主流分子钟定年方法的原理、误差来源和使用建议

Principles,error sources and application suggestions of prevailing molecular dating methods

近年来,分子钟定年方法(molecular dating methods)得以广泛运用,为宏观进化研究尤其是生物多样性及其格局形成历史的相关研究提供了不可或缺且十分详尽的进化时间框架。贝叶斯方法(Bayesianmethods)和马尔可夫链蒙特卡罗方法(Markov chain Monte Carlo)可容纳多维度、多类型的数据和参数设置,因此以BEAST、PAML-MCMCTree等软件为代表的贝叶斯节点标记法(Bayesian node-dating methods)逐渐成为分子钟定年方法中最为广泛使用的类型。贝叶斯框架的优势之一在于其可以利用复杂模型考虑各种不确定性因素,但是该类方法中各类模型和参数的设置都可能引入误差,从而影响进化分化时间估算的可靠性。本文介绍了贝叶斯分子钟定年方法的原理和主要类型,并以贝叶斯节点标记法为例,重点讨论了分子钟模型、化石标记的选择与放置、采样频率及化石标记点年龄先验分布等因素对节点定年的影响;提供了贝叶斯时间树构建软件的使用建议、节点年龄的讨论原则和不同模型下时间树的比较方法,针对常见的引起节点年龄潜在高估和低估风险的情况作了分析并给出了合理化建议。我们认为,合理整合多种贝叶斯方法和模型得出的结果并从中择优,能够提高定年结果的可靠性;研究人员应对时间树构建结果与其参数设置的关系开展讨论,从而为其他学者提供参考;化石记录的更新与分子钟定年方法的改进应同步不断跟进。

Background&Aims:Molecular dating methods have been applied widely in recent years and provide an indispensable and detailed evolutionary timescale for macroevolutionary researches,particularly for studies on the evolutionary history of biodiversity patterns.Bayesian methods and Markov chain Monte Carlo methods can accommodate multi-dimensional and various type of data and parameter settings,which have helped the node-dating methods implemented in softwares such as BEAST,PAML-MCMCTree to become the most widely used molecular dating methods.One of the advantages of Bayesian frameworks is that they can employ complex models to consider a variety of uncertainty factors to make more accurate estimations of evolutionary divergence times.Progress:We review the principles and main types of Bayesian molecular dating methods and use Bayesian node-dating methods as an example to discuss potential errors in molecular clock models,selection and placement of fossil calibrating points,frequency of sampling,and setting a prior distribution for node calibrations based on fossils.We further describe advantages associated with different Bayesian time tree reconstruction software packages,the discussing principle of node age,and the comparison method of time tree under different models.We also provide suggestions for overcoming the challenges of overestimation and underestimation bias of node ages.Integration of the output of various Bayesian methods and models and selection of the best among them often improve the reliability of molecular dating results.Prospect:Researchers should explicitly assess the relationship between model output of time tree construction and model parameter settings,which increases transparency and provides documentation and reference for future researches.We recommend that future research simultaneously focus on updating fossil records and improving molecular dating methods.