1 Treebuilding - Trees *are* Evolution
1
2
3
1
2
3
ONLY ONE TRUE TREE
1
Treebuilding 2 - Tree Parts
The Root Internal
Branches
Terminal
Internal nodes
A
B
C
D
Terminal nodes
The Root is a special node: it signifies the last common ancestor of all represented sequences Internal nodes are last common ancestors of subsets of the sequences Terminal nodes signify extant sequences Branches are the lineages; they are bounded by nodes
2
Treebuilding 3 - Tree Displays
# of events, rate
=
A B C D A B C D
=
B C A D
This axis has no meaning Sometimes trees are displayed on their side: B C A D Same topology, with branch lengths shown (preferable): B C A D
3
Treebuilding 4 - Rooted vs. Unrooted
A 1 B 2 1 2
C D
Same tree!
A
B
C
D
Anunrootedtreecanbeobtainedfromarootedtreebyremovingtheroot.Thetopology remainsthesame. Virtuallyalltreebasedinference(buildingandrateestimation)isdoneonunrootedtrees becausetheyhavecertainconvenientproperties.
4
Treebuilding 5 - Rooting the Unrooted Tree
ButButButBut what if we want a rooted tree??? UseanOUTGROUPwithwhichtorootthetree
This is the root of the tree were interested in: chordate ancestor
Dros Droso- Ciona Fish phila You Ciona
Fish You
5
Knowledge of the outgroup must be based on information independent from the analysis at hand
Treebuilding 6- Rooting the Unrooted
A
One unrooted tree for three sequences
1 2 3
Now lets add an outgroup to root the tree of A,B,C
B O A
C O B
1
B C
2
C A
O C
3
A B
O
A
B
C
O B
A
C
O
C
B
A
6
Treebuilding 7 - Number of Possible Trees 1
A
One unrooted tree for three sequences Three unrooted trees for four sequences
B O A B C O B
C C A O C A B
How many unrooted trees for 5 sequences?
7
Treebuilding 8 - Number of Possible Trees 2
Number of unrooted trees
(n = number of sequences) =
(2n-5)! 2n-3(n-3)!
for: 10 20 30 50 100
2027025 221643095512140477293
36
8.7x10
76
1.3x10 mycalculatorcan'tdothis
8
Treebuilding 9 - Alignment 1
Before Trees can be built, we must have a Multiple Alignment of homologous sequences. The alignment needs to be accurate, i.e., not only do the sequences need to be homologs, but Homologous Sites need to be correctly lined up in the columns of the alignment.
9
Treebuilding 10 - Alignment 2
Indelsobsurehomology.Becauseweneedtobesurethatwehave alignedhomologouspositions,regionsofuncertainhomologyare excludedfromanalysis. Atreegiventoyoubyamultiplesequencealignmentalgorithmisan impression,notarobusttree Itsoktoanalyzepartsofproteins(thosethatalignwell),andifa speciestreeistobebuilt,lotsofpartsfromdifferentproteins(more data=>betterresult) Theseconceptsalsoapplytosequencealignmentandtreebuilding usingDNAsequences
10
Treebuilding 11 - Alignment 3
Protein Multiple Alignment
Confined in scale ... proteins just arent very long ProbCons (Do, Batzoglou; probcons.stanford.edu) MUSCLE (Edgar, www.drive5.com) ClustalW (old workhorse, try not to use it)
Genomic DNA Multiple Alignment
Area of active research by several groups because
scale of problem is huge when genomes are aligned large-scale evolutionary events (large indels, rearrangements, duplications) require homology detection before actual nucleotide alignment
MLAGAN (Brudno, Batzoglou) http://lagan.stanford.edu/lagan_web/index.shtml TBA, BlastZ/MultiZ (Miller, Kent) Revolver (Asimenos, Edgar, Sidow, Batzoglou)
Gene211, Genomics (Winter): Genome concepts, PERL
11
Treebuilding 12 - Distance Methods 1
1. Convertsequencedatainto matrixofallpairwisedistances (countdifferencesandapply multiplehitcorrection) Fitthedistancestoa bifurcatingtree(veryfast)
1.
1234 10 0.40 0.50 0.65 20 0.30 0.45 30 0.35 40
Which Tree is the Right One? 1 2
0.3 0.1 0.1 0.25 0.1
Note:Distancemethodslose information(becauseofthe conversiontopairwisedata);only preferableifaccuracyhastobe sacrificedforspeed.
3 OR 4
1 4
2 3
12
(or the third possibility)
Treebuilding 13 - Distance Methods 2
Algorithmsthatclusterthemostsimilarsequencesfirstoften givethewronganswer
thisstatementalsoappliestoyourbrain
NeighborJoiningissmarterandisthedefactostandardfor distancemethodsbecauseofitsspeed Distancemethodsshouldonlybeusedwhenspeedlimitations precludetheuseofLikelihoodbasedcharacterstate methods(next)
13
Treebuilding 14 - Character-state Methods
...donotconvertsequencesintodistancesbutassesswhetherthe patternsofstatesatallsites/positionsinthealignment... Seq1...GATCG... Seq2...AATCG... Seq3...AATAA... Seq4...GATAA... ...favoronetreeoveranother(fromamongthetestedtrees)
1 3 1 2 1 2
vs.
2 4 3 4
vs.
4 3
(Maximum)Parsimonyand(Maximum)Likelihood Both require searching for the best tree, i.e., evaluating one tree after another. Smart algorithms do a good job at that unless the number of sequences is very large.
14
Treebuilding 15: Parsimony 1
Parsimony:OccamsRazor.Findthetreethatexplainsthesequence databyinvokingtheleastnumberofchanges. Basic Idea: Seq1...G... Seq2...G... Seq3...A... Seq4...A... favors:
G1
3
The other two trees would require two changes: Tree 2:
G1
2
Tree 3:
G G1
2
G
A
A3
4
A
A4
3
A
or
G2
4
or
2
A
G1
G
G1
2
G
One change from G to A (or A to G)
A3
4
A
A4
3
A 15
Treebuilding 16 - Parsimony 2
Pos/site123456 Seq1...ATACGA... Seq2...ATCCTC... Seq3...ATCCTT... Seq4...ACCTTA...
Pos 1: No change Pos 2: One change on lineage to 4 Pos 3: One change on lineage to 1 Pos 4, 5: Same as 2,3
Site 6: Two changes for all trees, no so discrimination: Tree 1
1A T3 1A
Tree 2
C2 1A
Tree 3
C2
or
2C
A4
3T
A4
4A
T3
None of these positions are informative as none of them favor one tree over another
16
Treebuilding 17 - Parsimony Exercise
Thetreerequiringthefewestchangeswins.
111 Pos:123456789012 Letscount changesfor eachposition. Seq1:GATCGATCGATC Seq2:AATCGTTAGATG Seq3:AACCGTGCAAAC
1 2 1 3 1 O
3 O 2 O 2 3
Outg:AACCGCGTGCAG
1 0 1 0 0 2 1 2 1 1 1 2
12
1 0 2 0 0 2 2 2 1 1 2 1
14
1 0 2 0 0 2 2 2 1 1 2 2
15
17
Treebuilding 18 - Summary
Parsimonyimplicitlyassumesthattherearenomultiplehits.Ithasno stochasticmodelforsubstitution.
Greatforlearning,notsogoodforactualscience Substitutionsmustberareonanygivenlineage Parsimonyscountingonlyapproachmaybeokforverycloselyrelated sequences
Likelihoodisthewaytobeonsoundstatisticalfooting(versusparsimony) withoutlossofinformation(versusdistancemethods).UseSEMPHYfor treebuilding,PAMLforinferringevolutionaryparameters. However(1):Forapplicationsotherthansequencebasedtreebuilding,a parsimonyapproach(Occamsrazor)canbeappropriate However(2):Formappingtraitsontoatree,parsimonycanbe appropriate
18
next
Do treebuilding
distance character states
Understand tree-based inference
orthology and paralogy evolutionary events
Discuss evolutionary constraint and how it matters
proteins genomes
Explore genome structure and how it is shaped by evolution
19
Gene Duplication 1
Both paralogs are retained if:
High expression level is needed
Actins, Tubulins, Cow stomach lysozyme, rRNA Concerted evolution after duplication by gene conversion or cycles of duplication and loss
The universal symbol for Gene Duplication
Neofunctionalization
New biochemical function or new expression domain p53 vs p63,73; RNA polymerases; hemoglobins; etc. independent evolution after duplication
Subfunctionalization
of Proteins specific function or of expression region many developmental regulators in vertebrates independent evolution after duplication
20
Gene Duplication 2
Species Tree
O
1
2
3
Where is the Gene Duplication on the Species Tree?
Gene Tree
a
b
O
1
2
3
1
2
3
21
Gene Duplication 3
There are only two types of nodes in a tree
Gene Duplication Last Common Species Ancestor
a
b
O
1
2
3
1
2
3
If independent evolution in the lineages was precipitated by Paralogy Orthology
a gene duplication: a speciation event:
(give some examples of orthologs and paralogs from the tree above) (hint: trace back to the last common node; is it a gene dup or a last common species ancestor?)
When a gene duplication occurred is inferred by comparison of the Gene Tree with the Species Tree
22
Gene Duplication 4 - Some Exercises
Human
Frog
ZFish
Ciona
ZFish1 ZFish2
Frog
Ciona
Hum2
Frog Hum1 ZFish
Ciona
Human ZFish
Frog
Ciona
23
Gene Duplication 5 - Rerooting Gene Trees
Same Tree
Left: rooted arbitrarily by treebuilding program Right: rooted by user under reasonable outgroup assumption TOPOLOGY IS THE SAME
24
Gene Duplication 6 - Some Gene Trees
p53, p63, p73
Patched
25
Mapping Traits onto Tree 1
Gene Tree Sequence name Mollusk p53/63/73 Trait: C-terminal domain present
Vertebrate p63
present
Vertebrate p73
present
Fish p53
absent
Chick p53
absent
Mouse p53
absent
26
Mapping Traits onto Tree 2
Gene Tree Sequence name Fly YFG Trait: expression domains eye
Vertebrate YFG1
fin/limb bud
Vertebrate YFG2
eye, fin/limb bud
Fish YFG3
eye
Chick YFG3
eye
Mouse YFG3
eye, placenta
27
next
Do treebuilding
distance character states
Understand tree-based inference
orthology and paralogy evolutionary events
Discuss evolutionary constraint and how it matters
proteins genomes
Explore genome structure and how it is shaped by evolution
28
Source and Use of Evolutionary Constraint
Constraint on Proteins:
- Structural: Folding or Packing - Functional: Catalysis or Binding
Constraint on DNA:
- Structural: ? - Functional: TFBS, Exons, etc
Constraint
determined
DeleteriousnessofMutation
determined
infer!
EvolutionaryRate estimate from extant homologous sequences by standard molecular evolutionary methodology
29
Constraint vs Conservation
Conservationimpliesbinarylogic:eitheritsconservedorits notconserved Constraintimpliesquantification,adegreetowhichthe evolutionofafunctionalunithasbeenrestrained
Theupperlimitofconstraintisnoevolutionarychange. Thelowerlimitofconstraintisneutral. Thewholerangeofevolutionaryratesisinbetween.
30
Alignments and Trees
Analyses of Constraint Require
Alignment of sequences
Protein or DNA
Tree relating the sequences
Topology only ... ... or species tree with neutral branch lengths (There are other types of constraint that evolutionary rates of single base substitutions do not capture -- not in this course)
31
Constraint in Proteins
Structure and Function
Inference of Constraint by Evolutionary Analysis
32
Constraint on Proteins - Rates of Evolution
p53, p63, p73
Patched
33
Example: p53 Single Site Constraint
Colors correspond to strength of evolutionary constraint. Red, weak constraint; Blue, strong constraint; Rainbow in between.
34
Example: p53 Regional Rate Plot
Rates of evolution of neighboring sites are averaged and smoothed in order to see regional features such as domains Normalized Rate of Evolution
Highly Constrained Regions
mean rate
Position in the Alignment
35
Regional vs Single Site Constraint
36