Table of Contents Author Guidelines Submit a Manuscript
The Scientific World Journal
Volume 2018, Article ID 1916094, 7 pages
https://doi.org/10.1155/2018/1916094
Research Article

The Probabilities of Trees and Cladograms under Ford’s -Model

Balearic Islands Health Research Institute (IdISBa) and Department of Mathematics and Computer Science, University of the Balearic Islands, 07122 Palma, Spain

Correspondence should be addressed to Francesc Rosselló; se.biu@ollessor.csec

Received 1 February 2018; Accepted 8 March 2018; Published 18 April 2018

Academic Editor: Béla Tóthmérész

Copyright © 2018 Tomás M. Coronado et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Ford’s -model is one of the most popular random parametric models of bifurcating phylogenetic tree growth, having as specific instances both the uniform and the Yule models. Its general properties have been used to study the behavior of phylogenetic tree shape indices under the probability distribution it defines. But the explicit formulas provided by Ford for the probabilities of unlabeled trees and phylogenetic trees fail in some cases. In this paper we give correct explicit formulas for these probabilities.

1. Introduction

The study of random growth models of rooted phylogenetic trees and the statistical properties of the shapes of the phylogenetic trees they produce was initiated almost one century ago by Yule [1] and it has gained momentum in the last 20 years: see, for instance, [28]. The final goal of this line of research is to understand the relationship between the forces that drive evolution and the topological properties of “real-life” phylogenetic trees [3, 9]; see also [10, Chapter 33]. One of the most popular such models is Ford’s -model for rooted bifurcating phylogenetic trees or cladograms [4]. It consists of a parametric model that generalizes both the uniform model (where new leaves are added equiprobably to any arc, giving rise to the uniform probability distribution on the sets of cladograms with a fixed set of taxa) and Yule’s model (where new leaves are added equiprobably only to pendant arcs, i.e., to arcs ending in leaves) by allocating a possibly different probability (that depends on a parameter and hence its name, “-model”) to the addition of the new leaves to pendant arcs or to internal arcs.

When models like Ford’s model are used to contrast topological properties of phylogenetic trees contained in databases like TreeBase (https://treebase.org), only their general properties (moments, asymptotic behavior) are employed. But, in the course of a research where we have needed to compute the probabilities of several specific cladograms under this model [11], we have noticed that the explicit formulas that Ford gives in [4, ] for the probabilities of cladograms and of tree shapes (unlabeled rooted bifurcating trees) are wrong, failing for some trees with leaves; see Propositions and in [4], with the definition of given in page 30 therein, for Ford’s formulas.

So, to help the future user of Ford’s model, in this paper we give the correct explicit formulas for these probabilities. This paper is accompanied by the GitHub page https://github.com/biocom-uib/prob-alpha where the interested reader can find a SageMath [12] module to compute these probabilities and their explicit values on the sets of cladograms with leaves labeled , for every from 2 to 8.

2. Preliminaries

2.1. Definitions, Notations, and Conventions

Throughout this paper, by a tree , we mean a rooted bifurcating tree. As it is customary, we understand as a directed graph, with its arcs pointing away from the root, which we shall denote by . Then, all nodes in have out-degree either 0 (its leaves, which form the set ) or 2 (its internal nodes, which form the set ). The children of an internal node are those nodes such that is an arc in , and they form the set . A node is a descendant of a node when there exists a directed path from to in . For every node , the subtree of rooted at is the subgraph of induced on the set of descendants of .

A tree is ordered when it is endowed with an ordering on every set . A cladogram (resp., an ordered cladogram) on a set of taxa is a tree (resp., an ordered tree) with its leaves bijectively labeled in . Whenever we want to stress the fact that a tree is not a cladogram, that is, it is an unlabeled tree, we shall use the term tree shape.

It is important to point out that although ordered trees have no practical interest from the phylogenetic point of view, because the orderings on the sets of children of internal nodes do not carry any phylogenetic information, they are useful from the mathematical point of view, because the existence of the orderings allows one to easily prove certain extra properties that can later be translated to the unordered setting (cf. Proposition 1).

An isomorphism of ordered trees is an isomorphism of rooted trees that moreover preserves these orderings. An isomorphism of cladograms (resp., of ordered cladograms) is an isomorphism of trees (resp., of ordered trees) that preserves the leaves’ labels. We shall always identify a tree shape, an ordered tree shape, a cladogram, or an ordered cladogram, with its isomorphism class, and in particular we shall make henceforth the abuse of language of saying that two of these objects, , are the same, in symbols , when they are (only) isomorphic. We shall denote by and , respectively, the sets of tree shapes and of ordered tree shapes with leaves. Given any finite set of taxa , we shall denote by and , respectively, the sets of cladograms and of ordered cladograms on . When the specific set is unrelevant and only its cardinal matters, we shall write and (with ) instead of and , and then we shall understand that is .

There exist natural isomorphism-preserving forgetful mappings

that “forget” the orderings or the labels of the trees. In particular, we shall call the image of a cladogram under its shape. Figure 1 depicts an example of images under these forgetful mappings.

Figure 1: An example of images under the forgetful mappings between (ordered and unordered) cladograms and tree shapes. In the ordered objects, the ordering is represented by the nodes’ colors: gray white.

Let us introduce some more notations. For every node in a tree , is its number of descendant leaves. For every internal node in an ordered tree , with children , its numerical split is the ordered pair . If, instead, is unordered and if with , then . In both cases, the multiset of numerical splits of is . For instance, if is the cladogram depicted in Figure 2, then

Figure 2: A cladogram in . The black nodes are its symmetric branch points.

A symmetric branch point in a tree is an internal node such that if and are its children, then the subtrees and of rooted at them have the same shape. For instance, the symmetric branch points in the cladogram depicted in Figure 2 are those filled in black.

Given two cladograms and on and , respectively, with , their root join is the cladogram on obtained by connecting the roots of and to a (new) common root ; see Figure 3. If are ordered cladograms, is ordered by inheriting the orderings on and and ordering the children of the new root as . If and are tree shapes, a similar construction yields a tree shape ; if they are moreover ordered, then becomes an ordered tree shape as explained above.

Figure 3: The root join .
2.2. The -Model

Ford’s -model [4] defines, for every , a family of probability density functions on that depends on one parameter , and then it translates this family into three other families of probability density functions on , on , and on , by imposing that the probability of a tree shape is equally distributed among its preimages under , , and , respectively.

It is well known [13] that every can be obtained in a unique way by adding recurrently to a single node labeled 1 new leaves labeled to arcs (i.e., splitting an arc into two arcs and and then adding a new arc from the inserted node to a new leaf) or to a new root (i.e., adding a new root and new arcs from to the old root and to a new leaf). The value of for is determined through all possible ways of constructing cladograms with shape in this way. More specifically,(1)if and denote, respectively, the only cladograms in and , let ;(2)for every , let be obtained by adding a new leaf labeled to . Then (3)When the desired number of leaves is reached, the probability of every tree shape is defined as

For instance, Figure 4 shows the construction of two cladograms in with the same shape and how their probability is built using the recursion in Step (2). If we generate all cladograms in with this shape, we compute their probabilities , and then we add up all these probabilities, we obtain the probability of this shape, which turns out to be ; cf. [4, Figure ].

Figure 4: Two examples of computations of the probability of a cladogram through its construction in Step (2) of the definition of the -model.

Once is defined on , it is transported to , , and by defining the probability of an object in one of these sets as the probability of its image in divided by the number of preimages of this image:(i)For every , if and it has symmetric branch points, thenbecause (see, e.g., [4, Lemma 31]).(ii)For every , if , thenbecause ( has different preimages under , obtained by taking all possible different combinations of orderings on the sets , ).(iii)For every , if and it has symmetric branch points, thenbecause (from the possible preimages of under , defined by all possible different combinations of orderings on the sets , , those differing only on the orderings on the children of the symmetric branch points are actually the same ordered tree shape).

The family of probabilities of ordered tree shapes satisfies the useful Markov branching recurrence (in the sense of [2, ]) given by the following proposition. In it and in the sequel, let, for every , where and is the mapping defined by and, for every , .

Proposition 1. For every and for every and ,

This recurrence, together with the fact that of a single node is , implies that, for every ,For proofs of Proposition 1 and (10), see Lemma 27 and Proposition 28 in [4], respectively.

3. Main Results

Our first result is an explicit formula for , for every and .

Proposition 2. For every , its probability under the -model is

Proof. Given , let be any ordered cladogram such that , and let and . If has symmetric branch points, then, by (4), (6), and (10), Now, on the one hand, it is easy to check that and therefore, since is symmetric, It remains to simplify this product. If, for every , we denote its children by and , then For every , the term appears twice in this product: in the denominator of the factor corresponding to itself and in the numerator of the factor corresponding to its parent. Therefore, all terms in this product vanish except (that appears in the denominator of its factor) and every with , a leaf. Thus, as we claimed.

Remark 3. Ford states (see [4, Proposition  32 and page 30]) that if , then where is the number of symmetric branching points in and If we simplify as in the proof of Proposition 2, this formula for becomeswhere is the number of internal nodes whose children have different numbers of descendant leaves. This formula does not agree with the one given in Proposition 2 above, because and, hence, it may happen that . The first example of a cladogram with this property (and the only one, up to relabeling, with at most 8 leaves) is the cladogram depicted in Figure 5. For it, our formula gives (see in the document ProblsAlpha.pdf in https://github.com/biocom-uib/prob-alpha)while expression (19) assigns to a probability of half this value:This last value cannot be right, for several reasons. Firstly, by [4, ], when , Ford’s model is equivalent to the uniform model, where every cladogram in has the same probabilityand when , Ford’s model gives rise to the Yule model [1, 14], where the probability of every is In particular, should be equal to and should be equal to . Both values are consistent with our formula, while expression (22) yields half these values.
As a second reason, which can be checked using a symbolic computation program, let us mention that if we take expression (22) as the probability of and hence of all other cladograms with its shape, and we assign to all other cladograms in the probabilities computed with Proposition 2, which agree on them with the values given by (19) (they are also provided in the aforementioned document ProblsAlpha.pdf), these probabilities do not add up 1.

Figure 5: The cladogram used in Remark 3.

Combining Proposition 2 and (4) we obtain the following result.

Corollary 4. For every with symmetric branch points,

This formula does not agree, either, with the one given in [4, Proposition 29]: the difference lies again in the same factor of 2 to the power of the number of internal nodes that are not symmetric branch points but whose children have the same number of descendant leaves.

The family of density mappings satisfies the following Markov branching recurrence.

Corollary 5. For every and for every and ,

Proof. If and , then as we claimed.

Remark 6. Against what is stated in [4], does not satisfy any Markov branching recurrence; that is, there does not exist any symmetric mapping such that, for every and for every and , Indeed, let be any two different tree shapes, both with leaves and symmetric branch points, for instance, the tree shapes in depicted in Figure 6. Then,In this case, has symmetric branch points and therefore while has symmetric branch points and therefore and . This shows that there does not exist any well-defined, single real number such that for every .

Figure 6: The tree shapes in mentioned in Remark 6.

Data Availability

The data used to support the findings of this study are available at the Github page that accompanies this paper.

Conflicts of Interest

The authors declare that there are no conflicts of interest regarding the publication of this paper.

Acknowledgments

This research was supported by Spanish Ministry of Economy and Competitiveness and European Regional Development Fund Project DPI2015-67082-P (MINECO/FEDER). The authors thank G. Cardona and G. Riera for several comments on the SageMath module that accompanies this paper.

References

  1. G. U. Yule, “A Mathematical Theory of Evolution, Based on the Conclusions of Dr. J. C. Willis, F.R.S.,” Philosophical Transactions of the Royal Society B: Biological Sciences, vol. 213, no. 402-410, pp. 21–87, 1925. View at Publisher · View at Google Scholar
  2. D. Aldous, “Probability distributions on cladograms,” in Random discrete structures, D. Aldous and R. Pemantle, Eds., vol. 76, pp. 1–18, Springer, New York, NY, USA, 1996. View at Publisher · View at Google Scholar · View at MathSciNet
  3. M. G. B. Blum and O. François, “Which random processes describe the tree of life? A large-scale study of phylogenetic tree imbalance,” Systematic Biology, vol. 55, no. 4, pp. 685–691, 2006. View at Publisher · View at Google Scholar · View at Scopus
  4. D. J. Ford, “Probabilities on cladograms: Introduction to the alpha model,” PhD Thesis (Stanford University), https://arxiv.org/abs/math/0511246.
  5. S. Keller-Schmidt, M. Tuğrul, V. M. Eguíluz, E. Hernández-García, and K. Klemm, “Anomalous scaling in an age-dependent branching model,” Physical Review E: Statistical, Nonlinear, and Soft Matter Physics, vol. 91, no. 2, Article ID 022803, 2015. View at Publisher · View at Google Scholar
  6. M. Kirkpatrick and M. Slatkin, “Searching for Evolutionary Patterns in the Shape of a Phylogenetic Tree,” Evolution, vol. 47, no. 4, pp. 1171–1181, 1993. View at Publisher · View at Google Scholar
  7. L. Popovic and M. Rivas, “Topology and inference for Yule trees with multiple states,” Journal of Mathematical Biology, vol. 73, no. 5, pp. 1251–1291, 2016. View at Publisher · View at Google Scholar · View at MathSciNet · View at Scopus
  8. R. Sainudiin and A. Véber, “A Beta-splitting model for evolutionary trees,” Royal Society Open Science, vol. 3, no. 5, Article ID 160016, 2016. View at Publisher · View at Google Scholar · View at Scopus
  9. A. O. Mooers and S. B. Heard, “Inferring evolutionary process from phylogenetic tree shape,” The Quarterly Review of Biology, vol. 72, no. 1, pp. 31–54, 1997. View at Publisher · View at Google Scholar · View at Scopus
  10. J. Felsenstein, Inferring Phylogenies, Sinauer Associates Inc., 2004.
  11. T. M. Coronado, A. Mir, F. Rosselló, and G. Valiente, “A balance index for phylogenetic trees based on quartets,” https://arxiv.org/abs/1801.05411.
  12. SageMath, the Sage Mathematics Software System (Version 7.6), The Sage Developers (2017) http://www.sagemath.org.
  13. L. L. Cavalli-Sforza and A. W. Edwards, “Phylogenetic Analysis: Models and Estimation Procedures,” Evolution, vol. 21, no. 3, pp. 550–570, 1967. View at Publisher · View at Google Scholar
  14. E. F. Harding, “The probabilities of rooted tree-shapes generated by random bifurcation,” Advances in Applied Probability, vol. 3, pp. 44–77, 1971. View at Publisher · View at Google Scholar · View at MathSciNet