The Tree Room : How to build a tree :
Other ways of building trees
Biologists use several different methods for reconstructing evolutionary trees. Parsimony analysis, discussed previously, is easy to explain and is commonly used; however, for complex molecular data sets, evolutionary biologists often turn to two other methods: Maximum Likelihood and Bayesian analyses. Based on probability theory, these are mathematical methods for building trees that can be used when something is known about the process of evolution underlying the data set i.e., when we have a "model" of how evolution works. For example, because of the structure of DNA, genetic mutations that change an A base to a G base are much more common than mutations that change an A base to a T or C base. Maximum Likelihood and Bayesian analyses can take this into account, generating trees that are more likely to be accurate based on biochemical information about which bases are likely to mutate into one another. Though the method used to reconstruct a tree (parsimony, Likelihood, or Bayesian) may affect the shape of the best-supported tree to some degree, different methods usually agree on the basic structure of the tree and additional data can then be used to sort out discrepancies.
The bases of DNA are different sizes. A and G are composed of two chemical rings, while C and T involve just one ring. Mutations that convert one base to another of the same size are more frequent than those that convert a base to one of a different size. Information like this can be used to build trees with the Maximum Likelihood and Bayesian methods.