Phylogeny by Hybridization

1
Phylogeny by Hybridization

• S. Angelov, B. Harb, S. Kannan,
• S. Khanna, J. Kim, and L. Wang.
• University of Pennsylvania

2
Motivation

• Estimating the phylogeny of species is a
  fundamental problem of comparative genomics
• Traditional methods use aligned molecular
  sequences from a small number of genes
• High throughput techniques are needed for
• Estimating the phylogeny of All Life
• Rapid identification of unknown organisms

3
Motivation (cont.)

• Tests using oligo hybridization arrays are one
  possible high-throughput approach
• Designing sequences of oligonucleotides (tags)
  for these tests presents a challenge
• Need homologous tagsi.e., shared tags among two
  genomes should indicate genealogical relationship
• Longer tags are more likely to be homologousbut
  the total address space is too large

4
Phylogenies

• Tree-graph representations of geneological
  relationships

5
Example
Unclassified Genome
ACGTCAG
TCGTACT
CATCGAC
GCGTACG
6
Example
Unclassified Genome
C G T
ACGTCAG
TCGTACT
CATCGAC
GCGTACG
7
Example
Unclassified Genome
C G T
ACGTCAG
TCGTACT
CATCGAC
GCGTACG
8
Example
Unclassified Genome
T A C
ACGTCAG
TCGTACT
CATCGAC
GCGTACG
9
Example
Unclassified Genome
T A C
T A C
ACGTCAG
TCGTACT
CATCGAC
GCGTACG
10
Our Research

• Design efficient algorithms for detecting all
  discriminating tags
• A tag is a short oligonucleotide sequence
• A tag is discriminating at some node if
• All genomes under one branch of contain
• No genomes under any other branch of contain

11
Outline

• Introduction
• Example
• A naïve algorithm
• Our basic algorithm
• Summary of results
• Experimental results
• Conclusion

12
Input and Goal

• Goal
• Find all discriminating tags for each node in
  

13
Naïve Algorithm

• Fix .
• Pick any in the left subtree of
• For each substring of check if
• Complexity

present in all genomes in the left subtree and
none of the genomes in the right subtree
14
Suffix Tree Example

• ACACG

• Every substring is a path from the root
• Every leaf corresponds to a suffix

15
Basic Algorithm

• Construct a generalized suffix tree for the
  species
• Each node in is a possible tag
• For each possible tag
• Determine the sequences of which is a
  substring
• Determine if is a discriminating tag
• Complexity

16
Basic Algorithm (cont.)

ACGTCAG
TCGTACT
GCGTACG
17
Summary of Results
18
Outline

• Introduction
• Example
• A naïve algorithm
• Our basic algorithm
• Summary of results
• Experimental results
• Conclusion

19
Cystic Fibrosis Phylogeny
1.2 MB Base Pairs / Specie
20
Cystic Fibrosis Tags
rat
cat
pig
dog
cow
fugu
zfish
tetra
chimp
mouse
human
baboon
chicken
21
Cystic Fibrosis Tags
rat
cat
pig
dog
cow
fugu
zfish
tetra
chimp
mouse
human
baboon
chicken
22
Random Permutation Phylogeny
23
Random Permutation Phylogeny Tags
24
Random Permutation Phylogeny Tags
25
Subphylogeny Test
rat
cat
dog
cow
fugu
zfish
chimp
human
baboon
chicken
26
Placing the Mouse

• Tags of length 20 for root node

• Mouse gene has one false positive (left branch)
• CAGACAGACAGACAGACAGA
• But matches only one of the extra right-branch
  tags
• Pig matches three of the extra right-branch tags

27
Future Work

• Tag prediction by evolutionary stochastic models
• Algorithmic Directions
• Too few tags Majority
• Too many tags Bipartition, tag clustering,
  etc.
• Approximate tags
• Experimental evaluation of complete genomes

28
Subphylogeny Test Tags
rat
cat
pig
dog
cow
fugu
zfish
tetra
chimp
mouse
human
baboon
chicken
29
Subphylogeny Test Tags
rat
cat
pig
dog
cow
fugu
zfish
tetra
chimp
mouse
human
baboon
chicken