni,VCF,known=false,training=true,truth=false,prior=12.0 /mnt/fred/home/efoss/gene_databases/GATK_resource_files/1000G_omni2.5.b37.sites.vcf -B:dbsnp,VCF,known=true,training=false,truth=false,prior=8.0 /mnt/fred/home/efoss/gene_databases/GATK_resource_files/dbsnp_132.b37.vcf -an QD -an HaplotypeScore -an MQRankSum -an ReadPosRankSum -an FS -an MQ -mode SNP -recalFile /mnt/fred/home/efoss/sequencing/MDS/all_five_patients_UnifiedGenotyper_output_SNPs_091711_1.recal -tranchesFile /mnt/fred/home/efoss/sequencing/MDS/all_five_patients_UnifiedGenotyper_output_SNPs_091711_1.tranches -rscriptFile /mnt/fred/home/efoss/sequencing/MDS/all_five_patients_UnifiedGenotyper_output_SNPs_091711_1.plots.R

I then ran ApplyRecalibration:

java -Xmx36g -jar /home/efoss/sequencing/GenomeAnalysisTK-1.1-31-gdc8398e/GenomeAnalysisTK.jar -T ApplyRecalibration -R /mnt/fred/home/efoss/gene_databases/human_genome/human_g1k_v37/human_g1k_v37.fasta -B:input,VCF /mnt/fred/home/efoss/sequencing/MDS/all_five_patients_UnifiedGenotyper_output_SNPs_091711_1.raw.vcf --ts_filter_level 99.0 -tranchesFile /mnt/fred/home/efoss/sequencing/MDS/all_five_patients_UnifiedGenotyper_output_SNPs_091711_1.tranches -recalFile /mnt/fred/home/efoss/sequencing/MDS/all_five_patients_UnifiedGenotyper_output_SNPs_091711_1.recal -o /mnt/fred/home/efoss/sequencing/MDS/all_five_patients_UnifiedGenotyper_output_SNPs_091711_1.recalibrated.raw.vcf


My problem is that there are plenty of bad calls getting through ApplyRecalibration. I pick up quite a few mutations that appear in 5 out of 5 (unrelated) individuals, so I can be confident that they are sequencing artifacts. However, these mutations are often listed as passing the filters. Here are two examples:

1 801943 rs7516866 C T 44026.78 PASS AC=10;AF=1.00;AN=10;BaseQRankSum=4.380;DB;DP=1449;DS;Dels=0.00;FS=0.000;HRun=1;HaplotypeScore=16.2975;MQ=47.88;MQ0=66;MQRankSum=7.651;QD=30.38;ReadPosRankSum=5.739;SB=-18779.46;VQSLOD=-4.0745;culprit=ReadPosRankSum GT:ADP:GQ:PL 1/1:10,268:278:99:7953,572,0 1/1:6,293:299:99:9124,682,0 1/1:5,268:273:99:7936,586,0 1/1:4,295:299:99:9809,761,0 1/1:10,290:300:99:8784,610,0
1 802191 rs71507475 G A 1686.49 PASS AC=5;AF=0.50;AN=10;BaseQRankSum=2.704;DB;DP=1499;DS;Dels=0.00;FS=5.855;HRun=0;HaplotypeScore=44.5371;MQ=22.55;MQ0=460;MQRankSum=-3.161;QD=1.13;ReadPosRankSum=-0.860;SB=-867.92;VQSLOD=-15469854734583906.0000;culprit=HaplotypeScore GT:ADP:GQ:PL 0/1:224,69:300:99:286,0,2161 0/1:225,71:299:99:264,0,2035 0/1:237,58:300:99:230,0,2248 0/1:217,75:300:99:525,0,1676 0/1:213,77:300:99:426,0,1708

I gathered up all of these 5/5 sequencing artifacts and checked the results of the exome analysis published in Nature Genetics and not a single one appeared in those results, so clearly I'm either running the analysis incorrectly or I'm interpreting the results i ncorrectly (or both).

I would appreciate any suggestions.

Thank you very much for the help.

Eric
efoss is offline   Reply With Quote
Old 09-20-2011, 02:21 PM   #2
efoss
Member
 
Location: Seattle

Join Date: Jul 2011
Posts: 98
Default

To add a bit more information: The number of sequence information lines in my unrecalibrated file is the same as in my recalibrated file, so clearly nothing was filtered out. Furthermore, the FILTER column almost always says "PASS" and when it doesn't, it says "TruthSensitivityTranche99.00to99.90".
efoss is offline   Reply With Quote
Old 09-21-2011, 07:23 AM   #3
gavin.oliver
Senior Member
 
Location: uk

Join Date: Jan 2010
Posts: 110
Default

I may be wrong, but isn't the recalibrator just scoring your variants with a new metric? Therefore the filtering itself would still remain to be done based on the VQSLOD values?
gavin.oliver is offline   Reply With Quote
Old 11-22-2011, 03:36 AM   #4
wanguan2000
Member
 
Location: shanghai

Join Date: Nov 2010
Posts: 24
Default

I have the same question. And I found the number of LowQual using ApplyRecalibration as the same as the LowQual in raw.vcf.
LowQual number does not change!!! and a little PASS change to TruthSensitivityTranche99.00to99.90.

should I filtering itself based on the VQSLOD values? but:
http://www.broadinstitute.org/gsa/wi...th_the_GATK_v3
don't say anything.
wanguan2000 is offline   Reply With Quote
Old 04-04-2012, 07:55 AM   #5
petriedish
Junior Member
 
Location: Colorado

Join Date: Nov 2010
Posts: 3
Default

I realise this is a very late reply, but if you look at those results there are rs numbers from dbSNP showing that those calls are common SNPs. For example, the first one, rs7516866 is seen in approximately 25% of alleles. With the huge number of calls it is not surprising that some of these very common SNPs will be shared between all of your 5 samples.
petriedish is offline   Reply With Quote
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off




All times are GMT -8. The time now is 02:24 PM.


Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2019, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO