How To Estimate Genome Size Using K-Mer Coverage

Question

[GAO Yang]

Hi, I just got a genome assembly (de novo), and I want to estimate the genome size. According to some published papers, it can be done using the k-mer coverage. But I am not quite following this one: How to cut the genome to chosen K-mer ? And how to summary the K-mer abundance and plot a Poisson distribution, just like in the papers?

Could anybody provide any software name,or Perl module, or even some Pseudo-code? Thanks for your help!

[Gjain]

can you mention the paper you are referring to...

[GAO Yang]

Sure, for example "The genome of the domesticated apple" Nature genetics 2010, supplementary Page9

[Michael]

There is no mention of k-mer in supplementary, I believe you confused kmer with read. Otherwise you should read about the Lander-Waterman statistics. The only difficult part is the fitting of the poisson distribution mentioned in the article.

[Arun]

+1. Even though I wrote about K-mers, this article seems to have nothing about K-mers.

[GAO Yang]

Sorry about that~ I mixed them up~ But plz check this one:"Genome sequencing reveals insights into physiology and longevity of the naked mole rat" supplementary P3,doi:10.1038/nature10533

[Arun]

In that case, you should know that the first link I have pointed to, which explains K-mer coverage related to genome size, is a tool (quake) to obtain all what you've asked for.

Answer 1

[Arun]

Regarding your first question, about K-mer coverage and genome size, there seems to be different methods/algorithms different softwares use. EDIT: The idea in general is explained very well here. I don't follow what you mean by "how to cut the genome to chosen K-mer, could you please elaborate? To speculate about the K-mer distribution, it is done by obtaining the histogram/density plot by binning K-mers over different coverage. You'll see a smooth curve that resembles a poisson distribution. If there is bias in your k-mer distribution, you'll normally see an initial peak like this, from which you can decide the cut-off of coverage that you'll have to use to get rid of this bias.

[GAO Yang]

Yeah, This is what I need! Thanks for that, I am going on with it @_@

[GAO Yang]

By the way, do you know how to apply this software on the Color-space reads (SOLiD output)? Maybe I need post another question about it ~ :)