Tree Pattern Matching in Phylogenetic Trees

1
Tree Pattern Matching in Phylogenetic Trees

• Automatic Search for Orthologs or Paralogs in
  Homologous Gene Sequence Databases
• By Jean-François Dufayard, Laurent Duret, Simon
  Penel, Manolo Gouy, François Rechenmann, and Guy
  Perrière

Presented by Jean Yeh
2
Background Information

• The authors have created three databases that
  gather genes into homologous families
• HOVERGEN vertebrates
• HOBACGEN prokaryotes
• HOGENOM completely sequenced organisms
• Among homologous genes, need to be able to
  differentiate orthologs from paralogs

3
Homologous Sequences

• Homologs Two genes related by descent from a
  common ancestral DNA sequence
• Orthologs Two genes in different species
  evolved from a single ancestral gene by
  speciation
• Paralogs Two genes related by duplication within
  a genome

4
Orthologs and Paralogs
http//www.ncbi.nlm.nih.gov/Education/BLASTinfo/or
thologs3.gif
5
Gene Function

• Gene function tends to change after gene
  duplication
• Orthologs are more reliable predictors of gene
  function than paralogs
• Evolutionary distance also plays a role
• Closely related paralogs probably more similar
  than distantly related orthologs

6
Goal

• Create algorithms that allow for automatic
  searching for orthologs or paralogs in their
  databases
• One algorithm for tree reconciliation
• One algorithm for tree pattern matching
• Implement under architecture used to query the
  databases

7
Tree Reconciliation

• Infers speciation and duplication events
• Compares gene tree G with species tree S to give
  a reconciled tree R
• Algorithm
• R S
• Step through G and R simultaneously
• If nodes are incongruent, insert duplication node
  in R and annotate gene losses

8
Tree Reconciliation
9
Tree Pattern Matching

• A tree pattern is a peculiar tree structure with
  taxonomic and evolutionary parameters contained
  in nodes and leaves
• Can be considered a subtree
• Want to match to a target tree
• E.g. pattern (X, Y, Z) matches ((X, Y), Z), (X,
  (Y, Z)), and ((X, Z), Y)

10
Tree Pattern Matching

• Uses a recurrence algorithm that takes into
  account different taxonomic levels as well as the
  specific branch constraints
• Cuts down on run time by checking the number of
  leaves in the pattern and the target tree
• Allows users to search for orthologs/paralogs

11
FamFetch Interface

• User interface to access the databases
• Incorporates both algorithms
• Pattern editor has two frames tool and pattern
• Pattern frame interactive editor to construct,
  load, save, and match patterns with a tree
  database
• Tool frame tools used in pattern frame

12
FamFetch
13
Tree Rooting

• For tree reconciliation, the trees must be rooted
• Authors use their reconciliation algorithm to
  find the most parsimonious solution the one
  that requires the least number of gene
  duplications
• Reconciliation algorithm relatively fast

14
Tree Pattern Search

• By forming their algorithm as a tree pattern
  search, the authors managed to increase possible
  queries for the users
• Can search for gene duplication or gene
  speciation events, not just orthologs and
  paralogs
• Also relatively fast algorithm, though lose the
  human flexibility of pattern matching

15
Automatic Search for Orthologs

• Previously done with pairwise BLAST searches and
  reciprocal hits
• Need all genes and if genes are wrong, results
  may be wrong
• Classifying genes into clusters of orthologs
  depends on evolutionary distance between species

16
Possible Improvement

• Have program estimate reliability of
  reconciliation
• While it allows for easier comparative sequence
  analysis, it was designed solely for databases
  the authors had already created
• Might be improved if it could be generalized for
  more databases