Deng H W, Fu Y X
Human Genetics Center, School of Public Health, University of Texas, Houston 77225, USA.
Genetics. 1996 Nov;144(3):1271-81. doi: 10.1093/genetics/144.3.1271.
Multiple hits at some sites of human mitochondrial DNA sequences suggest that the commonly assumed infinite-sites model can be violated. Under the neutral Wright-Fisher model without recombination and population subdivision, we investigated, by computer simulations, the effect of multiple hits on the estimation of the essential parameter theta = 4Nmu by FU's UPBLUE procedure. We found that with moderate mutation rate heterogeneity, UPBLUE performs very well in terms of unbiasness and efficiency. Under extreme mutation rate heterogeneity, if sample size is reasonably large (e.g., > 60), UPBLUE is still very satisfactory; otherwise we developed a new correction equation. Given knowledge of the degree of mutation rate heterogeneity, the performance of UPBLUE with the new correction equation was tested to be fairly satisfactory: there is almost no bias and the sampling variance is only slightly higher than the theoretical minimum variance. Thus, with an appropriate correction, UPBL.UE is relatively robust to the multiple hits. In genealogies reconstructed by UPGMA, we found that the total length of branches directly linked to the tips is underestimated, and those far away tend to be overestimated, while the total length of all branches is not biased.
人类线粒体DNA序列某些位点的多次击中表明，通常假定的无限位点模型可能会被违反。在无重组和群体细分的中性赖特 - 费希尔模型下，我们通过计算机模拟研究了多次击中对FU的UPBLUE程序估计基本参数θ = 4Nμ的影响。我们发现，在适度的突变率异质性条件下，UPBLUE在无偏性和效率方面表现非常出色。在极端突变率异质性条件下，如果样本量足够大（例如，> 60），UPBLUE仍然非常令人满意；否则我们开发了一个新的校正方程。在已知突变率异质性程度的情况下，带有新校正方程的UPBLUE的表现相当令人满意：几乎没有偏差，抽样方差仅略高于理论最小方差。因此，经过适当校正后，UPBLUE对多次击中相对稳健。在通过UPGMA重建的系统发育树中，我们发现直接与末端相连的分支的总长度被低估，而那些较远的分支往往被高估，而所有分支的总长度没有偏差。
