Advanced coverage predictor

[scrosby]

Hi all,

Is there some kind of a Poisson calculator out there that will predict not only mean coverage but coverage at various percents? For example, given a human genome, what mean coverage do I need to get 95% of my bases covered at greater than 30X?

Thanks!!

Seth

[MBekritsky]

Use ppois in R

If I'm not mistaken, what you're asking is essentially: if I assume a Poisson distribution, what mean rate do I need to have for at least 95% of the area under my curve above 29?

The answer to that can be found using a simple lapply or for loop in R. Here's what the for loop would look like:

for(i in 30:75) {
 # ppois returns the density under the curve, lower.tail=F
 # specifies that you want to return the area under the curve
 # at or above 30
  cat(i,ppois(29,i,lower.tail=F),"\n")
}

This prints out the mean coverage, and the percentage of bases with at least 30X coverage. The answer to your question appears to be that you should aim for mean coverage of at least 40X.

The code isn't that great, but it should give you an idea of how to answer the question.

[scrosby]

I made this. Does it look right?



It suggests that to get >95% 30-fold coverage I need an average coverage of 41...can that be right??

[MBekritsky]

Yup! I think I got that you need at least 40X average coverage.

I haven't done much WGS, mainly exome sequencing, but I'd also bear in mind that due to GC content, the genome you're sequencing, the instrument you're using, and other factors, you may not see 95% of the genome with at least 30X coverage.

[ECO]

scrosby: Fixed it for you using imgur.com. Sent you a PM.