Perfect Phylogeny¶
A tree that jointly displays character states without repeated introductions under a stated model.
Core Idea¶
A perfect phylogeny is one tree that jointly displays a set of discrete characters. For each character, the objects assigned to a particular state can occupy one connected region of that tree, separate from the regions for its other states. The same tree must work for all the characters under consideration. A different tree for each column does not make the complete character matrix compatible.[1]
“Perfect” describes exact fit to a declared model. It does not mean that the tree is a directly observed history or that recurrence, loss, error and uncertainty never occur in nature. General multistate compatibility can be stated on an unrooted tree; a rooted binary infinite-sites tree used to infer tumor clones is a narrower model with additional assumptions.[1][2]
Structural Signature¶
Sig role-phrases: labeled objects and character states → one candidate tree → connected state regions → joint compatibility decision → declared model scope.
- Objects and character-state matrix. Each object has a state for each relevant character. Without characters to display, a branching diagram is a tree but not a perfect phylogeny for a matrix.
- One candidate tree. A common topology must account for every included character. Substituting a separate tree for each troublesome column evades the joint condition.
- Connected state regions. For each character, occurrences of one state must be displayable as one connected part of the tree, distinct from that character's other states. If the same state requires separated regions on a proposed tree, that tree does not provide the stipulated exact display.[1]
- Compatibility witness or obstruction. A fitting tree witnesses compatibility; showing that no one tree can display the entire set establishes an obstruction. A fitting subset does not settle the larger set.
- Model assumptions. State whether characters are general multistate partitions or, for example, binary somatic-mutation patterns under rooted infinite-sites conditions. These assumptions determine what the display can mean.[2]
The roles specify a formal relationship, not a required biological sampling technique or software implementation.
What It Is Not¶
- Not any tree-shaped diagram. A topology without specified character states does not pass the display test.
- Not one tree per character. The compatibility question asks for a common witness.
- Not necessarily rooted. The general multistate construction in van Iersel, Jones and Kelk is unrooted. The tumor model below chooses a root for its own inference.[1][2]
- Not a universal no-loss historical verdict. An inferred compatible tree is a model fit; it cannot prove that the actual history lacked recurrence or reversal.
- Not a universal binary shortcut. The source notes that pairwise compatibility suffices in its two-state setting, while multistate local checks require care. Its eight-label example shows specifically that all three-character subsets may fit even when the four-character set does not.[1]
- Not automatically a tumor clone matrix. The tumor paper explicitly imposes an additional n-clonal condition beyond generic perfect-phylogeny compatibility.[2]
Scope of Application¶
The abstraction is used where discrete character-state observations are tested against a common branching explanation. In systematics, labels and traits may be modeled as partitions on an unrooted tree. In tumor evolution, a rooted clonal tree can encode binary mutation patterns under a monoclonal founder and infinite-sites assumptions. These are related compatibility problems, not interchangeable evidence about what actually happened.[1][2]
An object need not be a living species. What matters is the declared character matrix, the tree's interpretation and the rule for displaying states. The formal eight-label example below is synthetic. The tumor example applies an inferential model to sequencing measurements, which themselves are mixtures of clones.[2]
Clarity¶
Name the object set, character alphabet and data matrix before saying a phylogeny is perfect. Specify whether the claim is that a particular tree displays the matrix, that some such tree exists, or that a tested matrix has no perfect tree. Those are different assertions.
State the model's root and event conventions where they matter. In the cited tumor model, founder mutation r is present at the root and labels no acquisition edge; each non-founder mutation labels one acquisition edge. Treating r as one more edge event would misstate the paper's construction.[2]
Manages Complexity¶
One tree compresses many character constraints into a single proposed structure. Instead of offering a separate explanation of each column, the model exposes where a column or group of columns cannot coexist on the same tree.
That compression is conditional. A subset can have a simple display while an added character makes the full matrix incompatible. The model makes that failure visible rather than silently switching witnesses between subsets.[1]
Abstract Reasoning¶
Let C be a set of characters and T a tree. The question is whether T displays every character in C. If it does, every subset of C is displayed by T. The converse fails: the source constructs four characters on eight formal labels for which every three-character subset has a tree, yet the full four-character set has none. Thus local success is not, by itself, a global witness.[1]
Changing the character alphabet changes which compatibility results can be transferred. A binary pairwise condition cited in the source cannot be promoted to an arbitrary multistate decision rule without its own theorem. The level of formal scope is part of the reasoning, not a footnote.
Knowledge Transfer¶
Transfer the question—can one tree jointly explain all specified character states?—from formal taxon partitions to tumor mutation patterns. Preserve the differences: the former example is an unrooted multistate construction; the latter is rooted, binary and conditioned on an infinite-sites clone model with mixed-sample measurements.[1][2]
The transferable tool is a role map, not an automatic historical conclusion. A compatible tumor matrix permits a constrained inferred tree. It does not independently verify each acquisition or exclude unmodeled mechanisms such as copy-number changes or rearrangements.
Examples¶
Canonical formal case: an eight-label three-character subset¶
Van Iersel, Jones and Kelk define four partitions on formal labels a1–a4 and b1–b4. Their Figure 2 caption identifies a tree T1 displaying the exact three-character subset χ2, φ3 and ΩB. This is a positive formal compatibility case. The authors then show that the complete four-character set is incompatible even though each triple has a displayed tree. The labels are constructed taxa, not measured species; the detailed image topology is not independently reproduced here.[1]
Mapped back: Objects and states = eight formal labels and the selected three partitions; one candidate tree = T1; connected state regions = the display asserted for each selected partition; witness or obstruction = positive triple versus incompatible full four-character set; model scope = general unrooted multistate compatibility.
Applied inference case: tumor clonal trees¶
El-Kebir and colleagues analyze sequencing data from multiple tumor samples. A sample mixes cells from different clones, so mutation frequencies require inference of clone composition. Under their monoclonal, binary infinite-sites model, a rooted tree records one acquisition edge for each non-founder mutation, while founder mutation r is already in the root. The resulting clone-by-mutation patterns have nested or disjoint mutation-bearing sets. A compatible n-clonal matrix and tree are a model-constrained inference solution, not a directly witnessed cellular history.[2]
Mapped back: Objects and states = inferred clones and binary somatic-mutation patterns derived from mixed samples; one candidate tree = a rooted clonal tree; connected state regions = descendant sets for non-founder acquisitions under the model; witness or obstruction = the compatible inferred matrix/tree satisfying the extra n-clonal condition; model scope = monoclonal founder, infinite sites and no modeled copy-number/rearrangement events.
Structural Tensions¶
Exact compatibility versus recurrence-tolerant explanation. Exact display makes the constraint and its witness clear. Real or noisy data may call for recurrent changes, reversals or measurement-error models that permit more histories and more ambiguity. Accepting a recurrence-tolerant account sacrifices the clean exact-compatibility verdict; insisting on an exact tree can leave plausible observations unexplained. The cited works describe exact perfect phylogeny and its relation to near-perfect or model-constrained inference; no universal biological preference follows.[1][2]
Diagnostic: Which departures from exact state compatibility are plausible for this dataset, and how would they be distinguished from measurement error?
Local checking versus whole-matrix correctness. A valid binary local theorem can save work, but an unjustified multistate shortcut can accept a collection whose members do not fit one tree together. Checking the entire condition may cost more analysis; using the shortcut outside its scope costs correctness.[1]
Diagnostic: What alphabet and theorem license the proposed local test for this matrix?
Structural–Framed Character¶
Perfect phylogeny has a strongly structural formal identity within a framed biological modeling practice. The connected-region and one-tree conditions are formal relations; choosing character states, a root, an error model and the inferential question depends on the application.
- Evaluative weight. “Perfect” evaluates exact model fit, not historical certainty or biological merit. A poorer fit may be more honest when the assumed event model is false.
- Human-practice dependence. Researchers choose observations, coding and scope. Once specified, the mathematical compatibility test is not a matter of institutional preference.
- Institutional origin. Neither a paper's synthetic labels nor a tumor dataset fixes all possible instances; the compatibility rule can be reused on new matrices.
- Vocabulary travel. Object, character, state, tree and display map literally across the formal partition and clone-pattern cases, while their biological assumptions differ.
- Import versus recognition. Drawing an arbitrary tree imports a shape. Recognizing a perfect phylogeny requires verifying that it jointly displays the actual declared states.
Its character: an exact joint character/tree compatibility relation inside phylogenetic modeling, structurally testable once the data and assumptions are declared, and limited as evidence about observed history.
Structural Core vs. Domain Accent¶
The possible portable skeleton is several constraints sharing one satisfying structure. In this entry, that general idea does not by itself supply labeled biological objects, discrete inherited characters, tree display or connected state regions.
Those features are the domain accent and the selected identity. Replacing the formal taxon construction with a rooted clone model preserves the joint character/tree compatibility question, but replacing characters and tree with arbitrary constraints does not preserve a perfect phylogeny. The wider idea of shared constraint satisfaction may occur elsewhere, but its reach does not enlarge this named phylogenetic relation.
Instantiates / Related Primes¶
This entry is a kind of Tree (Graph Theory).
Tree (Graph Theory) is the broader mathematical kind: every perfect-phylogeny witness is a connected acyclic graph, while joint connected-state display of the declared characters distinguishes it from other trees. The unrooted formal witness and the rooted clone model both retain this tree carrier; its unique paths, bridges and minimum edge count do not by themselves make character inference easy or establish the actual evolutionary history.
Tree Structure describes branching organization; a tree alone does not meet this entry's character-display condition. Compatibility concerns co-satisfaction of constraints more broadly; perfect phylogeny adds the specific character-state/tree interpretation. These concepts help explain the model, but neither word alone supplies its complete definition.
Phylogenesis concerns evolutionary lineage formation; a compatible tree is a formal representation that may be used to infer a lineage, not the process itself. Phylogenetic Bracketing predicts unknown traits from known relatives on an assumed phylogenetic frame. It does not ask whether one tree exactly displays all observed characters.
Relationships to Other Abstractions¶
Current abstraction Perfect Phylogeny Domain-specific
Parents (1) — more general patterns this builds on
-
Perfect Phylogeny is a kind of Tree (Graph Theory) Domain-specific
A perfect phylogeny is a graph-theoretic tree specialized by joint connected-state display of a declared character matrix.The selected identity is the tree witness T for declared objects and characters, not the matrix alone or the decision procedure asking whether T exists. Every admitted witness is one finite connected acyclic graph; unrooted multistate T1 supplies it directly and the rooted clonal model adds a root and edge labels to the same tree carrier. Therefore unique paths, N−1 edges, bridges, the fundamental-cycle property, edge-minimal connectivity, leaf elimination and low treewidth apply as mathematical consequences, without claiming either source executes MST/treewidth algorithms or that perfect-phylogeny inference is easy. The stable differentia is that this same tree displays every included character with its state regions connected and disjoint under the stated model; not every graph-theoretic tree displays a given character set. A four-leaf star with character ab|cd is a tree but fails that display because the two state-spanning paths share the center. The nearer Tree genus already supplies Network, so no redundant direct Network edge is proposed. This strict edge may be emitted only after the separately approved exact Network-to-Reservoir-Flux deletion is installed and its current-catalog graph is independently cleared; the deletion is a release dependency, not an exception to the kind-of identity.
Hierarchy path (1) — routes to 1 parentless root
- Perfect Phylogeny → Tree (Graph Theory) → Network
Neighborhood in Abstraction Space¶
Perfect Phylogeny sits in a moderately populated region (59th percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.
Family — Unclustered & Miscellaneous (2721 abstractions)
Nearest neighbors
- Maximum Parsimony — 0.89
- Nested Set Model — 0.85
- Self-Verifying Finite Automaton — 0.84
- Gray Code — 0.84
- Generalized blockmodeling of binary networks — 0.84
Computed from structural-signature embeddings · 2026-10-17
Not to Be Confused With¶
- A generic genealogy or branching diagram: no character-state display condition is supplied.
- A tree fitted separately to each character: no joint witness is supplied.
- An observed history free of recurrence: compatibility under assumptions does not establish historical absence of repeated change.
- A rooted binary tumor tree as the whole category: it is a specialization with additional assumptions and n-clonal constraints.
- Phylogenetic bracketing: predicting a missing trait on a tree is a different operation from deciding matrix compatibility.
References¶
[1] Leo van Iersel, Mark Jones and Steven Kelk, “A Third Strike Against Perfect Phylogeny,” Systematic Biology 68(5):814–827 (2019), Introduction, “Example for Eight Taxa,” Figure 2 caption and incompatible-four discussion, and “Mathematical Definitions,” https://academic.oup.com/sysbio/article/68/5/814/5319938. Original publisher full text was inspected; Figure 2's detailed image pixels were not independently analyzed. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l
[2] Mohammed El-Kebir, Layla Oesper, Hannah Acheson-Field and Benjamin J. Raphael, “Reconstruction of clonal trees and tumor composition from multi-sample sequencing data,” Bioinformatics 31(12):i62–i70 (2015), Abstract, §2.1 Definitions 1–3 and perfect-phylogeny matrix discussion, and §3 Results, https://academic.oup.com/bioinformatics/article/31/12/i62/216528. Original publisher full text was inspected; the inferred tree remains conditional on the paper's model and mixed-sample measurements. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j