How low is OK to accept variants in high/low GC content regions?

Question

[Tania]

Hi Every one

If we have an interesting variant in a very high GC content, so we expect low coverage anyway. The variant has depth 18 with AD 15,3.

Is having 3 reads enough to say there is a variant at this site? GQ = 78.

Given that the all the reads has 58-60 Mapping quality, All the bases quality are between 35 -40 and the variant passes the hard filtering of GATK pipeline, like QD =2.7 (greater than 2) , FS <10 , SOR <3 and all other metrics given GATK suggestions pass..etc.

Thanks

[GenoMax]

Tania : This is an open forum for bioinformatics research projects. No one here is qualified to offer advice that can be applied in patient care. @Kevin has done the due diligence of recommending additional investigation but please keep in mind that the help provided on Biostars is for "research" applications only.

[Tania]

I understand, and I did not ask about any patient care advice,I am even a computer science student ! I asked how low is ok for high/low gc content. I never mentioned patient care.

[GenoMax]

Good to hear that. This part had raised some concern. Perhaps you were referring to other patients that are waiting to see him.

but my advisor is a medical professor, and he has his patients waiting. At least I can include it now to him if I have a logic behind low coverage and worrying AD, and he can decide what to do with the patients.

[Tania]

Not me, my advisor, and this is when Kevin mentioned patient care, I said I am checking the logic behind my analysis and he cab decide whatever he think with his patients :) Actually, this is my role, I discuss with my advisor my bioinformatics logic, and never think about this patient care stuff, he decides :) Thanks genomax for highlighting this anyway, I can be careful in discussions not to raise any concerns :)

Answer 1

[Kevin Blighe]

I have called Sanger-confirmed homozygous variants from just a single read, in the past, and Sanger-confirmed heterozygous from a minimum of 2 (clinical samples); however, calls at total read depths this low would never be included in a clinical setting, and the affected sample would be repeated.

In your case, the total read-depth over the position is actually 18. if it has passed all QC metrics, then you could 'proceed with caution'. The total read depth of 18 is actually the minimum required for certain labs in the UK National Health Service in terms of clinical genetic testing.

The imbalance on AD is worrying but not unusual.

Hopefully the GATK has called it heterozygous.

[Tania]

Thanks Kevin. Yes, GATK called it heterozygous with GQ =78. I also checked the region mappability, it has Duke uniqueness score 1. So, I think it is mappable as far as I understand.

[Kevin Blighe]

Yes, and you have stated that the MAPQs are good, too. Many regions of the genome suffer dips in depth of coverage, and a lot of this does have to do with GC content, which affects the unwinding ('melting') of the DNA double-helix and also the binding affinity of primers/adapters. Data shows that high-GC regions actually require a higher melting temperature than that which is typically used, and that this (coupled with a few other extra processing steps) can then make depth of coverage more uniform.

Look at it this way: If this variant was called in my former lab in the NHS and there was a patient waiting on the result, it would be preliminarily included but would only make the final report with Sanger confirmation. I can only guess that your work is research only, in which case I believe to include it is not an issue.

[Tania]

Thanks Kevin, I am not a medical doctor, but my advisor is a medical professor, and he has his patients waiting. At least I can include it now to him if I have a logic behind low coverage and worrying AD, and he can decide what to do with the patients. Thanks Kevin so much, appreciated.

[Kevin Blighe]

In that case, I would aim to confirm it via Sanger sequencing. This variant is obviously key; otherwise, you would not be discussing it here. As such, if a clinical decision could be made based on this variant, then you require more evidence.

[Tania]

Ok, thanks Kevin so much, really much appreciated.

[WouterDeCoster]

If an answer was helpful you should upvote it, if the answer resolved your question you should mark it as accepted.