Mardulyn Patrick, Vaesen Marie-Anne, Milinkovitch Michel C
Laboratory of Evolutionary Genetics, Université Libre de Bruxelles, Gosselies, Belgium.
PLoS One. 2008 Aug 13;3(8):e2960. doi: 10.1371/journal.pone.0002960.
Natural populations of known detailed past demographic history are extremely valuable to evaluate methods of historical inference, yet are extremely rare. As an alternative approach, we have generated multiple replicate microsatellite data sets from laboratory-cultured populations of a gonochoric free-living nematode, Caenorhabditis remanei, that were constrained to pre-defined demographic histories featuring different levels of migration among populations or bottleneck events of different magnitudes. These data sets were then used to evaluate the performances of two recently developed population genetics methods, BayesAss+, that estimates recent migration rates among populations, and Bottleneck, that detects the occurrence of recent bottlenecks. Migration rates inferred by BayesAss+ were generally over-estimates, although these were often included within the confidence interval. Analyses of data sets simulated in-silico, using a model mimicking the laboratory experiments, produced less biased estimates of the migration rates, and showed increased efficiency of the program when the number of loci and sampled genotypes per population was higher. In the replicates for which the pre-bottleneck laboratory-cultured populations did not significantly depart from a mutation/drift equilibrium, an important assumption of the program Bottleneck, only a portion of the bottleneck events were detected. This result was confirmed by in-silico simulations mirroring the laboratory bottleneck experiments. More generally, our study demonstrates the feasibility, and highlights some of the limits, of the approach that consists in generating molecular genetic data sets by controlling the evolution of laboratory-reared nematode populations, for the purpose of validating methods inferring population history.
已知详细过去人口统计学历史的自然种群对于评估历史推断方法极为有价值，但却极为罕见。作为一种替代方法，我们从一种雌雄异体的自由生活线虫——雷氏小杆线虫（Caenorhabditis remanei）的实验室培养种群中生成了多个重复的微卫星数据集，这些种群被设定为具有不同种群间迁移水平或不同程度瓶颈事件的预定义人口统计学历史。然后，这些数据集被用于评估两种最近开发的群体遗传学方法的性能，即用于估计种群间近期迁移率的BayesAss +和用于检测近期瓶颈事件发生的Bottleneck。BayesAss +推断出的迁移率通常是高估的，尽管这些值往往包含在置信区间内。使用模仿实验室实验的模型对计算机模拟数据集进行分析，得出的迁移率估计偏差较小，并且当每个种群的基因座数量和采样基因型数量较高时，该程序的效率有所提高。在瓶颈前实验室培养种群未显著偏离突变/漂变平衡（这是Bottleneck程序的一个重要假设）的重复实验中，仅检测到了一部分瓶颈事件。这一结果通过模拟实验室瓶颈实验的计算机模拟得到了证实。更普遍地说，我们的研究证明了通过控制实验室饲养线虫种群的进化来生成分子遗传数据集以验证推断种群历史方法的可行性，并突出了该方法的一些局限性。
