Statistical Genomics Foundry
Paper
domain/gene-family-evolutiondomain/ancestral-reconstruction

Count: evolutionary analysis of phylogenetic profiles with parsimony and likelihood

1. Citation

Csűrös M. 2010. “Count: evolutionary analysis of phylogenetic profiles with parsimony and likelihood.” Bioinformatics 26(15):1910–1912. DOI 10.1093/bioinformatics/btq315. Publisher URL: https://academic.oup.com/bioinformatics/article/26/15/1910/189891 Copyright line on the article page: ”© The Author 2010. Published by Oxford University Press. All rights reserved.” Access date: 2026-07-03.

2. Access note

Read via the OUP article page (WebFetch). Retrieved the abstract plus the extractable body text (method descriptions, availability/implementation section). This is a two-page Applications Note; figures and any dense equation typesetting are not fully captured. Where the page text did not supply a detail (e.g. concrete input file formats), that is recorded as a confident silence in §10, not filled from memory. No facts imported from the Count manual/website — paper only.

3. Thesis (1 sentence)

Count is a Java software package that performs ancestral reconstruction and lineage/family-specific inference on integer-valued phylogenetic profiles (chiefly gene-family phyletic distributions), implementing both parsimony (Dollo/Wagner) and probabilistic birth-and-death methods.

4. Problem & context

Gene-content evolution is studied via phyletic (presence–absence) profiles of homologous gene families across taxa on a phylogeny. Count generalizes presence–absence to numerical profiles Φ: 𝒳 ↦ {0, 1, 2, …} (integer family sizes / copy numbers), and reconstructs how those characters changed along an evolutionary tree, inferring family- and lineage-specific characteristics. It targets databases of clustered homolog families such as COG.

5. Method / approach

Given an evolutionary tree T with terminal taxa 𝒳 and a numerical profile, Count computes ancestral states ξ[u] at internal nodes u ∈ T and reconstructs character history. Two method families:

  • Parsimony methods (functional strings verbatim):

    • Dollo parsimony (cited to Farris)
    • Wagner parsimony (cited to Farris)
    • Asymmetric Wagner parsimony — penalizes gains and losses with different weights.
    • Propensity for Gene Loss (PGL) — computed by Count, cited to Krylov et al. 2003; associated with Dollo parsimony in the abstract.

  • Probabilistic / likelihood method:

    • A phylogenetic birth-and-death model (the paper’s repeated phrasing), built on linear birth-death-immigration processes (cited to Kendall, 1949).
    • Per-edge continuous-time Markov character evolution; each edge uv has a length τ.
    • Three branch-specific rate parameters (functional strings verbatim): gene loss rate μ, gene duplication rate λ, and gain rate κ.
    • Count allows rate variation across branches and across gene families.
    • Rate optimization (fitting the model) and ancestral reconstruction are the two main computational components.

Interface / implementation (functional strings verbatim):

  • Written entirely in Java (Java SE 6); tested on Windows, MacOS X and Linux.
  • Rich GUI supporting multiple datasets/models, session saving, and tab-delimited export.
  • The main components (rate optimization and ancestral reconstruction) can also be launched from the command line.
  • Distributed as a stand-alone Java application, a MacOS X application bundle, and via Java Webstart; ships with test data and a User’s Guide. Distributed under a BSD-style license; source available on request.

6. Key claims / findings

  • Count handles integer-valued profiles Φ ∈ {0,1,2,…}, not just binary presence–absence.
  • It implements Dollo, Wagner, and asymmetric Wagner parsimony, plus PGL.
  • It implements a probabilistic phylogenetic birth-and-death model with per-branch loss/duplication/ gain rates (μ, λ, κ) and supports rate variation across branches and families.
  • Per-family output: number of branches where a family was lost, gained, expanded, and contracted — under the probabilistic model these are estimated as expectations, hence fractional values.
  • Per-lineage (aggregate) output: counts of families lost, gained, expanded, and contracted on each branch.
  • Ancestral reconstruction output: inferred probabilities for family presence and absence at ancestral nodes (posterior probabilities conditioned on observed profiles), displayed on the tree.
  • Availability: http://www.iro.umontreal.ca/~csuros/gene_content/count.html.

7. Load-bearing statements (RESTRICTIVE license → own-words paraphrase; functional strings verbatim)

Mode: article is “All rights reserved,” so the statements below are paraphrased in own words; only functional strings (method names, Java version, rate symbols, the “expectations/fractional values” mechanism) are reproduced verbatim.

  • Count is a software package for analyzing numerical profiles on a phylogeny, primarily phyletic distributions of homologous gene families, but usable for any integer-valued evolutionary character. It performs ancestral reconstruction and infers family- and lineage-specific characteristics along the tree. (paraphrase of abstract)
  • It implements methods common in gene-content analysis — Dollo and Wagner parsimony, propensity for gene loss — as well as probabilistic methods using a phylogenetic birth-and-death model. (paraphrase of abstract; named methods verbatim)
  • The probabilistic model rests on linear birth-death-immigration processes (Kendall, 1949); character evolution on each edge uv of length τ is stochastically determined. (paraphrase; model term + τ verbatim)
  • Aggregate family-specific branch counts (lost/gained/expanded/contracted) are “estimated as expectations, hence the fractional values” (verbatim functional phrase).
  • Ancestral display: the tree shows inferred probabilities for family presence and absence at ancestral nodes (paraphrase; the “presence and absence at ancestral nodes” phrasing is the substance).
  • Written in Java (Java SE 6); distributed under a BSD-style license (functional strings).

8. Stated scope, assumptions, limitations

  • Scope: integer-valued (“numerical”) profiles on a phylogeny; primary application is gene-family phyletic distributions, but explicitly generalizable to any integer-valued evolutionary character.
  • The probabilistic model assumes character evolution as a continuous-time Markov birth-death- immigration process along tree edges parameterized by edge length τ and rates μ/λ/κ.
  • Rate variation is allowed across branches and across gene families.
  • The paper does not state input-tree constraints (see §9/§10).

9. Failure modes / invalidity patterns

The paper is an applications note and does not enumerate failure modes, diagnostics, error messages, or invalidity conditions. In particular:

  • No stated requirement that the input tree be ultrametric, time-scaled, or clock-like. The only tree/edge statement found is that “character evolution on each edge uv with length τ is stochastically determined” — i.e. edges carry a length τ used as the parameter of the per-edge process, but the paper gives no ultrametricity or non-negativity constraint on those lengths.
  • No statement about negative branch lengths or how they would be handled.
  • No stated symptom, threshold, or diagnostic value. [summarizer note: absence is the finding here, not an inference.]

10. What the source does NOT address (confident silences)

  • Concrete input file formats — no Newick / tab-delimited / specific format for either the tree or the profile table is specified in the article text (tab-delimited is mentioned only as an export option from the GUI).
  • Tree/branch-length requirements — ultrametricity, time-scaling, molecular clock, and negative branch lengths are never discussed (words “ultrametric,” “time,” “molecular clock” do not appear; “branch length”/edge length τ does).
  • Parsimony-vs-likelihood bias — the paper makes no comparative claim that parsimony under- or over-estimates gains or losses relative to likelihood.
  • Exact CLI syntax / flag names (e.g. any --ancestral_counts-style option) — not given; the paper only says the main components “can also be launched from the command line.”
  • Output file names — not specified.
  • Scalability / dataset-size limits / runtime performance — not discussed.
  • Software version — the paper describes the 2010 release; it pins no version number in text.

11. Open questions / ambiguities

  • Whether branch lengths τ must be strictly positive / how the birth-death likelihood behaves for zero or negative edge lengths — unresolved by the paper.
  • Exact input/output file conventions and CLI invocation — deferred to the (separate) User’s Guide, which this note does not read.
  • Whether “expansion/contraction” counts apply to both parsimony and probabilistic modes or only the latter — the fractional-expectation phrasing implies the probabilistic mode, but the paper is not fully explicit.

12. Guidance answers

  • What Count does / methods & outputs: Ancestral reconstruction of integer-valued gene-content (phyletic) profiles on a phylogeny, via BOTH parsimony — Dollo, Wagner, asymmetric Wagner, plus Propensity for Gene Loss (PGL) — AND a probabilistic phylogenetic birth-and-death model (linear birth-death-immigration, Kendall 1949) with per-branch rates μ (loss), λ (duplication), κ (gain). Outputs: per-family branch counts of families lost/gained/expanded/contracted (probabilistic values are expectations → fractional); per-lineage aggregate counts of families lost/gained/expanded/contracted per branch; and ancestral posterior probabilities of family presence/absence at internal nodes. (§5, §6)
  • Tree requirement (ultrametric / branch lengths / negative-length failure): The paper does NOT state that the input tree must be ultrametric or time-scaled. It refers only to each edge having a length τ used as the per-edge birth-death process parameter. No mention of negative branch lengths. → The skill’s claim “Count requires ultrametric” is not supported by this paper; if the skill needs that constraint it must be sourced elsewhere (manual/website) or labeled convention, not cited to this paper. (§9)
  • Parsimony vs. likelihood under/over-estimating gains or losses: The paper makes no such comparative statement. → The reconciliation-table claim “parsimony underestimates losses” is not sourced from this paper; treat as the skill’s inference or a community convention, not a Csűrös-2010 fact. (§10)
  • Input format / CLI / output file names / Java: It is a Java application (Java SE 6), GUI + command-line, BSD-style license. The paper does not specify input file formats, CLI flag names, or output file names. → CLI details like --ancestral_counts / Count.jar cannot be corroborated from this paper. (§5, §10)
  • Version pin: Paper is the 2010 release and pins no version; the skill’s “Count 11.0319+” pin postdates this paper — version-specific behavior is not sourceable here. (§10)
  • Must-quote functional strings captured verbatim: Dollo parsimony, Wagner parsimony, asymmetric Wagner parsimony, Propensity for Gene Loss (PGL), phylogenetic birth-and-death model, linear birth-death-immigration, rates μ/λ/κ, edge length τ, Java SE 6, BSD-style license, “estimated as expectations, hence the fractional values.”