IE68 Biological databases Phylogenetic analysis

1
IE68 - Biological databasesPhylogenetic analysis
2
Phylogenetic analysis

• Phylogeny
• a reconstruction of the evolutionary
  (genealogical) history of a group of
  organisms/genes or proteins from biological data
• organisms populations, species, genera,... gt
  taxa gt operational taxonomic units (OTUs)
• data molecular, morphological,
  archaeological,... gt characters
• Phylogenetic tree
• the graphical reconstruction of a phylogeny
• tree structure phylogram, cladogram

3
Phylogenetic tree
A tree consists of nodes connected by branches
polytomy
A B C D E
gt OTUs for which we have data
outgroup/midpoint
gt Ancestor of all the taxa that comprise the tree
notation ((A,B),(C,D,E))
4
Phylogenetics ltgt Phenetics

• Phenetics method of grouping taxa that is based
  on overall (dis)similarities of characters gt
  with no reference to evolution!
• Phylogenetics method of grouping taxa that is
  based on shared derived characters
  (synapomorphies) or a model of evolution

5
Why do we need phylogenies?

• Intrinsic interest in the tree gt tree of life
• origin of organisms

6
Why do we need phylogenies?

• Phylogenies can also be used as tools for
  investigating other problems
• e.g. biogeography
• phylogeny reflects the order of separation of
  the areas the different taxa occupy

T
7
Why do we need phylogenies?

• Phylogenies can also be used as tools for
  investigating other problems
• e.g. forensic science

8
(No Transcript)
9
Phylogenetic analysis

• Molecular Phylogenetics
• reconstruction of the evolutionary (geneological)
  history of a group of organisms from molecular
  data, i.e. DNA or protein sequences
• In this lecture, we will focus on phylogenetic
  analysis of organisms based on DNA sequence data

10
Molecular phylogenetics approach

• Step 1 PCR with primers that target cytoplasmic
  DNA or nuclear loci of taxa, followed by DNA
  sequence analysis
• Step 2 Multiple DNA sequence alignment
• Step 3 Phylogenetic analysis

11
PCR and DNA sequencing

• Which loci?
• DNA sequence information, primers, variability,
  single or low-copy, orthologous, neutral,
  recombination...
• Gene trees versus organismal trees
• phylogenies for genes do not always match those
  of their corresponding organisms gt analyse more
  than one gene

12
Confounding influence of gene duplication
2 types of homology orthology (speciation) and
paralogy (gene duplication)
13
Lineage sorting and coalescence
species alleles
14
Molecular phylogenetics approach

• Step 1 PCR with primers that target cytoplasmic
  DNA or nuclear loci of taxa, followed by DNA
  sequence analysis
• Step 2 Multiple DNA sequence alignment
• Step 3 Phylogenetic analysis

15
Multiple DNA sequence alignment

• Problem alternative alignments
• possible to align any two sequences by
  postulating some combination of gaps
  (insertion/deletions indels) and substitutions
• gt which one to choose?
• Basic task of sequence alignment is to find the
  alignment with the highest similarity, smallest
  distance, or lowest overall cost

16
Multiple DNA sequence alignment

• 2 sequences scoring scheme gt optimal alignment
• Scoring scheme
• - scoring matrix distance weights or similarity
  scores for each pair of aligned bases e.g.
  transition transversion matrix
• A T G C
• A 0 5 1 5
• T 5 0 5 1
• G 1 5 0 5
• C 5 1 5 0
• - gap weight, cost or penalty

17
Multiple DNA sequence alignment

• Cost of an alignment D s wg
• s no of substitutions, g total length of
  gaps
• w gap penalty cost of gap relative to
  substitution
• Gap penalty W makes implicit assumptions about
  how the sequences have evolved
• if indels are thought to be rare, then W should
  be large (and vice versa)
• gt have to use knowledge of biology e.g.
  translation (3 bp indel, position),
  transitionltgttransversion, ...

18
Multiple DNA sequence alignment

• Software programs e.g. CLUSTALW (global
  alignment)
• http//www.ebi.ac.uk/clustalw/index.html
• The optimal alignment is not always the true
  alignment gt new developments phylogenetic
  analysis without the multiple DNA sequence
  alignment step

19
Molecular phylogenetics approach

• Step 1 PCR with primers that target cytoplasmic
  DNA or nuclear loci of taxa, followed by DNA
  sequence analysis
• Step 2 Multiple DNA sequence alignment
• Step 3 Phylogenetic analysis

20
Inferring phylogenies from DNA sequences
C
Sequence alignment A ..AGCGTCT..B
..AGCGTGT..C ..AGGAGT..
A
B
Phylogenetic methods
unrooted tree
A
B
taxa
characters
C
rooted tree
21
Phylogenetic methods
Character-based methods
Non character-based methods
Methods based on an explicit model of evolution
Maximum-likelihood methods
Pairwise distance methods
Methods not based on an explicit model of
evolution
Maximum parsimony methods
22
Pairwise distance methods
3 taxa, 3 sequences

• Dissimilarity matrix count the number of
  differences between all possible pairs of
  sequences
• Convert dissimilarity to evolutionary distance by
  correcting for multiple events per site according
  to a certain model of evolution
• Infer tree topology on the basis of the
  evolutionary distances by using a clustering
  algorithm or optimality criterion

1 2 31 2 0.263 0.20 0.33
1 2 31 2 0.323 0.23 0.44
tree
23
Models of sequence evolution
expected ? observed difference gt correction
(linear) (not linear)
Apply a substitution model that tries to estimate
the correct number of substitutions
24
Models of sequence evolution

• Distance correction methodsconvert observed
  distances into measure that correspond to ACTUAL
  distance
• Several methods have been proposed, all with
  different assumptions about the nature of the
  evolutionary process
• Essentially they differ by the number of
  parameters they include
• We can use a general framework to show how these
  models are inter-related

25
Substitution models general framework
26
Substitution models general framework
27
e.g. Model of Jukes Cantor (JC)

• One of the first proposed perhaps the simplest
  model of evolution
• Assumes that all four bases have equal frequency
  and that all substitutions are equally likely
• Under this model, the distance between any two
  sequences is given by d -3/4ln(1-4/3p), where p
  is the proportion of nucleotides that are
  different in the two sequences

28
e.g. Kimura 2 parameter model (K2P)

• incorporates the observation that transitions
  accumulate more rapidly than transversion
• assumes all four bases have equal frequencies
  but that there are 2 rate classes for
  substitutions
• Under this model, the distance between any two
  sequences is given by d 1/2ln1/(1-2P-Q)
  1/4ln1/(1-2Q), where P and Q are the
  proportional differences between the two
  sequences due to transitions and transversions,
  respectively

29
Substitution models

• Other models adding more parameters
• Felsenstein model (F81)
• variation in base composition gt base frequency
  f ?A ?C ?G ?T may vary
• Hasewaga Kishino Yano (HKY) model
• unequal base frequency, transition/transversion
• General reversible model (REV) unequal base
  frequency, all six pairs of substitutions have
  different rates
• gt ideally, we want the simplest model we can get
  away with that still yields a reasonable
  estimate

30
Substitution models

• Assumptions of these models
• all nucleotide sites change independently
• base composition equilibrium
• substitution rate is constant over time and in
  different lineages
• each site in a sequence is equally likely to
  undergo substitutiongt gamma distribution has a
  parameter that specifies the range of rate
  variation among sites model ?

31

• Pairwise distance methods
• Dissimilarity matrix count the number of
  differences between all possible pairs of
  sequences
• Convert dissimilarity to evolutionary distance
  by correcting for multiple events per site
  according to a certain model of evolution
• Infer tree topology on the basis of the
  evolutionary distances by using a clustering
  algorithm

3 taxa, 3 sequences
1 2 31 2 0.263 0.20 0.33
1 2 31 2 0.323 0.23 0.44
tree
32
Clustering methods

• Clustering methods follow a set of steps (an
  algorithm) and arrive at a tree
• UPGMA (Unweighted Pair Group Method using
  Arithmetic Averages) results in an rooted and
  additive tree with molecular clock
• Neighbor-joining results in an unrooted and
  additive tree
• Other approaches least-squares, Fitch, Kitch,...

33
UPGMA clustering
A B C B 2 least differences C 4 4 D 6
6 6
1
A
1
B
Compute new distances between (AB) and other
OTUs d(AB)C (dAC dBC) /2 4 d(AB)D (dAD
dBD) /2 6
34
UPGMA clustering
1
A
AB C C 4 D 6 6
1
1
B
2
C
1
A
1
Compute new distances between (ABC) and other
OTUs d(ABC)D (d(AB)D dCD) /2 6
1
B
1
2
C
3
D
35
Clustering methods

• UPGMA additive and ultrametric distancesgt
  assumes a molecular clock gt very sensitive to
  unequal rate of evolution! gt relative-rate test
• Use other clustering methods for phylogenye.g.
  Neighbor-joining
• Goodness of fit statistics to select the
  metric tree that best accounts for the observed
  distances

36

• Pairwise distance methods
• Dissimilarity matrix count the number of
  differences between all possible pairs of
  sequences
• Convert dissimilarity to evolutionary distance
  by correcting for multiple events per site
  according to a certain model of evolution
• Infer tree topology on the basis of the
  evolutionary distances by using an optimality
  criterion

3 taxa, 3 sequences
1 2 31 2 0.263 0.20 0.33
1 2 31 2 0.323 0.23 0.44
tree
37
Minimum evolution

• Distance matrix gt unrooted metric trees
• Each tree has a length L, which is the sum of all
  the branch lengths
• Optimality criterionthe minimum evolution tree
  ME is the tree which minimizes L

38
Pairwise distance method

• Advantages
• very fast
• based on a model of evolution
• Disadvantages
• sequence information is reduced to one number
• branch lengths may not be biologically
  interpreted
• most methods provide only one tree topology
• dependent on the model of evolution used

39
Phylogenetic methods
Character-based methods
Non character-based methods
Methods based on an explicit model of evolution
Maximum-likelihood methods
Pairwise distance methods
Methods not based on an explicit model of
evolution
Maximum parsimony methods
40
Character-based methods

• Character-based (discrete) methods operate
  directly on sequences, rather than on pairwise
  distances
• Two major discrete methods
• Maximum parsimony (MP) chooses tree(s) that
  require fewest evolutionary changes
• Maximum Likelihood (ML) chooses tree(s) that is
  the one most likely to have produced the observed
  data

41
Maximum parsimony

• Maximum parsimony infers a phylogenetic tree by
  minimizing the total number of evolutionary steps
• Principle
• Investigate all possible tree topologies
• Reconstruct ancestral sequences
• Choose topology with smallest number of steps

42
Maximum parsimony - principle
A
1
3
2
4
1
2
B
3
4
1
2
C
3
4
possible tree topologies
43
Maximum parsimony - principle
44
Maximum parsimony - principle
45
Maximum parsimony - principle
46
Maximum parsimony - generalized

• In previous example, cost of each substitution
  was one step gt equal weight
• Instead, we can use different costs for different
  types of change (e.g. transitions vs
  transversions) to better match our assumptions
  about evolutionary processes gt weighted
  parsimonyaccording to Dollo, Wagner, Fitch, ...

47
Maximum parsimony - characters
48
Maximum parsimony search methods

• Number of tree topologies Nu
  (2n-5)!/2n-3(n-3)!i.e., 3 sequences 1 tree, 4
  seq 3 trees, 5 seq 15, 6 105, gt the more
  sequences ( taxa), the more trees gt
  computationally expensive
• Finding optimal trees
• Exhaustive search limited number of taxa
  (lt10)find the minimum tree of all possible trees
• Branch and bound small number of taxa (lt18)find
  the minimum tree without evaluating all trees by
  discarding families of trees during tree
  construction that cannot be shorter than the
  shortest tree found so far
• Heuristic search large number of taxa

49
Maximum parsimony search methods
- Heuristic searchexplore a subset of all
possible trees, by using stepwise addition of
taxa plus a rearrangement process (branch
swapping), but not guaranteed to find the minimal
tree
Global optimum
Local optimum
50
Maximum parsimony - output

• Consensus treeMP can yield multiple equally
  most parsimonious (optimal) trees gt
  relationships common to all the optimal trees are
  summarized with a consensus tree
• Strict consensus includes splits found in all
  trees
• Majority-rule consensus includes splits found in
  the majority of the trees (gt 50)

51
Maximum parsimony - output

• Consistency index (CI) - Retention index (RI)
• measures of the parsimony fit of a character to a
  tree, or of the average fit of all characters to
  a tree
• more specifically index of how much homoplasy
  the constructed tree has
• Value from 0 to 1
• higher value gt less homoplasy

52
(No Transcript)
53
Parsimony branch support and tree stability

• Bootstrap analysis
• is a resampling technique used to measure
  sampling error
• gives an idea about the reliability of branches
  and clusters
• original dataset gt resample gt construct trees
  gt compare trees to original trees
• gt70 quite confident of tree topology
• Decay index (Bremer support)
• gives us a sense of how many steps would be
  required before a grouping collapses
• higher value gt better branch support

54
Maximum parsimony

• Advantages
• based on shared derived characters
• evaluates different tree topologies
• does not reduce the information
• Disadvantages
• computationally intensive for large datasets
• no correction for multiple mutations
• sensitive to unequal rates of evolution (long
  branch attraction)

55
Phylogenetic methods
Character-based methods
Non character-based methods
Methods based on an explicit model of evolution
Maximum-likelihood methods
Pairwise distance methods
Methods not based on an explicit model of
evolution
Maximum parsimony methods
56
Maximum likelihood

• Statistical method
• If given some data D and a hypothesis H, the
  likelihood of that data is given byLD Pr (DH)
• Which is the probability of D given H?

57
Maximum likelihood

• In the context of molecular phylogenetics
• D is the set of sequences being compared
• H is a phylogenetic tree
• We want to find the likelihood of obtaining the
  observed data given the tree
• The tree that makes the data the most probable
  evolutionary outcome is the Maximum Likelihood
  estimate of the phylogeny

58
Maximum likelihood

• In other wordsWhich tree is most likely to have
  yielded these sequences (observed data) under a
  given model of evolution (JC, K2P, ...)?

59
Maximum likelihood

• Advantages
• Statistically well founded
• Based on a model of evolution
• Evaluates different topologies
• Uses all sequence information
• Often yields estimates that have lower variance
  than other methods
• Disadvantages
• Very slow (computationally intensive)
• Dependent on the model of evolution used

60
Software programs for phylogenetic analysis

• Overview http//evolution.genetics.washington.edu
  /phylip/software.html
• Most widely used software programs
• PHYLIP free available (downloadable or online
  http//bioweb.pasteur.fr/seqanal/phylogeny/phylip-
  uk.html)
• PAUP user friendly but not free available

61
Phylogenetic information on the internet

• http//tolweb.org/tree/phylogeny.html
• http//www.treebase.org/treebase/
• ....

62
If you need more information

• Jacqueline Vander Stappen
• K.U.Leuven
• Laboratory of Gene Technology
• Kasteelpark Arenberg 21
• B-3001 Leuven
• Jacqueline.vanderstappen_at_agr.kuleuven.ac.be