# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz SplitsTree4 is the leading application for computing unrooted phylogenetic networks from molecular sequence data. Given an alignment of sequences, a distance matrix or a set of trees, the program will compute a phylogenetic tree or network using methods such as split decomposition, neighbor-net, consensus network, super networks methods or methods for computing hybridization or simple recombination networks. SplitsTree4 is an all-new implementation of the SplitsTree3 software written in Java. Superficially, this new program is similar to previous versions of this software. However, there are substantial differences. The new program has many new features, see the manual for details. If you use SplitsTree in work in any way, then please cite the following paper: D. H. Huson and D. Bryant, Application of Phylogenetic Networks in Evolutionary Studies, Mol. Biol. Evol., 23(2):254-267, 2006. SplitsTree4 是从分子序列数据计算无根系统发育网络的主要应用。给定比对好的序列、距离矩阵或一组树，程序将使用分裂分解、邻域网络、共识网络、超级网络方法或计算杂交或简单重组网络的方法运算系统树或网络。 SplitsTree4 是用 Java 对 SplitsTree3 重新编写的全新应用。从表面上看，这个新程序与以前版本类似。然而，两者之间存在着实质性的差异。新程序有许多新功能，有关详细信息，请参阅使用手册。 如果您在工作中以任何方式使用 SplitsTree ，请引用以下文献： D. H. Huson and D. Bryant, Application of Phylogenetic Networks in Evolutionary Studies, Mol. Biol. Evol., 23(2):254-267, 2006. http://ab.inf.uni-tuebingen.de/software/splitstree4/
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz Statistical parsimony The statistical parsimony algorithm begins by estimating the maximum number of differences among haplotypes as a result of single substitutions (i.e. those that are not the result of multiple substitutions at a single site) with a 95% statistical confidence. This number is called the parsimony limit (or parsimony connection limit). After this, haplotypes differing by one change are connected, then those differing by two, by three and so on, until all the haplotypes are included in a single network or the parsimony connection limit is reached. The statistical parsimony method emphasizes what is shared among haplotypes that differ minimally rather than the differences among the haplotypes and provides an empirical assessment of deviations from parsimony. This method allows the identification of putative recombinants by looking at the spatial distribution in the sequence of the homoplasies defined by the network 统计简约法 统计简约算法首先估计单倍型之间的最大差异数，这是由 95% 的统计置信度的单个替换（即那些不是单个位点上多个替换的结果）引起的。这个数字被称为简约限制（或简约连接限制）。在此之后，将有一个不同点的单倍型连接起来，然后将两个不同点的单倍型连接起来，依次类推，直到所有的单倍型都包含在一个网络中或达到简约连接限制为止。统计简约法强调了最小差异单倍型之间的共同点，而不是单倍型之间的差异，并提供了对简约偏差的经验评估。这种方法可以通过观察网络定义的异源同形序列的空间分布来识别假定的重组。 Posada D , Crandall K A . Intraspecific gene genealogies: trees grafting into networks . Trends in Ecology and Evolution, 2001, 16(1):0-45.
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz Median-joining networks The median-joining network method begins by combining the MINIMUM-SPANNING TREES (MSTs) within a single network. With a parsimony criterion, median vectors (which represent MISSING INTERMEDIATES) are added to the network. Median-joining networks can handle large data sets and multistate characters. It is an exceptionally fast method that can analyze thousands of haplotypes in a reasonable amount of time and can also be applied to amino acid sequences. However, it requires the absence of recombination, which restricts the application of this method at the population level. 中间连接网 中间连接网络方法首先将最小生成树（ MST ）组合到单个网络中。用简约的标准，中值向量（代表缺失的中间物）被添加到网络中。中间连接网络可以处理大数据集和多状态字符。这是一种非常快速的方法，可以在合理的时间内分析数千个单倍型，也可以应用于氨基酸序列。然而，它要求不存在重组，这限制了该方法在群体层面的应用。 Posada D , Crandall K A . Intraspecific gene genealogies: trees grafting into networks . Trends in Ecology and Evolution, 2001, 16(1):0-45.
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz Median networks In the median-network approach, sequences are first converted to binary data and constant sites are eliminated. Each split is encoded as a binary character with states 0 and 1. Sites that support the same split are grouped in one character, which is weighted by the number of sites grouped. This leads to the representation of haplotypes as 0–1 vectors. Median or consensus vectors are calculated for each triplet of vectors until the median network is finished. For 30 haplotypes, the resulting median networks are impractical to display, owing to the presence of high-dimensional hypercubes. Fortunately, the network can be reduced (i.e. some loops can be solved) using predictions from coalescent theory (Box 2). All the most parsimonious trees are guaranteed to be represented in a median network. Although mainly aimed at mtDNA data, median networks can be estimated from other kinds of data, as long as the data are binary or can be reduced to binary data. 中值网络 在中值网络方法中，先将序列转换为二进制数据，然后消除保守位点。每个拆分都被编码为具有状态 0 和 1 的二进制字符。支持相同拆分的站点被分组为一个特征，该特征由分组的位点数加权。这导致单倍型的表现为 0-1 向量。对每三组向量计算中值向量或一致向量，直到中值网络完成。对于大于 30 个单倍型，由于存在高维超立方体，因此产生的中值网络无法显示。幸运的是，可以使用溯祖理论（框注 2 ）的预测来减少网络（即可以解决一些循环）。所有最简约的树都可以用中值网络表示。尽管中值网络主要针对线粒体 DNA 数据，但只要数据是二进制的或可以简化为二进制数据，就可以从其他类型的数据中估计出中值网络。 Posada D , Crandall K A . Intraspecific gene genealogies: trees grafting into networks . Trends in Ecology and Evolution, 2001, 16(1):0-45.
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz Split decomposition Any data set can be partitioned into sets (not necessarily of equal size) of sequences or ‘splits’. A network can be built by taking in turn those splits defined by the characters and combining them successively. Each split will define a branch connecting the two partitions delimited by the split. When splits are incompatible (i.e. they define contradictory groupings) a loop is introduced to indicate that there are alternative splits. The split decomposition method is fast, which means that a reasonable number of haplotypes (50) can be analyzed; that it can be applied to nucleotide or protein data; and that it allows for the inclusion of models of nucleotide substitution or amino acid replacement. The method is suitable also to bootstrap evaluation. 分割分解 任何数据集都可以分割成更小的数据集（大小不一定相等）或“分割”。反过来，可以通过这些由具体特征定义的“分割”迭代组合成网络。每次拆分将定义一次分支，连接由拆分分隔的两个分区。当拆分不兼容（即它们定义了相互矛盾的分组）时，将引入一个循环来指示存在可选拆分。分割分解法快速，这意味着可以分析相当多的单倍型（ 50 ）；它可以应用于核苷酸或蛋白质数据；允许包含核苷酸替代或氨基酸替代模型。该方法也适用于自举检验。 Posada D , Crandall K A . Intraspecific gene genealogies: trees grafting into networks . Trends in Ecology and Evolution, 2001, 16(1):0-45.
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz Pyramids The pyramids technique is an extension of the hierarchical clustering framework. Whereas traditional hierarchical methods such as Unweighted Pair Group with Arithmetic Means (UPGAM) represent a nested set of nonoverlapping clades, pyramids represent a set of clades that can overlap without necessarily being nested. The input data is a (Robinsonian) distance matrix (Box 1). The pyramid is obtained by using agglomerative algorithms. By allowing overlapping clusters, pyramids can be used to represent reticulate events, although these events are only allowed to be placed among terminal nodes that are sister taxa. 金字塔 金字塔拓扑结构是层次聚类框架的扩展。传统的层次方法，如不加权算术平均组对方法（ UPGAM ）表示一组嵌套的非重叠族，而金字塔表示一组可以重叠而不必嵌套的族。输入数据为（ Robinsonian ）距离矩阵（框注 1 ）。金字塔是用凝聚算法得到的。通过允许重叠的集群，金字塔可以用来表示网状事件，尽管这些事件只允许放置在作为姊妹分类群的终端节点之间。 Posada D , Crandall K A . Intraspecific gene genealogies: trees grafting into networks . Trends in Ecology and Evolution, 2001, 16(1):0-45.
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz Solution: network methods Phylogenetic methods that allow for persistent ancestral nodes, multifurcations and reticulations are needed to take these population phenomena into account. The advantage of networks over strictly bifurcating trees for estimating within-species relationships now becomes obvious. Networks can account effectively for processes acting at the species level and they might be able to incorporate predictions from population genetics theory (Box 2). In addition, networks provide a way of representing more of the phylogenetic information present in a data set (Fig. 2). For example, the presence of loops in a network might indicate recombination. In other cases, loops are the product of homoplasies and precisely indicate the occurrence of reverse or parallel mutations. Most network methods are distance methods, with the common idea of minimizing (with some specific restrictions) the distances (number of mutations) among haplotypes. In other cases, the likelihood function is maximized. 解决方案：网络方法 研究种群想象的系统发育方法需要考虑持续祖先节点、多岐分叉和网状结构。与严格的二岐分叉树相比，网络方法用于估计种内关系的优势现在变得越发明显了。网络可以有效地解释在物种层面上作用的过程，并且它们可能将群体遗传学理论的预测整合进来（框注 2 ）。此外，网络提供了一种表示数据集中存在的更多系统发育信息的方法（图 2 ）。例如，网络中的闭环可能表示重组。在其他情况下，闭环是趋同进化的产物，精确地指示了回复突变或平行突变的发生。大多数网络方法都是距离法，其共同的思想是最小化单倍型之间的距离（突变数）。在其他情况下，似然函数最大化。 Posada D , Crandall K A . Intraspecific gene genealogies: trees grafting into networks . Trends in Ecology and Evolution, 2001, 16(1):0-45.
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz Reticulation Evolutionary processes commonly acting at the population level, such as recombination between genes and hybridization between lineages, and HOMOPLASY (Box 1), generate reticulate relationships within the population. Traditional methods, based on bifurcating trees, make no explicit allowance for such reticulations. Instead, for instance, maximum parsimony deals with ambiguities arising from homoplasy by simply selecting a tree that minimizes the number of assumptions of parallel, convergent or reversing mutations without showing where these might have occurred. Recombinants are also typically forced into a nonreticulating tree topology, in which, in some fortunate instances, they might occupy positions intermediate between two clusters. In other cases, the recombinant will be placed in a basal lineage to the clade that includes its most derived parent3,4. 网状结构 通常作用于群体层面的进化过程，如基因重组和谱系间的杂交以及异源同形（框注 1 ），在群体内产生网状关系。传统的方法是基于分叉树，对这种网状结构没有明确的考虑。相反，例如，最大简约法通过简单地选择一棵树来处理由同形性引起的歧义，该树可以让并行、收敛或逆转突变的假设数量最小化，而不显示这些突变可能发生在哪里。重组子通常也被迫进入非连续的树拓扑结构，在某些幸运的情况下，它们可能占据两个集群之间的中间位置。在其他情况下，重组子将被置于包含其亲本的分支的基础谱系中。 Posada D , Crandall K A . Intraspecific gene genealogies: trees grafting into networks . Trends in Ecology and Evolution, 2001, 16(1):0-45.
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz Problems with interspecific methods at the intraspecific level Evolutionary relationships above and below the species level are different in nature. Relationships between genes sampled from individuals belonging to different species (phylogeny sensu stricto) are hierarchical. This is because they are the product of reproductive isolation and population fission over longer timescales, during which mutation combined with population divergence led to the fixation of different alleles and, ultimately, to nonoverlapping gene pools. By contrast, relationships between genes sampled from individuals within a species (sometimes called TOKOGENY 2 ) are not hierarchical, because they are the result of sexual reproduction, of smaller numbers of relatively recent mutations and, frequently, of recombination (Fig. 1). More traditional methods developed to estimate interspecific relationships, such as maximum likelihood, maximum parsimony and minimum evolution, cannot properly take account of the fact that, at the population level, several phenomena violate some of their assumptions. This leads to poor resolution or inadequately portrays genealogical relationships 种间研究方法去研究种上和种下的进化关系所存在的问题在本质上是不同的。从属于不同物种个体中提取的基因之间的关系是级联式的，这是狭义的系统发育。它们是在较长时间内生殖隔离和种群分裂的产物，在此过程中，突变和种群分化导致不同等位基因的固定，最终导致互不重叠的基因库。相比之下，从一个物种内的不同个体中提取的基因之间的关系（有时称为 TOKOGENY ）不是级联关系，因为它们是有性生殖的结果，相对较新的突变数量较少，而且经常是重组的结果（图 1 ）。为了估计种间关系而开发的传统方法，如最大似然法、最大简约法和最小进化法，不能适当地考虑到在种群水平上，一些现象违反了它们的一些假设。这会导致解决方案不好或无法充分描述谱系关系。 Posada D , Crandall K A . Intraspecific gene genealogies: trees grafting into networks . Trends in Ecology and Evolution, 2001, 16(1):0-45.
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz Intraspecific gene evolution cannot always be represented by a bifurcating tree. Rather, population genealogies are often multifurcated, descendant genes coexist with persistent ancestors and recombination events produce reticulate relationships. Whereas traditional phylogenetic methods assume bifurcating trees, several networking approaches have recently been developed to estimate intraspecific genealogies that take into account these population level phenomena. During the past decade, the explosion of molecular techniques has led to the accumulation of a considerable amount of comparative genetic information at the population level. At the same time, recent advances in population genetics theory, especially coalescent theory, have generated powerful tools for the analysis of intraspecific data. These two developments have converted intraspecific phylogenies into useful tools for testing a variety of evolutionary and population genetic hypotheses. Several phylogenetic methods, especially NETWORK (see Glossary) approaches, have been developed to take advantage of the unique characteristics of intraspecific data. In this article, we summarize some population genetics principles, explain why networks are appropriate representations of intraspecific genetic variation, describe and compare available methods and software for network estimation, and give examples of their application 种内基因进化不总能用分叉树来表示。相反，种群谱系通常是多歧分叉的，后代基因与长寿的祖先共存，重组事件产生网状关系。 传统的系统发育方法假设树是分叉的，最近已经开发出几种网络方法来估计种内谱系，将这些种群级别现象考虑进去。不是这些人口层面的现象。 在过去的十年中，分子技术的爆炸性发展导致了在种群水平上积累了大量的比较遗传信息。与此同时，群体遗传学理论的最新进展，特别是溯祖理论，为分析特定种内数据提供了强有力的工具。这两个进步已将种内系统发育转化为检验各种进化变异和种群遗传假说的有用工具。几种系统发育方法，特别是 NETWORK （见词汇表）方法，已被开发出来，以利用种内特定数据的独特特征。在本文中，我们总结了一些群体遗传学原理，解释了为什么网络是种内遗传变异的适当表示，描述并比较了现有的网络估计方法和软件，并举例说明了它们的应用。 Posada D , Crandall K A . Intraspecific gene genealogies: trees grafting into networks . Trends in Ecology and Evolution, 2001, 16(1):0-45.
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz In a haplotype network, how to determine which one is the root or where is the origin of the spp. and where does it invaded from. Coalescent theory predicts that ancestral haplotypes will be the haplotypes most frequently sampled. They are usually more broadly distributed geographically, and more likely to have a larger number of connections or descendant nodes/haplotypes. If I recall correctly, TCS calculates the haplotype that has the highest outgroup probability (potentially the most ancestral haplotype) and displays it as a square, whereas other haplotypes are displayed as ovals. Regarding from where the species invaded: Additional geographic, molecular dating, and natural history data would be helpful in determining the answer to that question. You might consider the geographic locality/region of the haplotype with the highest outgroup probability. if that haplotype is concentrated in a certain geographic area, it may indicate a glacial refugium or point of invasion. This would depend on the scale of your sampling, mobility of organism, etc. 在一个单倍型网络中，如何确定哪一个单倍型是根，或者物种的起源在哪里，以及它从哪里入侵。 根据溯祖理论预测，祖先的单倍型将是最常被抽样到的单倍型。它们通常在地理上分布更广，并具具有更多的连接或后代节点 / 单倍型。如果我记得正确的话， TCS 计算出具有最高外群（连接）概率的单倍型（可能是最原始的单倍型），并将其显示为一个正方形，而其他单倍型则显示为椭圆形。 关于物种入侵的来源： 额外的地理、分子年代测定和自然历史数据将有助于确定这个问题的答案。您可以考虑具有最高的外群概率单倍型的地理位置 / 区域。如果这种单倍型集中在某个地理区域，可能表明是冰川避难所或入侵点。这将取决于你的取样规模、生物体的流动性等。 https://www.researchgate.net/post/How_do_I_explain_the_haplotype_network_in_population_genetics
# 编者信息 熊荣川 明湖实验室 xiongrongchuan@126.com http://blog.sciencenet.cn/u/Bearjazz TCS analysis obtained five haplotype networks at 95% probability (connection limit = 15), corresponding to the five major lineages obtained by phylogenetic analyses (Figure 3). No alternative connections between haplotypes (‘loops’) were observed, indicating that no homoplasy was involved in the network. TCS 分析以 95% 的概率获得 5 个单倍型网络（连接极限 =15 步），与系统发育分析获得的 5 个主要支系相对应（图 3 ）。没有观察到单倍型（“环路”）之间的替代连接，这表明网络中没有非同源相似（平行演化）。 Li Z, Yu G, Rao D, et al. Phylogeography and Demographic History of Babina pleuraden (Anura, Ranidae) in Southwestern China . Plos One, 2012, 7(3):e34013.
标签: 单倍型网络
