Return to Main Index page

Required reading for the various lectures on Phylogenetics:

Avise text pp. 132-158 (photocopy in reading pile in BioSci 311)
  AND
Baldauf, SL. 2003. Phylogeny for the faint of heart. Trends in Genetics 19: 345-351. (pdf on WyoWeb course site)

    Download Excel spreadsheet HominTrees.XL demo. of 5-OTU UPGMA, Fitch-Margoliash and neighbor-joining tree routines from great ape distance matrix.

 Download Word doc of steps for UPGMA & NJ

  The Word (verbal) and Excel (quantitative) demos. will guide you through the steps needed to perform Homework 1.

Wed 27-Aug-08.  An overview of phylogenetic principles and their uses.

Phylogenetics attempts to uncover the branching pattern of the tree of life. Why?

1)  To understand the relationships among extant (and extinct) organisms -- where do the twigs and branches fit on the tree?

2) To understand the effect of history and phylogeny on development, function, adaptation, ecology, molecular evolution, behavior, mating systems, life cycles, speciation and biogeography. By standing back and looking at broad patterns we can see things we would never see with a narrow reductionist view. A few diverse examples should drive home the point:

How universal is sex? In sexual organisms how is sex (gender) determined?

How has flight evolved? Are the functional mechanisms similar or different in insects vs. vertebrates? Bats vs. birds?

How do new species arise? Do some lineages have more rapid radiations of species? If so, what factors make for rapid diversification in a lineage?

Are flippers, wings and arms homologous? That is, do they all derive from the same structure in the most recent common ancestor.  How about eyes and antennae?

How do the segments of worms relate to the major body portions of vertebrates (or do they?).

Why does Colombia have more species of birds than any other country in the world?

A phylogenetic perspective is important to any of these kinds of questions, and many other very different sorts of questions.

3)  Because (in close focus) individuals and their relatives show a reticulate pattern of evolution. Any valid theory of phylogeny must ultimately be able to cope with the reticulate patterns of population genetics, just as population genetics must be able to explain phylogenetic patterns in order to be completely general. A major issue that arises from the nature of fine-grained change is the problem of gene trees vs. species trees.



Fig. 2.1. Diagrammatic representation of reticulated evolution. The fuzzy tree shows a clean dichotomous branching pattern. At the level of individuals and their relatives, however, many of the paths cross as shown in the blowup to the right. This "crossing" is reticulation.



Fig. 2.2.  Diagrammatic representation of the difference between a gene tree and a species tree. The fuzzy tree shows a  dichotomous branching pattern for four species -- A, B, C and D. The thin lines within the fuzzy species tree show a gene tree that does not agree with the species tree. Taxa A, B and C share an allele that B and C do not share with D. That is, the splitting of the two alleles (Type 1 shared by A, B and C vs. Type 2 found in D) occurred at the circled Node 1 BEFORE the split that separated B and C from D (circled Node 2). Ancestral polymorphisms, therefore, can lead to discrepancies between gene trees and species trees.  If we pin our entire inference on one or two genes, we may be misled. 



Fig. 2.3. Simple depiction of gene tree vs. species tree from the Avise text (Avise Fig. 4.14, p. 149).   In the left panel, the gene tree (thin black lines) is concordant with the species tree (A & B are joined, then C).  In the right panel, the gene tree unites B & C, with A splitting off from the common ancestor further back in time, whereas the species tree unites A & B, with C splitting off earlier in time.  By using many genes, we can reduce the risk of producing an incorrect tree. 

4)  To illuminate and be illuminated by paleontological evidence. Current diversity fails to account for many important evolutionary "dead ends".

5)  Phylogenetic principles can apply completely outside evolutionary biology. For example, the study of manuscript traditions has benefited from phylogenetic logic. Much of the "evolution" of "the web" is obviously driven by branching processes that are similar to those driving phylogenies.

Clades and ingroups. A core concept is the monophyly of clades -- an ancestor and all of its descendant clades. Sometimes we find that OTUs (Operational Taxonomic Units -- these might be anything from populations to orders to phyla) we have considered to be monophyletic clades are actually paraphyletic. A paraphyletic group is one that includes an ancestor and some, but not all, of its descendants. One concrete example is the jays of the genus Aphelocoma. Some evidence suggests that the Mexican Jay Aphelocoma ultramarina actually arose from the Western Scrub-Jay A. occidentalis AFTER that species had diverged from the Florida Scrub-Jay A. coerulescens. If that were correct, the OTU "Scrub-Jay" would be paraphyletic ("Scrub-Jay" would include some, but not all, of the descendants of the most recent ancestor -- the Mexican Jay was left out, but belongs in the clade). An additional important concept is ingroup and outgroup. An ingroup is an assemblage of OTUs (often assumed to be monophyletic) that is the focus of interest. We bring in an outgroup as a way of broadening the analysis and providing a root for the tree.

Characters and their states and codings.  The UPGMA  trees we will build by hand use a distance matrix as the basis for the algorithms.  In many cases, however, the data we use to construct hypothesized phylogenies are characters -- sequence data, bone morphology or other evidence. Characters can have various states (e.g., 0/1, Present/Absent, 0/1/2/3). For cladists, the key to a useful character is when it shows synapomorphies -- shared derived characters that unite the members of a monophyletic clade. The great nemesis of phylogenetic inference is homoplasy -- similarity in state for reasons other than common ancestry. For example, convergent evolution can result in homoplasy (e.g., the spines and succulent stems of New World cacti and African euphorbs). Another example is a trait that changes to a new state (e.g., a base pair in a sequence that changes from A to C) then changes back -- a reversal. Characters may be either discrete or continuous. The way one codes characters can have important effects on the resulting phylogenetic inferences. For example, how does one deal with polymorphisms in a taxon? (E.g., Taxon A has a character with States I and II, whereas Taxon B has states II and III -- the usual default, in contrast, is 0/1 presence/absence).
    Another issue, which arises when using DNA sequence data, is that of sequence alignment.  In order to decide where changes have occurred between two or more OTUs in a phylogeny, we need to decide how the stretches of sequence from the OTUs should be aligned.  This includes the problem of where to insert "gaps" -- places where the sequence for a particular OTU clearly lacks some of the sequence that is present in other OTUs.  Refer to the handout I gave out in class for an example (Fig. 14, p. 378 in the Molecular Systematics text).

MAJOR ALTERNATIVE APPROACHES:

I.  The cladistic or branching approach -- at its strictest, this approach is based largely on synapomorphies that serve to unite monophyletic clades. The cladistic tradition stems from the work of Willi Hennig (1966. Phylogenetic Systematics. U. Illinois Press). It is clearly the current dominant approach in the field of phylogenetics. A well-known contemporary "cladist" is avian phylogeneticist Joel Cracraft of the American Museum in New York. Important theorists in the field include Joe Felsenstein (U. Wash., developer of much of the maximum likelihood approach to phylogeny and author of the PHYLIP software; Dave Swofford (Lab. Mol. Systematics, Smithsonian; developer of much of the parsimony approach and author of the PAUP software); Wayne Maddison (U. of Arizona, author of MacClade software).  Some eminent biologists consider that strict adherence to cladistic principles leads to nonsensical conclusions that ignore many obvious and unavoidable complexities of the real world (such as the phenomenon of reticulate evolution).  [See Avise, J.C. 2000. Cladists in Wonderland. Evol. 54: 1828-1832. This is a very entertaining review of Species Concepts and Phylogenetic Theory. Q.D. Wheeler and R. Meier, eds.]

Parsimony vs. maximum likelihood vs. Bayesian methods: A major subcluster in essentially cladistic approaches is between parsimony and maximum likelihood approaches. Parsimony relies on the concept that the tree needing the fewest changes in state along its branches is the best (unfortunately, evolution is not always parsimonious). Maximum likelihood uses an explicit function that can be minimized to give the "most likely" explanation for the data (unfortunately evolutionary outcomes are not always the result of the most likely process). [See pp. 87-89 of Hall, B.G. 2001. Phylogenetic Trees Made Easy. Sinauer Assoc., Sunderland, MA].  An approach that is rapidly gaining ground uses Bayesian methodology to construct trees.  [See Lewis, P.O., and D.L. Swofford. 2001. Back to the future: Bayesian inference arrives in phylogenetics. Trends Ecol. Evol. 16: 600-601 and pp. 120-122 in Phylogenetic Trees Made Easy].
Character states vs. distance measures: A major implementation approach (with no unifying philosophical theme and sometimes involving no biological assumptions) is the use of distance techniques (e.g., Cavalli-Sforza chord distances or Nei's distances). The UPGMA, Fitch-Margoliash and neighbor-joining approaches you will practice, can all be constructed from distance matrix inputs. The alternative (used, for example, with sequence data and morphological characters) is a matrix of character states -- often a binary 0/1 matrix. We will examine the use of some of these data matrices in software practica, but will not work through them "by hand".
It may be the case that the choice of methodology is less important than the quality and quantity of the input data.  [See Hughes, A.L. 2002. Strength in numbers. Nature 417: 795].
II.  Phenetic or clustering approaches -- now very little used. The "numerical taxonomy" of Sokal and Sneath (1963. Principles of Numerical Taxonomy. W.H. Freeman Co.) relied on clustering OTUs based on their overall similarity in a multivariate analysis of multiple characters and character states.

For a time considerable excitement and optimism existed about the potential for a "molecular clock" an underlying constant rate of mutation and change in all lineages that would allow dating of phylogenetic splits. It is now clear that no absolute, unchanging linear molecular clock exists. Nevertheless, rates of molecular change may often be essentially linear and invariant within particular OTUs, allowing interplay between molecular and paleontological evidence, for example.  In a few exceptional cases geological and biological history provide an unambiguous sequential record that can be timed fairly precisely.  An interesting application for avifauna and geology of the Hawaiian archipelago is: Fleischer, R.C., C.E. McIntosh, and C.L. Tarr. 1998. Evolution on a volcanic conveyor belt: using phylogeographic reconstructions and K-Ar-based ages of the Hawaiian Islands to estimate molecular rates. Mol. Ecol. 7: 533-545.

In parsimony approaches, branch length may refer to the number of presumed state changes (evolutionary steps) rather than time per se. The intent of parsimony is to find the trees that require the fewest steps. In maximum likelihood approaches, one is often trying to minimize a function describing the amount of evolutionary change -- the top priority, however, is the likelihood of the tree given the matrix of data values rather than the minimization of the number of steps (that is, the maximum likelihood tree may not be the most parsimonious tree).

Taxon sampling

An unsettling aspect of phylogenetic analyses is their sensitivity to the number and nature (e.g., choice of outgroup) of the taxa included in the analysis.  Incomplete taxon sampling has often been cited as a flaw in studies that precede a later analysis (based on more taxa) that supports a different phylogeny.  It may be the case, however, that the number of taxa included is less important   than the quality and quantity of the input data.  [See Rosenberg, M.S., and S. Kumar. 2001. Incomplete taxon sampling is not a problem for phylogenetic inference. PNAS USA 98 (19): 10751-10756].

Some pitfalls of molecular phylogenetics:

It has become steadily easier to apply molecular data to phylogeny reconstruction. One of the problems this raises is that traditional systematists with a firm grounding in the intricacies of morphological, distributional and behavioral variation in a given taxon are becoming less "fashionable". Wayne Maddison [Molecular Systematics 1996, Ch. 3, pp. 57-61] and others are concerned about a loss of emphasis on basic natural history and a strong organism-centered focus. Return to top of page