we are planning to do some whole genome sequencing of human subjects to look at signatures of selection and to check out associations of known SNPs to extended phenotypes. We have quotes for 15x depth and 30x depth sequencing and obviously 30x is much more expensive. so my question is what exactly do you lose at 15x depth sequencing in humans. Most importantly can you put a figure on it? For example, if we are going to miss SNPs entirely then what % will we miss? thanks in advance for any answers.
Hi John,
There are two aspects to this: how much local depth is needed for a good SNV call, and how global depth translates to local depth.
Alison Meynert's attached papers will help with the first point. Her conclusions agree with our own results, that around 13-15X local depth (depth at a given locus) is needed for a confident (>95% sensitivity) het detection.
The other part of the equation is how global depth corresponds to local depth. This depends on the technology and the characteristics of the locus, but in our hands (TruSeq Nano, HiSeq X), we find that 30X global depth gives >=15X coverage at 97.4% of exonic loci, whereas 15X global depth gives only 60.3% coverage. This work is not yet published, but I intend to have it on bioRxiv in a couple of weeks.
Other considerations may factor in (eg. the feasibility of imputation, or the tradeoff between sample count and individual sample coverage), but I'm not an expert on these aspects.
Hope that helps,
Mark
Article Quantifying single nucleotide variant detection sensitivity ...
Article Variant detection sensitivity and biases in whole genome and...
There is a recent report in PNAS of more than 10,500 human genomes completed to 30 - 40x coverage. I am not sure if this data is publicly avaiable or not. Even if it is not public the authors may allow you to collaborate with them on a study with their data.
http://www.pnas.org/content/early/2016/10/03/1613365113.full.pdf?with-ds=yes
- Badges
- Science topic
- Similar topics
- Filing