Three desirable properties for any method of selecting a subset of evolutionary units (EUs) for conservation or for genomic sequencing are discussed. These properties are spread, stability, and applicability. We are motivated by a practical case in which the maximization of phylogenetic diversity (PD), which has been suggested as a suitable method, appears to lead to counterintuitive collections of EUs and does not meet these three criteria. We define a simple greedy algorithm (GREEDYMMD) as a close approximation to choosing the subset that maximizes the minimum pairwise distance (MMD) between EUs. GREEDYMMD satisfies our three criteria and may be a useful alternative to PD in real-world situations. In particular, we show that this method of selection is suitable under a model of biodiversity in which features arise and/or disappear during evolution. We also show that if distances between EUs satisfy the ultrametric condition, then GREEDYMMD delivers an optimal subset of EUs that maximizes both the minimum pairwise distance and the PD. Finally, because GREEDYMMD works with distances and does not require a tree, it is readily applicable to many data sets.

中文翻译：

---

选择要保存或排序的分类群：理想的标准和贪婪的解决方案

讨论了为保护或基因组测序选择进化单元 (EU) 子集的任何方法的三个理想特性。这些特性是铺展性、稳定性和适用性。我们受到一个实际案例的启发，在该案例中，系统发育多样性 (PD) 的最大化被认为是一种合适的方法，但似乎导致了违反直觉的 EU 集合，并且不符合这三个标准。我们将简单的贪心算法 (GREEDYMMD) 定义为选择最大化 EU 之间最小成对距离 (MMD) 的子集的近似方法。GREEDYMMD 满足我们的三个标准，可能是现实世界中 PD 的有用替代方案。特别是，我们表明这种选择方法适用于生物多样性模型，其中特征在进化过程中出现和/或消失。我们还表明，如果 EU 之间的距离满足超度量条件，则 GREEDYMMD 会提供一个最优的 EU 子集，该子集最大化最小成对距离和 PD。最后，因为 GREEDYMMD 处理距离并且不需要树，所以它很容易适用于许多数据集。