Notes 20100713.115800 CS 798 Presentation Genetic Algorithm

From SnOwy - Ed's Wiki Notebook

Genetic Algorithm used for Phylogenies

• Phylogenetic tree reconstruction
• Parsimonious phylogenies
• Population of phylogenies
• Crossover: combines several parent phylogenies
• Mutation: randomly changes an individual
• Fitness: parsimony

Results

• Compares GaPhyl and Phylip
• Datasets: Lamiiflorae and Angiosperms (23, 49 respectively); phenome, binary characters (29, 61 respectively);
• Both GaPhyl and Phylip both have a stochastic component.
• GaPhyl was able to find 250 most parsimonious trees
• Phylip found a subset of 75
• Note: parsimony scores are the best found -- i.e. no guarantee that any have the best tree-- but these are the best.

Questions: Are we worried that there ARE so many best trees?

• In general, this is a common enough phenomenon that we've accepted it
• Err... however, likelihood methods are less likely to have such ties-- but only because their ranges are continuous
• Not sure -- probably not worried-- the angiosperms feature hybridization → phlogenetic network instead of a tree