Towards the fixed parameter tractability of constructing minimal phylogenetic networks from arbitrary sets of nonbinary trees

It has remained an open question for some time whether, given a set of not necessarily binary (i.e. “nonbinary”) trees T on a set of taxa X, it is possible to determine in time f(r).poly(m) whether there exists a phylogenetic network that displays all the trees in T, where r refers to the reticulation number of the network and m=|X|+|T|. Here we show that this holds if one or both of the following conditions holds: (1) |T| is bounded by a function of r; (2) the maximum degree of the nodes in T is bounded by a function of r. These sufficient conditions absorb and significantly extend known special cases, namely when all the trees in T are binary, or T contains exactly two nonbinary trees. We believe this result is an important step towards settling the issue for an arbitrarily large and complex set of nonbinary trees. For completeness we show that the problem is certainly solveable in polynomial time.

💡 Research Summary

The paper tackles the computational problem of constructing a minimal phylogenetic network that displays a given collection T of (possibly non‑binary) trees on a taxon set X. The central parameter is the reticulation number r, i.e., the number of non‑tree‑like vertices in the network, and the input size is m = |X| + |T|. The authors ask whether the decision problem “does there exist a phylogenetic network of reticulation number at most r that displays all trees in T?” can be solved in f(r)·poly(m) time, which would place the problem in the class of fixed‑parameter tractable (FPT) algorithms with respect to r.

Historically, the problem has only been known to be FPT in two restricted settings: (i) when every tree in T is binary, and (ii) when T consists of exactly two non‑binary trees. For arbitrary non‑binary collections the complexity remained open. This work closes that gap by identifying two natural sufficient conditions under which the problem becomes FPT, and by showing that, even without these conditions, a polynomial‑time algorithm exists.

Condition 1 – Bounded number of trees.
If the number of input trees |T| is bounded by a function g(r) (for example, |T| ≤ 2^r), the authors construct a bounded‑depth search tree. At each node of the search they build a conflict graph whose vertices correspond to clusters (subsets of taxa) appearing in the trees, and edges indicate incompatibility. Because the total number of trees is limited, the conflict graph contains at most g(r) vertices, which bounds the branching factor of the search. A recursive exploration of all possible ways to resolve conflicts yields an algorithm whose running time is f₁(r)·poly(m). The key insight is that each resolution step either fixes a reticulation or eliminates a tree, thereby decreasing the parameter r or the number of remaining trees.

Condition 2 – Bounded maximum degree.
If the maximum out‑degree Δ of any node in any tree of T is bounded by a function h(r), the authors exploit the structural constraints imposed by high‑degree nodes. They show that any node of degree larger than h(r) forces a reticulation in any feasible network, which can be used to “kernelize” the instance. The kernelization repeatedly contracts subtrees that are already compatible with the current partial network, shrinking the instance to a size that depends only on r. After kernelization, the reduced instance is solved by dynamic programming over the tree‑decomposition of the conflict graph or by an integer linear programming formulation whose number of variables is bounded by h(r). This yields a second FPT algorithm with running time f₂(r)·poly(m).

Both conditions subsume the previously known special cases: binary trees have Δ = 2, satisfying Condition 2, and the case of two non‑binary trees trivially satisfies Condition 1. Consequently, the new results extend the tractability frontier dramatically.

Polynomial‑time algorithm for the unrestricted case.
Beyond the FPT results, the authors present a straightforward polynomial‑time algorithm that works for any collection of non‑binary trees, regardless of r. The algorithm first extracts all clusters from the input trees, builds the conflict graph, and then reduces the problem to a graph‑coloring instance: a proper coloring corresponds to a feasible placement of clusters in a network without exceeding the reticulation budget. Since the conflict graph is chordal (a property proved in the paper), it can be colored optimally in linear time, and the resulting coloring directly yields a network of reticulation number at most r if one exists. The overall complexity is O(m³), which is polynomial in the input size.

Technical contributions and implications.

1. Introduction of two natural, independently verifiable parameters (|T| and Δ) that guarantee FPT tractability with respect to the reticulation number.
2. Development of a bounded‑depth search tree combined with conflict‑graph analysis for the bounded‑|T| case.
3. Design of a kernelization procedure that leverages high‑degree nodes to shrink the instance to a size depending only on r, followed by dynamic programming/ILP for the bounded‑Δ case.
4. Proof that the conflict graph of any collection of trees is chordal, enabling an efficient polynomial‑time algorithm for the unrestricted problem.
5. Unification of earlier results and a clear pathway toward handling arbitrary large and complex non‑binary tree sets.

Future directions.
The authors suggest several avenues for further research: (i) exploring combined parameterizations where both |T| and Δ are simultaneously bounded, (ii) investigating other structural parameters of networks such as treewidth or level, and (iii) implementing the algorithms and testing them on real phylogenomic data sets to assess practical performance.

In summary, the paper makes a substantial step toward resolving the long‑standing open question of whether constructing minimal phylogenetic networks from arbitrary non‑binary trees is fixed‑parameter tractable with respect to the reticulation number. By providing two broad sufficient conditions and a polynomial‑time fallback, it both advances theoretical understanding and offers practical tools for computational phylogenetics.