allele frequency from different read depths

Question

[rthapa]

Hello, I want to estimate allele frequency at different read depth (10X, 20X, 30X, and so on). I have variants call output in vcf format. I also estimated read depths for all SNPs. I am thinking to subsets the vcf at different depth for e.g. 10X, 20X etc. I am not sure if this is a correct way to proceed. Is there any better way to estimate allele frequency at different depths? Any suggestion will be appreciated. Thanks

Answer 1

[pbpanigrahi]

If you are talking about Variant AF (VAF) then it is independent of read depth.

VAF = (forward non-ref + reverse non-ref alleles) / (forward ref alleles + reverse ref + forward non-ref + reverse non-ref alleles)

So irrespective of depth at a given position, VAF can be calculated by above formula. No depth information is used. So its relative term.

So your approach of filtering vcf at different depth would be fine because first you are filtering at a given depth and analyzing VAF. You don't have to recalculate VAF.

This thread may be useful

Hope this is what expects. If not please elaborate more with some dummy example vcf.

[rthapa]

Is it not that the sequencing depth affects the counts of ref allele and alternate allele? I think the depth could affect the VAF.

[pbpanigrahi]

If you compare below variants,

variant1: Depth = 50, and nonref allele freq = 25, VAF = 25/50 = 0.5
variant1: Depth = 100, and nonref allele freq = 50, VAF = 50/100 = 0.5
variant3: Depth = 10, and nonref allele freq = 5, VAF = 5/10 = 0.5

So u see VAF for all three cases is 0.5 even if depth are different.

However for variant2, confidence is stronger than variant1 and 3 due to higher depth.

So when you apply different depth filter, basically you are segregating the variants in different confidence bins (higher depth, higher confidence)

Hope it was useful.

[rthapa]

Thank you. How about the minor allele frequency? Can we still detect with low sequence depth?

[pbpanigrahi]

One educational note, if you find any answers/suggestion useful, in this forum, you should upvote. This way the answer gets highlighted. So in future if someone wants to know about it, they can directly look into the right answer.

Coming to your query, yes we can detect MAF at low sequencing depth but higher is better to gain confidence on the variant call. What low depth you are expecting.

You can find many articles on google

you can refer some papers which work in similar domain.

you can search for MAF in this forum.