Evolutionary Biology and Ecology, Université Libre de Bruxelles, CP 160 ⁄ 12, av FD Roosevelt 50, 1050 Brussels, Belgium.
Mol Ecol. 2012 Jul;21(14):3385-90. doi: 10.1111/j.1365-294X.2012.05622.x.
Phylogenetic trees and networks are both used in the scientific literature to display DNA sequence variation at the intraspecific level. Should we rather use trees or networks? I argue that the process of inferring the most parsimonious genealogical relationships among a set of DNA sequences should be dissociated from the problem of displaying this information in a graph. A network graph is probably more appropriate than a strict consensus tree if many alternative, equally most parsimonious, genealogies are to be included. Within the maximum parsimony framework, current phylogenetic inference and network-building algorithms are both unable to guarantee the finding of all most parsimonious (MP) connections. In fact, each approach can find MP connections that the other does not. Although it should be possible to improve at least the maximum parsimony approach, current implementations of these algorithms are such that it is advisable to use both approaches to increase the probability of finding all possible MP connections among a set of DNA sequences.
系统发生树和网络都用于科学文献中显示种内 DNA 序列变异。我们应该使用树还是网络？我认为，在推断一组 DNA 序列中最简约的系统发育关系时，应该将这一过程与在图中显示该信息的问题分开。如果要包含许多替代的、同样最简约的系统发育，则网络图可能比严格一致树更合适。在最大简约框架内，当前的系统发育推断和网络构建算法都不能保证找到所有最简约 (MP) 的连接。事实上，每种方法都可以找到其他方法找不到的 MP 连接。尽管应该有可能至少改进最大简约方法，但这些算法的当前实现方式是，最好同时使用这两种方法，以增加在一组 DNA 序列中找到所有可能的 MP 连接的概率。
