Hello everyone,
I experienced quite strange Ti/Tv ratios in my GATK VariantEval report (I cut out some columns to make it fit better in here):
[...]
Analysis Name: Quality Metrics by allele count
Analysis Description: Shows various stats binned by allele count
Table Name : MetricsByAc
filter_name novelty_name AC n Ti/Tv
------------------------------------------
called known 0 268411 0.40751131364086857
called known 1 372920 2.1674523293837855
called known 2 655967 2.1697342797914443
called novel 0 345 0.3372093023255814
called novel 1 111817 1.3576654648196176
called novel 2 52628 1.326716477297847
[...]
This lists the Ti/Tv ratios according to known(dbSNP) and novel SNPs and different allele counts (AC).
I have two questions here:
(1) I don't quite understand, why GATK calls 345 novel SNPs with AC=0. If I understand it right, this means that both alleles are reference alleles. Why did GATK make a call here, if there is no SNP?
(2) What could be a reason for the low Ti/Tv ratio (~0.4) for the 268.411 called known SNPs? I would expect it to be the other way round, a high value over the standard 1,7005.
There is another strange thing in another table (just a subset again):
[...]
Analysis Name: Ti/Tv Variant Evaluator
Analysis Description: Ti/Tv Variant Evaluator
filter_name novelty_name ti/tv_ratio ti/tv_ratio_standard
---------------------------------------------------------
called all 1.4964 14756673 1.7005
called known 1.5172 14756673 1.7005
called novel 1.3440 14756673 1.7005
filtered all 1.5269 14756673 1.7005
filtered known 2.1049 14756673 1.7005
filtered novel 1.0674 14756673 1.7005
[...]
I guess the Ti/Tv ratios for filtered SNPs look ok. But the ratios for the actual calls are quite bad and pretty much the same for known and novel SNPs. I'd expect it to be the other way round here as well. Does anyone have an idea how to explain this?
I used mainly the pipeline and parameters suggested on the GATK homepage including "Base quality score recalibration", "Local realignment around indels" and the "VariantFiltrationWalker" filtering SNPs in low-complexity areas. The threshold for calling and emitting in the "Unified genotyper" are 30 and 10 respectively. Everything else is pretty much default. My GATK version is 1.0.4705; a bit old, but I started this analysis quite a while ago and do not want to change the version inbetween.
My data is a single Illumina paired-end sequencing run of human DNA with 75bp reads and ~6x coverage. Not that much, but I also saw similar tendencies in another run with 20x coverage.
Hope I'm not missing some important information. Otherwise I'd give any info needed for answering.
Would be very thankful if someone has some ideas what might have gone wrong and maybe even how to fix it.
Thanks!
Christoph
I experienced quite strange Ti/Tv ratios in my GATK VariantEval report (I cut out some columns to make it fit better in here):
[...]
Analysis Name: Quality Metrics by allele count
Analysis Description: Shows various stats binned by allele count
Table Name : MetricsByAc
filter_name novelty_name AC n Ti/Tv
------------------------------------------
called known 0 268411 0.40751131364086857
called known 1 372920 2.1674523293837855
called known 2 655967 2.1697342797914443
called novel 0 345 0.3372093023255814
called novel 1 111817 1.3576654648196176
called novel 2 52628 1.326716477297847
[...]
This lists the Ti/Tv ratios according to known(dbSNP) and novel SNPs and different allele counts (AC).
I have two questions here:
(1) I don't quite understand, why GATK calls 345 novel SNPs with AC=0. If I understand it right, this means that both alleles are reference alleles. Why did GATK make a call here, if there is no SNP?
(2) What could be a reason for the low Ti/Tv ratio (~0.4) for the 268.411 called known SNPs? I would expect it to be the other way round, a high value over the standard 1,7005.
There is another strange thing in another table (just a subset again):
[...]
Analysis Name: Ti/Tv Variant Evaluator
Analysis Description: Ti/Tv Variant Evaluator
filter_name novelty_name ti/tv_ratio ti/tv_ratio_standard
---------------------------------------------------------
called all 1.4964 14756673 1.7005
called known 1.5172 14756673 1.7005
called novel 1.3440 14756673 1.7005
filtered all 1.5269 14756673 1.7005
filtered known 2.1049 14756673 1.7005
filtered novel 1.0674 14756673 1.7005
[...]
I guess the Ti/Tv ratios for filtered SNPs look ok. But the ratios for the actual calls are quite bad and pretty much the same for known and novel SNPs. I'd expect it to be the other way round here as well. Does anyone have an idea how to explain this?
I used mainly the pipeline and parameters suggested on the GATK homepage including "Base quality score recalibration", "Local realignment around indels" and the "VariantFiltrationWalker" filtering SNPs in low-complexity areas. The threshold for calling and emitting in the "Unified genotyper" are 30 and 10 respectively. Everything else is pretty much default. My GATK version is 1.0.4705; a bit old, but I started this analysis quite a while ago and do not want to change the version inbetween.
My data is a single Illumina paired-end sequencing run of human DNA with 75bp reads and ~6x coverage. Not that much, but I also saw similar tendencies in another run with 20x coverage.
Hope I'm not missing some important information. Otherwise I'd give any info needed for answering.
Would be very thankful if someone has some ideas what might have gone wrong and maybe even how to fix it.
Thanks!
Christoph