Comparing Phylogenies
Transcript of Comparing Phylogenies
![Page 1: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/1.jpg)
Norbert ZehDalhousie University
Comparing Phylogenies
Kernelization, Depth-Bounded Search and Beyond
![Page 2: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/2.jpg)
Phylogenetic Trees
Halophiles
Methanosarcina
Methanobacterium
Methanococcus
T. celerThermoprot.
Pyrodicticum
EntamoebaeSlimemolds Animals
Fungi
PlantsCiliates
Flagellates
Trichomonads
Microsporidia
Diplomonads
GreenFilamentous
bacteriaSpirochetesGram
positives
Proteobacteria
Cyanobacteria
Planctomyces
BacteroidesCytophaga
Thermotoga
Aquifex
Archaea EucaryotaBacteria
Comparing Phylogenies
Norbert Zeh
![Page 3: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/3.jpg)
Reticulation Events
Lateral gene transfer (subtree prune-and-regraft)
a b c d e
Comparing Phylogenies
Norbert Zeh
![Page 4: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/4.jpg)
Reticulation Events
Lateral gene transfer (subtree prune-and-regraft)
a b c d e
Comparing Phylogenies
Norbert Zeh
![Page 5: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/5.jpg)
Reticulation Events
Lateral gene transfer (subtree prune-and-regraft)
a b d eca b c d e
Comparing Phylogenies
Norbert Zeh
![Page 6: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/6.jpg)
Reticulation Events
Lateral gene transfer (subtree prune-and-regraft)
Hybridization
a b d eca b c d e
a b c d eComparing Phylogenies
Norbert Zeh
![Page 7: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/7.jpg)
Reticulation Events
Lateral gene transfer (subtree prune-and-regraft)
Hybridization
a b d eca b c d e
a b c d eComparing Phylogenies
Norbert Zeh
![Page 8: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/8.jpg)
Reticulation Events
Lateral gene transfer (subtree prune-and-regraft)
Hybridization
a b d eca b c d e
a b c d eComparing Phylogenies
Norbert Zeh
![Page 9: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/9.jpg)
Reticulation Events
Zeina the ZonkeyOwklawn Farm Zoo, Nova Scotia
Lateral gene transfer (subtree prune-and-regraft)
Hybridization
a b d eca b c d e
a b c d eComparing Phylogenies
Norbert Zeh
![Page 10: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/10.jpg)
Reconciling Phylogenies
1. Tree distances
ā¢ SPR distance: number of SPR operations to transform one tree into theother
NP-hard [Bordewich/Semple 2005]
ā¢ Hybridization number: minimum number of nodes with two parents inany network that displays both trees
NP-hard [Bordewich/Semple 2007]
ā¢ Robinson-Foulds distance: number of bipartitions that disagree
Linear-time, but . . .
a b c d e f
Ļ
fe dcba
Ļ
a b c d e f
Ļ
f edc ba
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 11: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/11.jpg)
Reconciling Phylogenies
2. Supertrees
ā¢ MRP supertrees [Ragan 1992]ā¢ RF supertrees [Bansal et al. 2010]ā¢ SPR supertrees [Whidden/Zeh/Beiko 2012]ā¢ . . .
ā
c ge f
Ļ
a bce
Ļ
a b c d
Ļ
c ge f
Ļ
bda
Comparing Phylogenies
Norbert Zeh
![Page 12: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/12.jpg)
Reconciling Phylogenies
2. Supertrees
ā¢ MRP supertrees [Ragan 1992]ā¢ RF supertrees [Bansal et al. 2010]ā¢ SPR supertrees [Whidden/Zeh/Beiko 2012]ā¢ . . .
ā
c ge f
Ļ
a bce
Ļ
a b c d
Ļ
c
Ļ
bda
Comparing Phylogenies
Norbert Zeh
![Page 13: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/13.jpg)
Reconciling Phylogenies
2. Supertrees
ā¢ MRP supertrees [Ragan 1992]ā¢ RF supertrees [Bansal et al. 2010]ā¢ SPR supertrees [Whidden/Zeh/Beiko 2012]ā¢ . . .
ā
c ge f
Ļ
a bce
Ļ
a b c d
Ļ
ce
Ļ
ba
Comparing Phylogenies
Norbert Zeh
![Page 14: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/14.jpg)
Reconciling Phylogenies
2. Supertrees
ā¢ MRP supertrees [Ragan 1992]ā¢ RF supertrees [Bansal et al. 2010]ā¢ SPR supertrees [Whidden/Zeh/Beiko 2012]ā¢ . . .
ā
c ge f
Ļ
a bce
Ļ
a b c d
Ļ
c ge f
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 15: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/15.jpg)
Reconciling Phylogenies
2. Supertrees
ā¢ MRP supertrees [Ragan 1992]ā¢ RF supertrees [Bansal et al. 2010]ā¢ SPR supertrees [Whidden/Zeh/Beiko 2012]ā¢ . . .
3. Phylogenetic networks
ā¢ DLT networks [Hallet/Lagergren 2011, Doyon et al. 2011]ā¢ Recombination networks [Gusfield et al. 2003]ā¢ Level-k hybridization networks [van Iersel/Kelk 2011]ā¢ MAAF of multiple trees [Chen/Wang 2012]
ā
c ge f
Ļ
a bce
Ļ
a b c d
Ļ
c ge f
Ļ
bda
Comparing Phylogenies
Norbert Zeh
![Page 16: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/16.jpg)
Computing SPR Distance
Comparing Phylogenies
Norbert Zeh
![Page 17: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/17.jpg)
Agreement Forests
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 18: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/18.jpg)
Agreement Forests
[Bordewich/Semple 2005]
a d e
Ļ
cb f
What is the largest forest we canobtain from T1 and T2 using edgedeletions and forced contractions?
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 19: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/19.jpg)
Agreement Forests
[Bordewich/Semple 2005]
a d e
Ļ
cb f
What is the largest forest we canobtain from T1 and T2 using edgedeletions and forced contractions?
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 20: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/20.jpg)
Agreement Forests
[Bordewich/Semple 2005]
a d e
Ļ
cb f
What is the largest forest we canobtain from T1 and T2 using edgedeletions and forced contractions?
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
What is the smallest hybridizationnetwork that displays both T1 and T2?
a b c d e f
Ļ
fedcb a
Ļ
afd bc e
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 21: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/21.jpg)
Agreement Forests
[Bordewich/Semple 2005]
a d e
Ļ
cb f
What is the largest forest we canobtain from T1 and T2 using edgedeletions and forced contractions?
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
What is the smallest hybridizationnetwork that displays both T1 and T2?
a b c d e f
Ļ
fedcb a
Ļ
afd bc e
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 22: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/22.jpg)
Agreement Forests
[Bordewich/Semple 2005]
a d e
Ļ
cb f
What is the largest forest we canobtain from T1 and T2 using edgedeletions and forced contractions?
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
What is the smallest hybridizationnetwork that displays both T1 and T2?
a b c d e f
Ļ
fedcb a
Ļ
afd bc e
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 23: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/23.jpg)
Agreement Forests
[Bordewich/Semple 2005]
a d e
Ļ
cb f
What is the largest forest we canobtain from T1 and T2 using edgedeletions and forced contractions?
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
What is the smallest hybridizationnetwork that displays both T1 and T2?
What is the largest acyclic agreementforest of T1 and T2?
a b c d e f
Ļ
fedcb a
Ļ
afd bc e
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 24: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/24.jpg)
Agreement Forests
[Bordewich/Semple 2005]
a d e
Ļ
cb f
What is the largest forest we canobtain from T1 and T2 using edgedeletions and forced contractions?
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
What is the smallest hybridizationnetwork that displays both T1 and T2?
What is the largest acyclic agreementforest of T1 and T2?
a b c d e f
Ļ
fedcb a
Ļ
afd bc e
Ļ
agreement forest
a b c d e f
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 25: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/25.jpg)
Agreement Forests
[Bordewich/Semple 2005]
a d e
Ļ
cb f
What is the largest forest we canobtain from T1 and T2 using edgedeletions and forced contractions?
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
What is the smallest hybridizationnetwork that displays both T1 and T2?
What is the largest acyclic agreementforest of T1 and T2?
a b c d e f
Ļ
fedcb a
Ļ
afd bc e
Ļ
agreement forest
a b c d e f
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 26: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/26.jpg)
Agreement Forests
[Bordewich/Semple 2005]
a d e
Ļ
cb f
[Bordewich/Semple 2007]
What is the largest forest we canobtain from T1 and T2 using edgedeletions and forced contractions?
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
What is the smallest hybridizationnetwork that displays both T1 and T2?
What is the largest acyclic agreementforest of T1 and T2?
a b c d e f
Ļ
fedcb a
Ļ
afd bc e
Ļ
agreement forest
a b c d e f
Ļ
acyclicagreement forest
a b c d e f
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 27: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/27.jpg)
Agreement Forests
[Bordewich/Semple 2005]
a d e
Ļ
cb f
[Bordewich/Semple 2007]
What is the largest forest we canobtain from T1 and T2 using edgedeletions and forced contractions?
How many SPR operations does ittake to turn T1 into T2?
a b c d e f
Ļ
f edcba
Ļ
a f d b c e
Ļ
What is the smallest hybridizationnetwork that displays both T1 and T2?
What is the largest acyclic agreementforest of T1 and T2?
a b c d e f
Ļ
fedcb a
Ļ
afd bc e
Ļ
agreement forest
a b c d e f
Ļ
acyclicagreement forest
a b c d e f
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 28: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/28.jpg)
Kernelization for Maximum Agreement Forest (SPR Distance)
Rule 1: Prune agreeing subtrees
Rule 2: Compress agreeing chains
Running time: O((56k)k + poly(n)) [Bordewich/Semple 2005]
c d e f c d e f
ā A A
ā
a1
a2
a3
a4
a5
a6
a1
a2
a3
a4
a5
a6
a1
a2
a3
a1
a2
a3
Comparing Phylogenies
Norbert Zeh
![Page 29: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/29.jpg)
Depth-Bounded Search for MAF [Whidden/Zeh 2009]
An MAF of T1 and T2 can be obtained by cutting edges in F2.
T1 T2
T1 F
F1
F2 F
F2
Comparing Phylogenies
Norbert Zeh
![Page 30: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/30.jpg)
Depth-Bounded Search for MAF [Whidden/Zeh 2009]
Case 1: A whole tree in F2 agrees with a subtree of T1
Comparing Phylogenies
Norbert Zeh
![Page 31: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/31.jpg)
Depth-Bounded Search for MAF [Whidden/Zeh 2009]
Case 1: A whole tree in F2 agrees with a subtree of T1
Case 2: Two agreeing subtrees are adjacent in T1 and F2
Comparing Phylogenies
Norbert Zeh
![Page 32: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/32.jpg)
Depth-Bounded Search for MAF [Whidden/Zeh 2009]
Case 3: Subtrees A and B are adjacent in T1 but not in F2
One branch per edge
a c
aā²
bā²
cā²
a
c
aā² bā²
cā²b
Comparing Phylogenies
Norbert Zeh
![Page 33: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/33.jpg)
Depth-Bounded Search for MAF [Whidden/Zeh 2009]
Case 3: Subtrees A and B are adjacent in T1 but not in F2
ā¢ Number of recursive calls = 3k
ā¢ Each costs O(n) time
Running time: O(3kn)
One branch per edge
a c
aā²
bā²
cā²
a
c
aā² bā²
cā²b
Comparing Phylogenies
Norbert Zeh
![Page 34: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/34.jpg)
Improved Branching Rules [Whidden/Beiko/Zeh 2010]
Case 3.1: a and b belong to different subtrees of F2
2 recursive callswith parameterkā 1
a c
aā² cā²
a c
aā² cā²
Comparing Phylogenies
Norbert Zeh
![Page 35: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/35.jpg)
Improved Branching Rules [Whidden/Beiko/Zeh 2010]
Case 3.1: a and b belong to different subtrees of F2
Case 3.2: One pendant subtree on path from a to b in F2
2 recursive callswith parameterkā 1
1 recursive callwith parameterkā 1
a c
aā² cā²
a c
aā² cā²
a c
aā² cā²
a
c
aā² bā²
cā²b
Comparing Phylogenies
Norbert Zeh
![Page 36: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/36.jpg)
Improved Branching Rules [Whidden/Beiko/Zeh 2010]
Case 3.3: mā„ 2 pendant subtrees on path from a to b in F2
3 recursive callswith parametersā¢ kā 1ā¢ kā 1ā¢ kā² ā¤ kā 2a c
aā² cā²
a
c
aā²
cā²
Comparing Phylogenies
Norbert Zeh
![Page 37: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/37.jpg)
Improved Branching Rules [Whidden/Beiko/Zeh 2010]
Case 3.3: mā„ 2 pendant subtrees on path from a to b in F2
3 recursive callswith parametersā¢ kā 1ā¢ kā 1ā¢ kā² ā¤ kā 2
Number of recursive invocations
I(k)ā¤ 2I(kā 1) + I(kā 2)ā¤ (1+p
2)k ā 2.41k
a c
aā² cā²
a
c
aā²
cā²
Comparing Phylogenies
Norbert Zeh
![Page 38: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/38.jpg)
Multifurcating Trees [Whidden/Beiko/Zeh 2012]
Problem 1: What is a meaningful definition of an agreement forest?
a b c
Ļ
a b c
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 39: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/39.jpg)
Multifurcating Trees [Whidden/Beiko/Zeh 2012]
Problem 1: What is a meaningful definition of an agreement forest?
a b c
Ļ
a b c
Ļ
An MAF of two multifurcating phylogenies T1 and T2 is the largest forest thatis an AF of two binary resolutions of T1 and T2.
Comparing Phylogenies
Norbert Zeh
![Page 40: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/40.jpg)
Multifurcating Trees [Whidden/Beiko/Zeh 2012]
Problem 1: What is a meaningful definition of an agreement forest?
a b c
Ļ
a b c
Ļ
An MAF of two multifurcating phylogenies T1 and T2 is the largest forest thatis an AF of two binary resolutions of T1 and T2.
Problem 2: Sibling pairs become sibling groups.
a1 a2 a3 a4 a1 a2 a3
a4
B1,1
B1,2
B3,1
B3,2
Comparing Phylogenies
Norbert Zeh
![Page 41: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/41.jpg)
Multifurcating Trees [Whidden/Beiko/Zeh 2012]
Itās FPT, alright . . .
ā¢ 5 cases depending on the structure of F2ā¢ The worst: I(k) = 1+ 2I(kā 1) + 3I(kā 2)
Comparing Phylogenies
Norbert Zeh
![Page 42: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/42.jpg)
Multifurcating Trees [Whidden/Beiko/Zeh 2012]
Itās FPT, alright . . .
ā¢ 5 cases depending on the structure of F2ā¢ The worst: I(k) = 1+ 2I(kā 1) + 3I(kā 2)
. . . but it aināt fast.
Comparing Phylogenies
Norbert Zeh
![Page 43: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/43.jpg)
Multifurcating Trees [Whidden/Beiko/Zeh 2012]
a1 a2 a3 a4 a1 a2 a3
a4
B1,1 B3,1B2,1
Itās FPT, alright . . .
ā¢ 5 cases depending on the structure of F2ā¢ The worst: I(k) = 1+ 2I(kā 1) + 3I(kā 2)
. . . but it aināt fast.
Comparing Phylogenies
Norbert Zeh
![Page 44: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/44.jpg)
Multifurcating Trees [Whidden/Beiko/Zeh 2012]
ā¢ Until the protected edges are eliminated, every recursive call becomes a2-way branch.
ā¢ Each such sequence of 2-way branches ends in a ā1-way branchā.
Running time: O(2.42kn)
a1 a2 a3 a4 a1 a2 a3
a4
B1,1 B3,1B2,1
Itās FPT, alright . . .
ā¢ 5 cases depending on the structure of F2ā¢ The worst: I(k) = 1+ 2I(kā 1) + 3I(kā 2)
. . . but it aināt fast.
Comparing Phylogenies
Norbert Zeh
![Page 45: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/45.jpg)
Binary Trees Even Faster [Whidden/Beiko/Zeh 2012]
ā¢ Edge protection idea from the multifurcating algorithm
ā¢ A couple of new cases
ā¢ A hairy analysis
Running time: O(2kn)
Comparing Phylogenies
Norbert Zeh
![Page 46: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/46.jpg)
Clustering [Linz/Semple 2009]
An MAF of the two input trees can be computed by computing MAFs of theclusters . . . with a twist.
a b c
Ļ1
a bc
Ļ1
e f g h
Ļ2
e fgh
Ļ2
d
Ļ
iĪ±1 Ī±2 d
Ļ
iĪ±1Ī±2
a b c d e f
Ļ
g h i a bcd e f
Ļ
gh i
Comparing Phylogenies
Norbert Zeh
![Page 47: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/47.jpg)
Branch and Bound
ā¢ For each invocation, compute 3-approximation kā² of number of edges leftto be cut.
ā¢ If kā² > 3k, abort.
Added cost per invocation: O(n) [Whidden/Zeh 2009]
kā² > 3k
not explored
Comparing Phylogenies
Norbert Zeh
![Page 48: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/48.jpg)
Experimental Results
Comparing Phylogenies
Norbert Zeh
![Page 49: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/49.jpg)
Hybridization [Whidden/Beiko/Zeh 2012]
Observation: While F2 is not an AF of T1 and T2, at least one of the branchesin each case of the MAF algorithm makes progress towards an MAAF.
Comparing Phylogenies
Norbert Zeh
![Page 50: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/50.jpg)
Hybridization [Whidden/Beiko/Zeh 2012]
Observation: While F2 is not an AF of T1 and T2, at least one of the branchesin each case of the MAF algorithm makes progress towards an MAAF.
Case 3.2ā²: One pendant subtree on path from a to b in F2
2 recursive callswith parameterkā 1a c
aā² cā²
a
c
aā² bā²
cā²b
Comparing Phylogenies
Norbert Zeh
![Page 51: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/51.jpg)
Hybridization [Whidden/Beiko/Zeh 2012]
Observation: While F2 is not an AF of T1 and T2, at least one of the branchesin each case of the MAF algorithm makes progress towards an MAAF.
Case 3.2ā²: One pendant subtree on path from a to b in F2
2 recursive callswith parameterkā 1a c
aā² cā²
a
c
aā² bā²
cā²b
Once an AF is obtained, cut edges to eliminate cycles.
Comparing Phylogenies
Norbert Zeh
![Page 52: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/52.jpg)
Hybridization [Whidden/Beiko/Zeh 2012]
Cycle graph
Ļ
h i
f g
d e
a b c
a b c d e f g h i
Ļ
a b c d e f gh i
Ļ
Comparing Phylogenies
Norbert Zeh
![Page 53: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/53.jpg)
Hybridization [Whidden/Beiko/Zeh 2012]
Breaking cycles
ā¢ 2k edges between componentsā¢ For each, may need to eliminate the path to the root of the parent
component
ā O(22k Ā· 2.42kn) = O(9.68kn) time
Comparing Phylogenies
Norbert Zeh
![Page 54: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/54.jpg)
Hybridization [Whidden/Beiko/Zeh 2012]
Breaking cycles
ā¢ 2k edges between componentsā¢ For each, may need to eliminate the path to the root of the parent
component
ā O(22k Ā· 2.42kn) = O(9.68kn) time
Reducing the number of candidate edges
ā¢ Can get away with considering only k of the 2k edges
ā O(2k Ā· 2.42kn) = O(4.84kn) time
Comparing Phylogenies
Norbert Zeh
![Page 55: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/55.jpg)
Hybridization [Whidden/Beiko/Zeh 2012]
A better analysis
ā¢ If the AF has kā² ā k edges, the refinement step considersļæ½ k
kākā²ļæ½
ļæ½ 2k
choicesā¢ If the AF has kā² ā 0 edges, the refinement step considers at most 2kā² ļæ½ 2k
choicesā¢ If the AF has kā² ā k/2 edges, the refinement step considers
ļæ½ kkākā²ļæ½
ā 2k
choices, but this situation can arise only 2.42kā² ļæ½ 2.42k times
ā O(3.18kn) time
Comparing Phylogenies
Norbert Zeh
![Page 56: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/56.jpg)
Application: SPR Supertrees
Comparing Phylogenies
Norbert Zeh
![Page 57: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/57.jpg)
SPR Supertrees [Whidden/Zeh/Beiko 2012]
Open problem: Computational complexity of computing an optimal SPRsupertree.
Comparing Phylogenies
Norbert Zeh
![Page 58: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/58.jpg)
SPR Supertrees [Whidden/Zeh/Beiko 2012]
Open problem: Computational complexity of computing an optimal SPRsupertree.
Heuristic
ā¢ Build up initial supertreeā¢ Iterative improvement using SPR operations
Comparing Phylogenies
Norbert Zeh
![Page 59: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/59.jpg)
SPR Supertrees [Whidden/Zeh/Beiko 2012]
Open problem: Computational complexity of computing an optimal SPRsupertree.
Heuristic
ā¢ Build up initial supertreeā¢ Iterative improvement using SPR operations
Initial tree construction
ā¢ Start with 4-leaf tree consistent with one of the gene treesā¢ Attach one leaf at a timeā¢ For each leaf, choose the location that minimizes SPR distance
Comparing Phylogenies
Norbert Zeh
![Page 60: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/60.jpg)
SPR Supertrees [Whidden/Zeh/Beiko 2012]
Open problem: Computational complexity of computing an optimal SPRsupertree.
Heuristic
ā¢ Build up initial supertreeā¢ Iterative improvement using SPR operations
Initial tree construction
ā¢ Start with 4-leaf tree consistent with one of the gene treesā¢ Attach one leaf at a timeā¢ For each leaf, choose the location that minimizes SPR distance
Iterative improvement
ā¢ Try all O(n2) SPR operations on current supertree and choose the one thatminimizes the SPR distance from gene trees
Comparing Phylogenies
Norbert Zeh
![Page 61: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/61.jpg)
SPR Supertrees [Whidden/Zeh/Beiko 2012]
Limit number of SPR moves to consider
ā¢ Consider only SPR operations across r = O(1) edgesā O(n) moves
ā O(tn) exact SPR computations
ā¢ Rank moves based on approximate SPR distance of resulting tree to genetrees
ā¢ Try moves in this order and choose the first one that gives an improvement
ā O(tn2) approximate SPR computations + O(t) exact SPR computations
Comparing Phylogenies
Norbert Zeh
![Page 62: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/62.jpg)
SPR Supertrees [Whidden/Zeh/Beiko 2012]
Limit number of SPR moves to consider
ā¢ Consider only SPR operations across r = O(1) edgesā O(n) moves
ā O(tn) exact SPR computations
ā¢ Rank moves based on approximate SPR distance of resulting tree to genetrees
ā¢ Try moves in this order and choose the first one that gives an improvement
ā O(tn2) approximate SPR computations + O(t) exact SPR computations
MAF-driven improvements
ā¢ In each iteration, every gene tree initiates one SPR move on supertree thatreduces its distance by one
ā¢ Choose this move using the MAF of gene tree and supertree
ā t exact SPR computations
Comparing Phylogenies
Norbert Zeh
![Page 63: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/63.jpg)
Conclusions
Comparing Phylogenies
Norbert Zeh
![Page 64: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/64.jpg)
Ongoing and Future Work
Faster supertree search
ā¢ FPT approximation to handle really large trees
Comparing Phylogenies
Norbert Zeh
![Page 65: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/65.jpg)
Ongoing and Future Work
Faster supertree search
ā¢ FPT approximation to handle really large trees
Faster M(A)AF algorithms
ā¢ Substantially break the 2k barrier to handle trees with 1,000s of leaves
Comparing Phylogenies
Norbert Zeh
![Page 66: Comparing Phylogenies](https://reader031.fdocuments.in/reader031/viewer/2022012512/618a8caa8ff1db57044b9180/html5/thumbnails/66.jpg)
Ongoing and Future Work
Faster supertree search
ā¢ FPT approximation to handle really large trees
Faster M(A)AF algorithms
ā¢ Substantially break the 2k barrier to handle trees with 1,000s of leaves
Compute all M(A)AFs [Abrecht et al. 2012]
ā¢ Provide more biological insight
Comparing Phylogenies
Norbert Zeh