What Is The Typical Density Of Heterozygous Alleles In A Diploid Human Genome?

Question

[Ryan Thompson]

I'm doing a theoretical project on haplotype assembly, and I need a reasonable method for generating a realistic diploid stretch of DNA. For example, if the density of heterozygous alleles is approximately one every thousand bases, then I would generate a diploid version by randomly mutating every position with probability 1/1000.

So in a typical human genome, what is the density of heterozygous alleles? Or in other words, what is the mean number of base pairs between heterozygous alleles?

Answer 1

[lh3]

It is 1e-3 for Africans and ~0.8e-3 for typical non-Africans. If a population went through a severe bottleneck recently, the heterozygosity is smaller. Uniformly generating heterozygotes in a random way is good enough for some purposes, but if you want a more realistic simulation, you should use a coalescent simulator, such as ms or MaCS. In a coalescent process, which approximates human evolution quite well, the distance between adjacent heterozygotes follows a distribution much fatter than a poisson distribution. As a result, you more frequently get a region with a lot of heterozygotes or a very long stretch of homozygosity.

Another way is to take the consensus sequence from a published personal genome, which is more realistic to some extend, but also has more bias.

[Sean Davis]

Thanks, Heng. I went for simple in my answer, but clearly there is much more detail in "doing this right".

Answer 2

[Sean Davis]

You could generate intervals between variants using a poisson distribution. You could use sampling from known snp allele frequencies for genotype probabilities, with the knowledge that "known" snps are skewed toward common snps.