On the stability of the Bayenv method in assessing human SNP-environment associations.
On the stability of the Bayenv method in assessing human SNP-environment associations.
复制标题
DOI：
10.1186/1479-7364-8-1
复制
发表时间：
2014-01-09
期刊：
影响因子：
4.5
通讯作者：
Feldman MW
中科院分区：
文献类型：
--
作者：
Blair LM;Granka JM;Feldman MW
在许多物种之间和物种内部，表型随环境梯度的变化已经被记录下来，在某些情况下，遗传变异已被证明与这些梯度有关。Bayenv是一种相对较新的方法，用于检测与环境梯度相关的多态性模式。使用贝叶斯马尔可夫链蒙特卡罗（MCMC）方法，Bayenv评估了基于观察到的中性标记的频率，将种群等位基因频率与环境变量相关的线性模型是否比零模型更可能。尽管该方法已被用于检测包括人类、植物、鱼类和蚊子在内的许多物种的环境适应性，但该MCMC算法的独立运行之间的稳定性尚未得到表征。在本文中，我们探讨了运行之间结果的可变性和促成它的因素。使用来自全球60个人群样本的全基因组单核苷酸多态性（SNP）数据，按照先前的Bayenv方法应用，对Bayenv程序进行了独立运行。为了评估影响该方法稳定性的因素，我们使用了不同数量的MCMC迭代，并分析了第二个修改后的数据集，该数据集排除了两个具有极端气候变量的西伯利亚种群。在任意两次运行之间，贝叶斯因子与经验p值尾部snp重叠之间的相关性惊人地低。经验尾部基因与非基因snp的富集比经验p值更为稳健；然而，对某些环境变量的富集的重要性在不同的运行中仍然有所不同，这与以前发表的结论相矛盾。具有更多MCMC迭代次数的运行略微降低了运行到运行的可变性，并且排除西伯利亚种群对运行的稳定性没有很大的影响。由于运行间的高可变性，我们建议不要仅基于一次运行的Bayenv算法得出全基因组适应模式的结论，并建议在解释仅使用一次运行的先前研究时谨慎。下一步，我们建议进行多次独立的Bayenv运行，并在运行之间平均Bayes因子，以获得更稳定可靠的结果。有了这些改进，未来使用Bayenv方法发现物种内部的环境适应将更加准确，可解释，并且易于在研究之间进行比较。
Phenotypic variation along environmental gradients has been documented among and within many species, and in some cases, genetic variation has been shown to be associated with these gradients. Bayenv is a relatively new method developed to detect patterns of polymorphisms associated with environmental gradients. Using a Bayesian Markov Chain Monte Carlo (MCMC) approach, Bayenv evaluates whether a linear model relating population allele frequencies to environmental variables is more probable than a null model based on observed frequencies of neutral markers. Although this method has been used to detect environmental adaptation in a number of species, including humans, plants, fish, and mosquitoes, stability between independent runs of this MCMC algorithm has not been characterized. In this paper, we explore the variability of results between runs and the factors contributing to it. Independent runs of the Bayenv program were carried out using genome-wide single-nucleotide polymorphism (SNP) data from samples from 60 worldwide human populations following previous applications of the Bayenv method. To assess factors contributing to the method's stability, we used varying numbers of MCMC iterations and also analyzed a second modified data set that excluded two Siberian populations with extreme climate variables. Between any two runs, correlations between Bayes factors and the overlap of SNPs in the empirical p value tails were surprisingly low. Enrichments of genic versus non-genic SNPs in the empirical tails were more robust than the empirical p values; however, the significance of the enrichments for some environmental variables still varied among runs, contradicting previously published conclusions. Runs with a greater number of MCMC iterations slightly reduced run-to-run variability, and excluding the Siberian populations did not have a large effect on the stability of the runs. Because of high run-to-run variability, we advise against making conclusions about genome-wide patterns of adaptation based on only one run of the Bayenv algorithm and recommend caution in interpreting previous studies that have used only one run. Moving forward, we suggest carrying out multiple independent runs of Bayenv and averaging Bayes factors between runs to produce more stable and reliable results. With these modifications, future discoveries of environmental adaptation within species using the Bayenv method will be more accurate, interpretable, and easily compared between studies.
登录
查看更多内容
影响因子：
4.5
作者：
Hancock, Angela M.;Witonsky, David B.;Gordon, Adam S.;Eshel, Gidon;Pritchard, Jonathan K.;Coop, Graham;Di Rienzo, Anna
通讯作者：
Di Rienzo, Anna
影响因子：
3.3
作者：
Fang, Zhou;Pyhajarvi, Tanja;Ross-Ibarra, Jeffrey
通讯作者：
Ross-Ibarra, Jeffrey
DOI：
10.1073/pnas.0914625107
发表时间：
2010-05-11
影响因子：
11.1
作者：
Hancock, Angela M.;Witonsky, David B.;Di Rienzo, Anna
通讯作者：
Di Rienzo, Anna
影响因子：
3.3
作者：
Coop, Graham;Witonsky, David;Pritchard, Jonathan K.
通讯作者：
Pritchard, Jonathan K.
解偶联蛋白的群体遗传分析支持 UCP3 在人类抗寒性中的作用。
影响因子：
10.7
作者：
Hancock AM;Clark VJ;Qian Y;Di Rienzo A
通讯作者：
Di Rienzo A