Using Data-Display Networks for Exploratory Data Analysis in Phylogenetic Studies
Using Data-Display Networks for Exploratory Data Analysis in Phylogenetic Studies
复制标题
DOI：
10.1093/molbev/msp309
复制
发表时间：
2010-05-01
影响因子：
10.7
通讯作者：
Morrison, David A.
中科院分区：
文献类型：
--
作者：
Morrison, David A.
探索性数据分析（EDA）是生物学数据分析中一个经常被低估的部分。它包括在对手头的科学问题进行最终分析之前，对数据的特征进行评估。对于系统发育分析，一个有用的EDA工具是数据显示网络。这种类型的网络旨在显示数据集中的任何字符（或树）冲突，而无需预先假设这些冲突的原因。这些冲突可能是由1)数据收集或分析中的方法问题，2)同源性，或3)某种水平基因流动引起的。在这里，我将使用拆分网络探索13个已发布的数据集，作为在EDA中使用数据显示网络的示例。在每种情况下，我都对所提供的数据执行了原始EDA，以突出显示结果网络的各个方面，这些方面对于解释系统发育非常重要。在每种情况下，至少有一个重要的点（可能被原作者遗漏）可能会影响系统发育分析。我的结论是，EDA应该在系统发育分析中发挥比现在更大的作用。
Exploratory data analysis (EDA) is a frequently undervalued part of data analysis in biology. It involves evaluating the characteristics of the data "before" proceeding to the definitive analysis in relation to the scientific question at hand. For phylogenetic analyses, a useful tool for EDA is a data-display network. This type of network is designed to display any character (or tree) conflict in a data set, without prior assumptions about the causes of those conflicts. The conflicts might be caused by 1) methodological issues in data collection or analysis, 2) homoplasy, or 3) horizontal gene flow of some sort. Here, I explore 13 published data sets using splits networks, as examples of using data-display networks for EDA. In each case, I performed an original EDA on the data provided, to highlight the aspects of the resulting network that will be important for an interpretation of the phylogeny. In each case, there is at least one important point (possibly missed by the original authors) that might affect the phylogenetic analysis. I conclude that EDA should play a greater role in phylogenetic analyses than it has done.