Estimating heterozygosity from kmer frequency distribution

[MeganS]

Is there a program that can estimate the heterozygosity of a sample using the kmer frequency distribution of the raw reads? I have whole genome, Illumina data (100bp PE reads, from 300bp fragments). The kmer frequency plot has a clear bimodal distribution, so I can get a rough estimate by eyeballing the areas under the curves for the two peaks. I am hoping to find a more robust method and more automated since I have over 100 samples.

[ebioman]

push

Actually I have no responds neither, I am afraid.
I am just asking myself the same question and wondered whether you were able to solve that question ?

[Melissa]

Perhaps you want to look into Ka/Ks estimation.

[WormSeeq]

I just came across this paper on arxiv "Estimation of genomic characteristics by analyzing k-mer frequency in de novo genome projects"

http://arxiv.org/abs/1308.2012

I have not tried their tool though! It is available at ftp://ftp.genomics.org.cn/pub/gce/
Best,
~wormSeeq.

[Brian Bushnell]

Hmmm, I wrote a program that does this. Well, two, actually. Their usage is about the same.

khist.sh in=reads.fq khist=khist.txt peaks=peaks.txt
or
kmercountexact.sh in=reads.fq khist=khist.txt peaks=peaks.txt

The first uses approximate counts, while the second uses exact counts (and thus potentially more memory). The peaks file header contains estimates of genome size and heterozygousity. You can also add the flag "ploidy=2" for diploid organisms, so that it won't need to autodetect the ploidy (and thus potentially make a mistake).

These are both distributed with BBTools.