QC's contamination data

[maotian06]

@tyamaguchi-ucla
I am reading my OSCC QC's contamination data, and here is one of the sample outputs:

sample	contamination	error
MTOSIMHO000033-N001-A01-F	0.12376796182545992	0.0074568334509377415

But my other ATC call-gSNP sample is like:

sample	contamination	error
ATCWGS21N	8.74087704075081E-5	9.059114090615396E-6

Based on GATK's code, I think the contamination is based on the ratio of the contamination:
contaminationEstimate = contaminationOppositeDepth / totalDepthWeightedByOppositeFrequency

So 0.1 = 10%.

But for the error bar of the contamination level, the output doesn't have the estimate.high and estimate.low. Can I directly use the +error / -error as the error bar?

[tyamaguchi-ucla]

According to https://gatk.broadinstitute.org/hc/en-us/articles/360050814292-CalculateContamination, it "calculates the fraction of reads coming from cross-sample contamination, given results from GetPileupSummaries". So, the equation makes sense. It looks like the OSCC sample was highly contaminated. (~12%) Any other highly contaminated samples? How about call-sSNV results? They might have shown hypermutation then.

[maotian06]

Yes, that means the OSCC data was highly contaminated.
I check my ATC dataset, most of the samples the contamination rate is below 1%. So I guess it is just the OSCC dataset.
Any possible way to save this data?

[tyamaguchi-ucla]

I would drop samples with cross individual contamination > 3~4%. Any normal and tumour pairs with less contamination? All samples just have one pair of FASTQs so partial clean-up is not practically possible.