Hybridization Graphs (H-graphs)

• H-graphs are intersection graphs derived from connected subgraphs of a subdivided host graph H, unifying traditional classes like interval, circular-arc, and chordal graphs.
• They exhibit bounded structural parameters such as mim-width and Boolean-width, which enable efficient dynamic programming and meta-algorithmic applications in optimization problems.
• Research on H-graphs addresses a spectrum of issues from NP-complete recognition challenges to fixed-parameter tractability in clique, coloring, and induced subgraph problems.

A hybridization graph (H-graph) is a graph-theoretic abstraction defined as the intersection graph of connected subgraphs of a subdivision of a fixed host graph $H$. This model encompasses and generalizes multiple important graph classes—including interval graphs, circular-arc graphs, and chordal graphs—providing a unifying perspective for structural and algorithmic analysis. The theory of H-graphs, initiated by Biró, Hujter, and Tuza (1992), has catalyzed significant research on geometrical representations, combinatorial optimization, computational complexity, and structural graph parameters (Lima et al., 20 Oct 2025, Çağırıcı et al., 2022, Fomin et al., 2017, Chaplick et al., 2017).

1. Formal Definition and Canonical Examples

Given a fixed, finite, simple graph $H$, an H-graph $G$ is constructed as follows:

• Select a subdivision $H'$ of $H$ (replace each edge $e = ab$ of $H$ by a path $P_e$ of length at least one, with new internal vertices that are pairwise disjoint and non-adjacent between paths).
• Assign to each vertex $v \in V(G)$ a nonempty subset $\Phi(v) \subseteq V(H')$ such that $H$0 is connected.
• For any distinct $H$1, define $H$2.

The pair $H$3 is called an H-representation, and $H$4 is the "segment" of $H$5 in $H$6. The class of $H$7-graphs generalizes several well-studied classes:

$H$8	Induced Graph Class	Structural Representation
Path ($H$9)	Interval graphs	Intersection graphs of subpaths of a path
Cycle ($G$0)	Circular-arc graphs	Intersection graphs of arcs in a circle
Any tree	Chordal graphs	Intersection graphs of subtrees of a tree

This framework extends to more complex host graphs $G$1 to define novel classes with rich combinatorial properties (Lima et al., 20 Oct 2025).

2. Structural Parameters: Mim-width, Boolean-width, and Minimal Separators

H-graphs are characterized by several key structural parameters critical to algorithmic tractability:

• Mim-width: The maximum size of an induced matching across any bipartition of the vertex set in a branch decomposition of $G$2. For any fixed $G$3, every $G$4-vertex $G$5-graph admits a branch decomposition of mim-width $G$6, computable in polynomial time. This yields (Fomin et al., 2017):

$G$7

• Boolean-width: The logarithm of the number of neighborhood-equivalence classes across a cut. For fixed $G$8, boolean-width is bounded logarithmically:

$G$9

• Minimal Separators: The number of minimal separators in $H'$0 is polynomial for fixed $H'$1:

$H'$2

Bounded mim-width and boolean-width enable the application of meta-algorithmic frameworks for efficient dynamic programming, while the polynomial bound on minimal separators underpins algorithms for a variety of induced subgraph and separation problems (Fomin et al., 2017).

3. Computational Complexity and Algorithmic Meta-theorems

Algorithmic advances for H-graphs hinge on the interplay between their structural decomposition and logical definability:

• For fixed $H'$3, numerous classical problems—including Maximum Clique, Maximum Independent Set, and Minimum Dominating Set—are solvable in polynomial time on $H'$4-graphs. This is realized via LC-VSP solvers that leverage bounded mim-width or polynomial minimal separators, and extends to other CMSOL-definable problems (Fomin et al., 2017).
• For problems expressible in Counting Monadic Second-order Logic (CMSOL) and with bounded treewidth constraints, meta-theorems guarantee polynomial-time algorithms on $H'$5-graphs (Chaplick et al., 2017).

For problem-specific tractability:

• Clique and Coloring: When $H'$6 is a cactus (every edge in at most one cycle) or when the H-representation has the Helly property, MAX-CLIQUE is solvable in polynomial time (Chaplick et al., 2017).
• Induced-Path Problems: Longest Induced Path, Induced Disjoint Paths, and similar are solvable in $H'$7 time given a decomposition.
• Parameterized Algorithms: $H'$8-CLIQUE and List $H'$9-Coloring are fixed-parameter tractable (FPT parameterized by $H$0) for fixed $H$1 due to the treewidth bound $H$2 (Chaplick et al., 2017).

Notably, while general polynomial-time recognition algorithms exist for $H$3 corresponding to paths, cycles, and trees (i.e., for interval, circular-arc, and chordal graphs, respectively), the recognition problem for $H$4-graphs is NP-complete if $H$5 contains two distinct cycles, as established via reductions from bipartite 2-track graph problems (Çağırıcı et al., 2022).

4. Combinatorial Optimization: Hardness and Fixed-Parameter Results

Complexity analyses of H-graphs exhibit a nuanced tractability landscape:

• APX-Hardness: For any $H$6 containing as a minor the multigraph $H$7 of three nodes with two parallel edges between each pair, MAX-CLIQUE on $H$8-graphs is APX-hard; graph isomorphism is GI-complete under the same condition (Chaplick et al., 2017).
• W[1]-Hardness: Both Maximum Independent Set and Minimum Dominating Set are W[1]-hard when parameterized by $H$9, even with an $e = ab$0-representation given. Reductions are constructed from Multicolored Clique and Multicolored Independent Set (Fomin et al., 2017).
• FPT Cases: Dominating Set is fixed-parameter tractable (FPT) parameterized by $e = ab$1 when $e = ab$2 is a tree $e = ab$3 (equivalently, on chordal graphs of bounded leafage). A dynamic programming approach yields an algorithm of complexity $e = ab$4, where $e = ab$5 is the leafage of the chordal graph (Fomin et al., 2017).

For MAX-CLIQUE, a kernel with at most $e = ab$6 vertices is computable in $e = ab$7 time, maintaining all $e = ab$8-cliques of $e = ab$9 (Fomin et al., 2017).

5. Recognition Complexity: Tractability and Intractability Frontier

The complexity of deciding whether a graph $H$0 is an $H$1-graph is sensitive to the structure of $H$2:

• Polynomial-Time Recognition:
  • For $H$3 (interval graphs) and $H$4 (circular-arc graphs), classical polynomial-time algorithms exist (Çağırıcı et al., 2022).
  • For $H$5 (tree), recent results provide polynomial-time recognition for every fixed $H$6; the algorithm proceeds via clique-tree and subtree-model analysis (Çağırıcı et al., 2022).
  • For specific unicyclic $H$7 (e.g., lollipop graphs $H$8), polynomial-time algorithms exist by structural decomposition—partitioning components with respect to maximal cliques and applying models for interval and circular-arc graphs (Çağırıcı et al., 2022).
• NP-Completeness:
  • If $H$9 contains two different cycles, recognition of $P_e$0-graphs is NP-complete. The construction extends from core butterfly minors to more complex cyclic host graphs.
  • For medusa graphs (unions of $P_e$1-graphs as $P_e$2 ranges over unicyclic graphs), recognition is NP-complete, while the restriction to Helly models admits efficient algorithms (Çağırıcı et al., 2022).

Open problems remain, such as the recognition complexity for general unicyclic $P_e$3-graphs (other than cycles and lollipops) and the fixed-parameter tractability of recognition parameterized by $P_e$4.

6. Longest Path Transversals and Open Structural Questions

H-graphs admit a bounded longest path transversal number: For connected $P_e$5 with at least two vertices, any connected $P_e$6-graph $P_e$7 satisfies

$P_e$8

This upper bound follows from decomposing $P_e$9 via tree-decomposition, constructing a vertex set $v \in V(G)$0 intersecting all longest paths by Helly-type arguments, and refining $v \in V(G)$1 through combinatorial domination properties, yielding a small transversal $v \in V(G)$2 (Lima et al., 20 Oct 2025). In certain classical classes (interval, circular-arc, chordal with bounded forbidden minors), tighter bounds exist. Whether the dependency on $v \in V(G)$3 can be improved, or whether a universal constant suffices for all connected graphs, remains open.

7. Perspectives and Research Directions

The theory of H-graphs provides a unified combinatorial and algorithmic framework bridging classical structural graph theory and modern algorithmic meta-theorems. Research continues on several fronts:

• Improving the bounds for transversal numbers and separator cardinality.
• Clarifying the recognition complexity for various host classes, including further unicyclic, cactus, and "medusa" configurations.
• Investigating parameterized and approximation algorithms for optimization problems, especially in variants admitting Helly representations or other geometric restrictions.
• Extending the H-graph model to accommodate additional structural features, such as hierarchical and weighted relations relevant for network science and applications.
• Developing recognition and isomorphism algorithms for host graphs $v \in V(G)$4 beyond classical cases.

Ongoing work aims to fully delineate the landscape of tractable and intractable cases, with implications for both foundational graph theory and algorithm design in information systems and computational biology (Lima et al., 20 Oct 2025, Fomin et al., 2017, Chaplick et al., 2017, Çağırıcı et al., 2022).