Meise Botanic Garden, Nieuwelaan 38, Meise, Belgium.
Unidad Ejecutora Lillo, UEL (CONICET - Fundación Miguel Lillo), Miguel Lillo 251, 4000, San Miguel de Tucumán, Argentina.
Cladistics. 2024 Aug;40(4):456-467. doi: 10.1111/cla.12571. Epub 2024 Feb 12.
Wheeler (Cladistics 2023, 39, 475) recently suggested that the issues with inapplicable characters in phylogenetic analysis can be dealt with directly by treating observed absences of a feature not in a separate absence/presence character but as insertion/deletion events in a complex character that describes the feature in all its variation; and that this dynamic homology view can be achieved by imposing a sequence or linear order on a set of characters and by analysing the resulting sequence character using custom alphabet tree alignment algorithms. As Wheeler observed, this approach can lead to considering inappropriate character states (such as a head state and a foot state) homologous. We show that it is also sensitive to the specific ordering assumption used and that such different character orders can lead to a preference for different trees. We present a simple four-taxon dataset with observations of absence, but no inapplicable characters or other kinds of character dependence, for which the dynamic homology framework gives different results to classic algorithms for independent characters, including an optimal tree with biologically impossible reconstructions at inner nodes (every terminal has a head but the inner nodes are headless). We show how these issues can be solved by removing the character ordering assumption that the approach requires. Doing so, the dynamic homology framework reduces in general to Maddison's (Syst. Biol. 1993, 42, 576) well-known proposal to deal with inapplicability using step matrix analysis of complex characters. If in addition costs are interpreted in terms of homology, it reduces to Goloboff et al.'s (Cladistics 2021, 37, 596) step matrix implementation for maximization of homology as applied to inapplicable characters. However, if used with homogeneous costs, as Wheeler suggested, it reduces to unordered analysis of such complex characters, which is known to treat tails that may share many observed features as irrelevant for establishing kinship when they differ in just one feature, e.g. colour.
惠勒（Cladistics 2023，39，475）最近提出，在系统发育分析中处理不可应用特征的问题，可以直接通过将特征的观察缺失不作为单独的缺失/存在特征来处理，而是将其视为描述特征所有变体的复杂特征中的插入/缺失事件；并且可以通过对字符集施加序列或线性顺序并使用自定义字母树对齐算法分析所得序列字符来实现这种动态同源性观点。正如惠勒所观察到的，这种方法可能导致认为不适当的字符状态（例如头部状态和足部状态）是同源的。我们表明，它也对所使用的特定排序假设敏感，并且这种不同的字符顺序可能导致对不同树的偏好。我们提出了一个简单的四分类数据集，其中观察到缺失，但没有不可应用的字符或其他类型的字符依赖性，对于这个数据集，动态同源性框架给出了与独立字符的经典算法不同的结果，包括一个具有生物学上不可能的内部节点重建的最优树（每个末端都有一个头部，但内部节点没有头部）。我们展示了如何通过取消该方法所需的字符排序假设来解决这些问题。这样做，动态同源性框架通常会简化为 Maddison 的（Syst. Biol. 1993，42，576）著名建议，即使用复杂字符的步矩阵分析来处理不可应用问题。如果另外将成本解释为同源性，它将简化为 Goloboff 等人的（Cladistics 2021，37，596）步矩阵实现，用于最大化应用于不可应用字符的同源性。然而，如果像惠勒建议的那样使用同质成本，它将简化为这种复杂字符的无序分析，众所周知，当尾巴在仅一个特征上不同但共享许多观察到的特征时，它会被视为与建立亲缘关系无关，例如颜色。
