Hypothesis Testing

Phylogenetics

Hypothesis Testing

  • Presenter: Joe Ironside

  • A critical aspect of phylogenetics involving evaluation of proposed evolutionary trees, assessing their validity and reliability.

Robustness of Phylogenies

  • Essential Question: Is our tree robust?

    • Inquiry into the reliability of the constructed phylogenetic tree.

    • Exploration of confidence levels regarding the tree's reflection of the true phylogeny.

Consensus Trees

  • A tool to summarize a whole set of phylogenetic trees. Differentiated into types:

    • Strict Consensus Tree:

    • Clades must be present in all trees.

    • Semi-Strict Consensus Tree:

    • Clades must be present in at least one tree and not contradicted by any tree.

    • Majority Rule Consensus Tree:

    • Clades must be present in a specified proportion of trees.

    • Typically requires presence in at least 50% of trees.

    • Clades not meeting these criteria are collapsed, generating polytomies.

Examples include:

  • Mammals

  • Turtles

  • Lizards

  • Crocodilians

  • Birds

Bootstrapping

  • A technique to assess the reliability of phylogenetic trees:

    • Description: Pseudoreplication is employed where a new dataset is created by randomly sampling characters from the original dataset.

    • Procedure:

    • Phylogeny is reconstructed for each pseudoreplicate.

    • This process is repeated many times.

    • Majority Rule Consensus Tree is then constructed using optimal trees derived from all pseudoreplicates.

    • Bootstrap Support for a Branch:

    • Defined as the percentage of optimal trees in which the branch appears.

Results from bootstrapping may show supportive values:

  • Example: Leads to bootstrap values (e.g., Mammals: 83, Turtles: 34, Lizards: 98, Crocodilians: 47).

Monophyly

  • Monophyletic Group:

    • A group that includes all descendants of a common ancestor, thus depicting a complete lineage.

    • Hierarchical classification includes:

    • Pisces

    • Amphibia

    • Mammalia

    • Testudines

    • Lepidosauria

    • Crocodylia

    • Archosauria

    • Diapsida

    • Reptilia

    • Amniota

    • Tetrapoda

    • Vertebrata

    • Aves

Paraphyly

  • Paraphyletic Group:

    • A group that excludes one or more descendants of a common ancestor, which are placed in a separate category. Examples include:

    • Reptilia:

      • This group excludes birds.

    • Artiodactyla (Even-toed Ungulates):

      • This group excludes cetaceans (whales, dolphins).

    • Invertebrates:

      • This group excludes vertebrates.

    • Prokaryotes:

      • This group excludes eukaryotes.

Polyphyly

  • Polyphyletic Group:

    • A group that excludes the common ancestor, and some of its descendants may also be excluded.

    • Examples:

    • Warm-blooded vertebrates exclude the common ancestor of birds and mammals, as well as all reptilian descendants of this common ancestor.

Comparison of Group Types

  • Monophyly: All descendants from a common ancestor included.

  • Paraphyly: Excludes one or more descendants from a common ancestor, thus not fully inclusive.

  • Polyphyly: Excludes the common ancestor itself, leading to a more unrelated grouping.

Additional Hypothesis Testing

  • Fundamental Question: Is one tree significantly better than another?

Comparing Two Trees

  • Involves specifying two or more trees a priori without utilizing data for a tree search.

  • Utilizes the Kishino-Hasegawa (KH) test for this comparison and evaluation.

Using Constrained Trees

  • Allows for hypothesis testing even without specifying trees a priori:

    • Procedure:

    • A constrained tree is generated by specifying constraints on the analysis.

    • A tree search is performed utilizing parsimony, likelihood, or Bayesian criteria.

    • Only trees meeting the specified constraints are retained.

    • Example:

    • Constrained Tree: Birds and crocodilians form a monophyletic group.

    • Unconstrained Tree: Birds and crocodilians may not form a monophyletic group.

Hypothesis Testing Methods

Parsimony Testing

  1. With Constraint Enforced: Produces the most parsimonious tree(s) that meet the constraints.

  2. Without Constraint: Produces the most parsimonious tree(s) irrespective of constraints.

  3. Tests: Check for significant differences in parsimony using:

    • Templeton Test

    • Winning Sites Test

Likelihood Testing

  1. With Constraint Enforced: Generates likeliest tree(s) that fulfill constraints.

  2. Without Constraint: Generates likeliest tree(s) independent of constraints.

  3. Tests: Assess significant differences in likelihood using:

    • Topology-based Tests

    • Shimodaira-Hasegawa (SH) Test

    • Approximately Unbiased (AU) Test

Bayesian Testing

  • Involves Markov Chain Monte Carlo (MCMC) Simulation:

    • Yields a set of best trees at a state of stationarity.

    • Within this set, identifies the number of trees that are consistent with the constraint tree.

    • Computes posterior probability of the hypothesis by dividing consistent tree counts by total tree counts.

Phylogenetic Comparative Methods

  • Multispecies Comparative Studies:

    • Design to test hypotheses by comparing traits across different species.

    • Example: Investigating if polygynous bird species tend to exhibit bright male coloration.

Addressing Phylogenetic Confounding

  • Acknowledge the confounding effects of phylogeny, as species may not be independent data points due to similarities resulting from recent common ancestry.

  • Experimental Design:

    • Should mitigate this confounding effect, specifically avoiding phylogenetic pseudoreplication (e.g., comparing polygynous pheasants with monogamous geese).

Phylogenetically Independent Contrasts (PICs)

  • A methodological approach designed to account for phylogenetic relationships, thus correcting for the influence of phylogeny in statistical analyses.