Reconstructing accurate trees is one of the most challenging topics in phylogenetics. In this talk, two new algorithms will be presented in attempt to improve the accuracy of gene tree building by incorporating species phylogeny. Both of them are presented in the CFG (Context-Free Grammar) framework. The first algorithm is called tree merge, which takes a set of gene trees built by different algorithms and then merges them together to get a new tree that can be better than all the input. The second algorithm calculates a likelihood with respect to the species evolution and incorporates this likelihood to the likelihood calculated from the sequence alignment. Maximum likelihood tree can then be searched. The two algorithms have been implemented in NJTREE software which is the core engine of the TreeFam database. NJTREE is also extensively used in the orthology prediction pipeline of Ensembl Compara, and is proved to be more accurate than traditional algorithms in reconstructing phylogenetic trees in a multi-gene family.
Related Links
* http://treesoft.sourceforge.net - NJTREE software website
* http://www.treefam.org - TreeFam database
view more