Haploid genomes

gilgi

Dear GATK team,

I know that in the past GATK was not suitable for haploid genomes. I wanted to ask if this possibly changed since then - and whether it is possible to use GATK for haploid genomes.

Thanks a lot, Gilgi

Mark_DePristo

The ug in gatk2 can call haploid sequence natively now. You just set ploidy to 1.

gilgi

Thanks! This is great news! I'll try to work with it.

gilgi

Hi, We tried to download gatk2 from here: http://www.broadinstitute.org/gatk/download

But after download saw that the version doesn't seem to be 2 but: version 1.6-596-g3b9929c

Is this the correct version?

Thanks, Gilgi

ebanks

The 2.0 release should be coming out later today.

gilgi

Thanks a lot. I downloaded the gatk 2.0.

When trying to use the UnifiedGenotyper with --sample_ploidy 1 I get an error: MESSAGE: Incorrect genotype calculation model chosen. Only [POOLSNP|POOLINDEL|POOLBOTH] supported with this walker if sample ploidy != 2

What does this mean? My data isn't pool, I have individual (barcoded) haploid sequenced strains. I tried to add: --genotype_likelihoods_model POOLBOTH

But then I get: MESSAGE: Incorrect AF Calculation model. Only POOL model supported if sample ploidy != 2

I tried to look for the answer in the guide - without success. Can you help please?

delangel

Gilgi - sorry, we should have made documentation more explicit. I've added a new post in http://gatk.vanillaforums.com/discussion/1214/calling-non-diploid-organisms-with-unifiedgenotyper

That speficies how to run this. You need to additionaly put in -pnrm POOL into your command line.

gilgi

Thanks a lot!!!

hilli

Ive used the -pnrm POOL option but still getting the same error as gilgi. So I have

java -Xmx30g -jar /usr/local/gatk2/GenomeAnalysisTK.jar -T UnifiedGenotyper -R ref -I bam -I bam -pnrm POOL -polidy 1 -o vcf

Any help? Ali

ebanks

We have changed the arguments so that they more accurately reflect what they are doing: http://www.broadinstitute.org/gatk/gatkdocs/org_broadinstitute_sting_gatk_walkers_genotyper_UnifiedGenotyper.html#--genotype_likelihoods_model So you'll want e.g. -pnrm GeneralPloidySNP

flow

With v2.0-39 I had been using -ploidy 1 -pnrm POOL -glm POOLSNP. Can you confirm that for calling of snps/indels in haploid genomes as of v2.1 would now be -ploidy 1 -pnrm EXACT -glm GeneralPloidySNP. Is the -pnrm POOL option now defunct?

delangel

You just need -ploidy 1. "-pnrm EXACT" will work but there's no other option :). "-glm GeneralPloidySNP" will not work - you need either SNP, INDEL or BOTH.

flow

Thanks for your reply. Under what circumstance should -glm GeneralPloidySNP/GeneralPloidyINDEL be used?

aunderwo

I am experiencing a problem with the ploidy 1 option. Having used GATK2 unified genotyper with the params --sample_ploidy 1 --genotype_likelihoods_model BOTH -rf BadCigar I get the following line in a vcf file (see sample in bold)

Staphylococcus 1553115 . A G 24454.01 . AC=13;AF=0.813;AN=16;BaseQRankSum=1.072;DP=1040;Dels=0.00;FS=32.822;HaplotypeScore=3.3463;MLEAC=13;MLEAF=0.813;MQ=40.20;MQ0=47;MQRankSum=-10.543;QD=32.13;ReadPosRankSum=-1.148;SB=-9.076e+03 GT:AD:DP:GQ:MLPSAC:MLPSAF:PL 1:0,29:29:99:1:1.00:1015,0 1:0,62:62:99:1:1.00:2053,0 1:0,106:106:99:1:1.00:3210,0 1:0,102:102:99:1:1.00:3305,0 1:0,88:88:99:1:1.00:2750,0 1:0,41:41:99:1:1.00:1324,0 1:0,76:76:99:1:1.00:2448,0 1:0,39:39:99:1:1.00:1303,0 0:64,40:104:99:0:0.00:0,1334 1:0,41:41:99:1:1.00:1373,0 1:0,49:49:99:1:1.00:1668,0 0:72,50:122:99:0:0.00:0,1258 1:0,59:59:99:1:1.00:1852,0 1:0,38:38:99:1:1.00:1192,0 1:0,31:31:99:1:1.00:961,0 0:53,0:53:99:0:0.00:0,1633

The sample in bold is called as WT (genotype 0) with a high GQ despite there being 72 reads of genotype 0 and 50 of genotype 1. Examining the bam file suggests that this is a mapping error in a repetitive phage region

If I set ploidy to be 2 the equivalent line in the resulting vcf file is

Staphylococcus 1553115 . A G 24788.02 . AC=28;AF=0.875;AN=32;BaseQRankSum=0.947;DP=1040;Dels=0.00;FS=32.822;HaplotypeScore=3.3463;InbreedingCoeff=0.4286;MLEAC=28;MLEAF=0.875;MQ=40.20;MQ0=47;MQRankSum=-10.096;QD=25.11;ReadPosRankSum=-1.177;SB=-9.871e+03 GT:AD:DP:GQ:PL 1/1:0,29:29:81:986,81,0 1/1:0,62:62:99:1895,156,0 1/1:0,106:106:99:2992,247,0 1/1:0,102:102:99:3169,268,0 1/1:0,88:88:99:2452,193,0 1/1:0,41:41:99:1243,99,0 1/1:0,76:76:99:2283,193,0 1/1:0,39:39:99:1233,105,0 0/1:64,40:104:99:886,0,1706 1/1:0,41:41:99:1298,108,0 1/1:0,49:49:99:1581,129,0 0/1:72,50:122:99:1235,0,2126 1/1:0,59:59:99:1649,132,0 1/1:0,38:38:87:1065,87,0 1/1:0,31:31:69:821,69,0 0/0:53,0:53:99:0,138,1588

As can be seen from the bold text, the same position is called as heterozygote which based on the number of the reads mapping would be likley except for the fact this is a bacterial haploid genome. Previously I would have discarded this since the heterozygous call indicates mis-mapping as the bam file confirms. I had been hoping to use the sample_polidy option set to 1 for bacterial genomes but this results concerns me. I could obviously filter based on AD but the wonder why the sample was given a high GQ when the polidy is set to 1 and the AD suggests the call of genotype 0 should be doubted.

Any suggestions on what is going on here?? Many thanks

Anthony

delangel

The code is actually doing what it's designed to do - when you're using -ploidy 1, there are only 2 possible genotype assignments, and the assignment "0" is by far the most likely one even if 40% of your reads have another base. In the default diploid case, the most likely genotype is the 0/1 one, which is exactly what you're getting. Even in the haploid case, there's considerable evidence that favors the "0" genotype (plus the population prior), so you'll get a high value of GQ anyway - your PL values of 0,1258 indicate that, statistically, it's 10^125 likelier that your data came from a reference site than from an alt site based on all the available reads.

flow

In the current documentation (v2.3-9) for the Unified Genotyper there is a caveat stating "We only handle diploid genotypes". Has something changed or can -ploidy still be safely set to 1?

Geraldine_VdAuwera

That's an old caveat, I'll remove it from the doc. You're good to go.