A Comparison of Phylogenetic Tree Algorithms

1
A Comparison of Phylogenetic Tree Algorithms
Alicia Morelli Bioinformatics Major Project
Advisor Ralph Morelli April 27, 2006
2
About the Project

• Essentially,
• Compares resulting trees produced by three tree
  creation algorithms
• Why do it?
• Challenging, Instructive but not Impossible
• What are the components?
• Tree Creation Algorithms
• UPGMA(1960s)
• Neighbors Relation (1977, S. Sattath, A Tversky)
• Neighbor Joining (1987 N. Saitou, M.Nei)
• GUI
• Create and mutate strand of DNA
• Convert DNA data to distance matrices
• Run algorithms and produce trees
• Algorithm Comparison
• Compares two algorithms at a time

3
What is a phylogenetic tree?
Leaves
Species
Leaves
Leaves
Species
Species
Branches
Branches
Relationships
Relationships
Nodes
Ancestors
Nodes
Ancestors
Branches
Relationships
Nodes
Ancestors
4
Tree Creation Flowchart
Newick Format (((((B,F),G),(A,D)),C),E)
Raw Data Taxa Characters Species
A ATGGCTATTCTTATAGTACG Species B ATCGCTAGTCTTATATT
ACA Species C TTCACTAGACCTGTGGTCCA Species
D TTGACCAGACCTGTGGTCCG Species E TTGACCAGTTCTCTAGT
TCG
Distance Based Algorithms
Tree Drawing Algorithm
Tree Creation
UPGMA
Simplified
Pairwise Alignment Algorithm
Strand Comparison
Neighbors Relation
Neighbor Joining
Matrix Reduction Algorithm
5
Overview of a Distance Based Algorithm
6
Differences and Similarities

• All Algorithms
• Pick a value in the distance matrix
• Reduce the matrix
• Repeat
• UPGMA
• Uses the smallest distance in the matrix
• Reduces by averaging old distances
• Neighbor Joining
• Uses largest value in an intermediate matrix
• Reduces using an adjusted average
• Neighbors Relation
• Use largest value in an intermediate matrix
• Reduces by averaging old distances

7
The Basis of Comparison

• Different topologies
• How often do the different algorithms produce the
  same topologies?
• Expectation
• Neighbor Joining and Neighbors Relation will
  produce trees with the same topologies
• UPGMA will produce trees with different topologies

8
Run Applet!
9
Results
10
Results!
11
Conclusion

• Performance
• NJ and NR are more likely to create same topology
  tree
• UPGMA vs NR and UPGMA vs NJ produce same topology
  trees more often than would happen randomly
• Future
• Create more accurate DNA mutation simulation
• Implement Tree Drawing Algorithm in GUI

12
Citations

• Java Resources
• Morelli, Ralph. Java, Java, Java Object Oriented
  Problem Solving. 2nd ed. Prentice Hall, 2002.
• JavaTM 2 Platform Standard Edition 5.0 API
  Specification. 2004. Sun Microsystems, Inc. 26
  April 2006. http//java.sun.com/j2se/1.5.0/docs/ap
  i/.
• Gilleland Michael. Combination Generator. 2006.
  Merriam Park Software. 26 April, 2006.
  http//www.merriampark.com/comb.htm.
• ColorPane.java. Code Examples from Java Swing
  2E. 2003. O'Reilly Associates. 26 April 2006.
  http//examples.oreilly.com/jswing2/code/ch22/Colo
  rPane.java.
• Bioinformatics Resources
• Krane, Dan and Michael Raymer. Fundamental
  Concepts of Bioinformatics. San Francisco
  Benjamin Cummings, 2003.
• Opperdoes, Fred. Neighbor-Joining Method. 1997.
  Christian de Duve Institute of Cellular
  Pathology. 26 April 2006. http//www.icp.ucl.ac.be
  /opperd/private/neighbor.html.
• Felsenstein, Joe. NEIGHBOR -- Neighbor-Joining
  and UPGMA methods. 2006. Phylip Home Page. 26
  April 2006. http//evolution.genetics.washington.e
  du/phylip/doc/neighbor.html.
• Images
• BembidionSpecies.jpg. 2005. Tree Of Life Web
  Project. 26 April 2006. http//tolweb.org/tree/lea
  rn/concepts/ConceptsImg/BembidionSpecies.jpg.
• Stewart, C.B., Pupko, T. Phylogenetic Analysis
  Based On Two Talks. Benny Chor Homepage. 26
  April 2006. http//www.cs.tau.ac.il/bchor/CG05/CG
  7-trees.ppt.
• Application of UPGMA Clustering Method On
  Selected Cytochrome C Data to Calculate
  Phylogenetic Relations. 2006. New Mexicans For
  Science and Reason. 26 April 2006.
  http//www.nmsr.org/upgma2.gif.
• Savai, P., Abulleef, H., Lee, L., Skvortso, D.
  Possible Rooted Trees 2002. Phylogenetic
  Analysis. 26 April 2006. http//www.cmb.usc.edu/cb
  mp/2002/PhylogeneticAnalysis/possible_rooted_trees
  .jpg.
• Thanks to
• Ralph Morelli and Chris Armen for their help with
  the project!

13
THE END