Fearnhead P
Department of Statistics, University of Oxford, Oxford, United Kingdom.
Theor Popul Biol. 2001 Jun;59(4):263-79. doi: 10.1006/tpbi.2001.1514.
We consider using the ancestral selection graph (ASG) to simulate samples from population genetic models with selection. Currently the use of the ASG to simulate samples is limited. This is because the computational requirement for simulating samples increases exponentially with the selection rate and also due to needing to simulate a sample of size one from the population at equilibrium. For the only case where the distribution of a sample of size one is known, that of parent-independent mutations, more efficient simulation algorithms exist. We will show that by applying the idea of coupling from the past to the ASG, samples can be simulated from a general K-allele model without knowledge of the distribution of a sample of size one. Furthermore, the computation involved in generating such samples appears to be less than that of simulating the ASG until its ultimate ancestor. In particular, in the case of genic selection with parent-independent mutations, the computational requirement increases only quadratically with the selection rate. The algorithm is demonstrated by simulating samples at a microsatellite locus.
我们考虑使用祖先选择图（ASG）来模拟具有选择作用的群体遗传模型中的样本。目前，使用ASG来模拟样本存在局限性。这是因为模拟样本的计算需求会随着选择率呈指数增长，还因为需要在平衡状态下从群体中模拟出大小为一的样本。对于唯一已知大小为一样本分布的情况，即与亲本无关的突变情况，存在更有效的模拟算法。我们将表明，通过将过去耦合的思想应用于ASG，可以在不知道大小为一样本分布的情况下，从一般的K等位基因模型中模拟样本。此外，生成此类样本所涉及的计算似乎比模拟ASG直至其最终祖先的计算量要小。特别是，在具有与亲本无关突变的基因选择情况下，计算需求仅随选择率呈二次方增长。通过在微卫星位点模拟样本对该算法进行了演示。
