## Performance of supertree methods on various dataset decompositions (2004)

### Cached

### Download Links

Venue: | Phylogenetic Supertrees: Combining Information to Reveal the Tree of Life, volume 3 of Computational Biology |

Citations: | 17 - 9 self |

### BibTeX

@INPROCEEDINGS{Roshan04performanceof,

author = {Usman Roshan and Bernard M. E and Moret Tiffani and L. Williams and Tandy Warnow},

title = {Performance of supertree methods on various dataset decompositions},

booktitle = {Phylogenetic Supertrees: Combining Information to Reveal the Tree of Life, volume 3 of Computational Biology},

year = {2004},

pages = {301--328},

publisher = {Kluwer Academic Publishers}

}

### OpenURL

### Abstract

Many large-scale phylogenetic reconstruction methods attempt to solve hard optimization problems (such as Maximum Parsimony (MP) and Maximum Likelihood (ML)), but they are severely limited by the number of taxa that they can handle in a reasonable time frame. A standard heuristic approach to this problem is the divide-and-conquer strategy: decompose the dataset into smaller subsets, solve the subsets (i.e., use MP or ML on each subset to obtain trees), then combine the solutions to the subsets into a solution to the original dataset. This last step, combining given trees into a single tree, is known as supertree construction in computational phylogenetics. The traditional application of supertree methods is to combine existing, published phylogenies into a single phylogeny. Here, we study supertree construction in the context of divide-and-conquer methods for large-scale tree reconstruction. We study several divide-and-conquer approaches and experimentally demonstrate their advantage over Matrix Representation Parsimony (MRP), a traditional supertree technique, and over global heuristics such as the parsimony ratchet. On the ten large biological datasets under investigation, our study shows that the techniques used for dividing the dataset into subproblems as well as those used for merging them into a single solution strongly influence the quality of the supertree construction. In most cases, our merging technique—the Strict Consensus Merger (SCM)—outperforms MRP with respect to MP scores and running time. Divideand-conquer techniques are also a highly competitive alternative to global heuristics such as the parsimony ratchet, especially on the more challenging datasets. 1

### Citations

2293 |
The neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecular Biology and Evolution
- Saitou, Nei
- 1987
(Show Context)
Citation Context ... usually return a totally unresolved tree because too many quartets will be in error. In an earlier study [33], we compared various quartet-based methods and the fast and simple neighbor-joining (NJ) =-=[31]-=- method on simulated data. Quartet Puzzling, which merges quartet trees using a greedy heuristic, clearly dominated the other quartet-based methods, but was much slower and clearly less accurate than ... |

1482 |
PAUP *: phylogenetic analysis using parsimony (*and other methods), v. 4.0
- SWOFFORD
- 1997
(Show Context)
Citation Context ...s, what is the best method to assemble them into a single supertree? • How do divide-and-conquermethods fare when compared to “global” approaches, such as the heuristic searches for MP in PAUP*4.0b10 =-=[36]-=-? To investigate the first two questions, we compare methods that differ explicitly in how they decompose the dataset and how they merge subtrees into a supertree. The family of Disk-Covering Methods ... |

1132 |
Algorithmic graph theory and perfect graphs
- Golumbic
- 1980
(Show Context)
Citation Context ...j ≤ q. The graph is then minimally triangulated, i.e., edges are added to the graph until every cycle of length at least four has a chord (an edge connecting two nonconsecutive vertices on the cycle) =-=[9, 14]-=-, while attempting to minimize the weight of the largest edge added. Obtaining an optimal triangulation of a graph is in general NP-hard [7], but threshold graphs are usually triangulated or close to ... |

1098 |
ModelTest: testing the model of DNA substitution
- POSADA, CRANDALL
- 1998
(Show Context)
Citation Context ...the data particularly well, since it affects only the choice of edges in the threshold graph; nevertheless, it is possible that a better distance correction (such as could be obtained using MODELTEST =-=[27]-=-) would yield better results. In other words, our results with the DCM-based approaches should be regarded as pessimistic—they establish a lower bound, but are subject to further improvement. 3.2 Impl... |

956 |
Evolutionary trees from DNA sequences: a maximum likelihood approach
- Felsenstein
- 1981
(Show Context)
Citation Context ...of supertree methods in a different context: as part of divide-and-conquer methods that can be used to solve difficult optimization problems such as Maximum Parsimony (MP) and Maximum Likelihood (ML) =-=[11, 12, 16, 34]-=-. These two problems are sufficiently hard that a biologically acceptable phylogenetic analysis can take a very long time (months, perhaps) to derive. Our conjecture, which we study in this paper, is ... |

745 |
A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide sequences
- Kimura
- 1980
(Show Context)
Citation Context ...ph as the first step of its computation—and also that the distance matrix does not play any role in the phylogenetic reconstruction beyond this first step. We used the Kimura 2-parameter (plus Gamma) =-=[20, 39]-=- distance correction formula to compute a distance matrix for each dataset, using parameter values of κ = 2 and α = 1 (the “default” values). We do not need the model to fit the data particularly well... |

202 |
Quartet puzzling: a quartet maximum likelihood method for reconstructing tree topologies
- Strimmer, Haeseler
- 1996
(Show Context)
Citation Context ...), since these are the datasets where a divide-andconquer methodology will have the largest impact. Some of the earliest divide-and-conquer methods are quartet-based methods, such as Quartet Puzzling =-=[35]-=-, Short Quartet methods [10], and Quartet Cleaning [2]. Quartet methods are at one extreme of divide-and-conquer methods, since they decompose the datasets into the smallest possible subsets for which... |

141 |
Maximum-likelihood estimation of phylogeny from dna sequences when substitution rates differ over sites
- Yang
- 1993
(Show Context)
Citation Context ...ph as the first step of its computation—and also that the distance matrix does not play any role in the phylogenetic reconstruction beyond this first step. We used the Kimura 2-parameter (plus Gamma) =-=[20, 39]-=- distance correction formula to compute a distance matrix for each dataset, using parameter values of κ = 2 and α = 1 (the “default” values). We do not need the model to fit the data particularly well... |

130 |
The Steiner problem in phylogeny is NPComplete
- Foulds, Graham
- 1982
(Show Context)
Citation Context ...of supertree methods in a different context: as part of divide-and-conquer methods that can be used to solve difficult optimization problems such as Maximum Parsimony (MP) and Maximum Likelihood (ML) =-=[11, 12, 16, 34]-=-. These two problems are sufficiently hard that a biologically acceptable phylogenetic analysis can take a very long time (months, perhaps) to derive. Our conjecture, which we study in this paper, is ... |

110 | Two strikes against perfect phylogeny
- Bodlaender, Fellows, et al.
(Show Context)
Citation Context ...ge connecting two nonconsecutive vertices on the cycle) [9, 14], while attempting to minimize the weight of the largest edge added. Obtaining an optimal triangulation of a graph is in general NP-hard =-=[7]-=-, but threshold graphs are usually triangulated or close to it [17]—and our experience shows that even the simple greedy heuristic we use produces triangulations that do not have very long edges. We t... |

99 | A few logs suffice to build (almost) all trees (I), Submitted, Random Structures and Algorithms
- Erdos, Steel, et al.
(Show Context)
Citation Context ...ets where a divide-andconquer methodology will have the largest impact. Some of the earliest divide-and-conquer methods are quartet-based methods, such as Quartet Puzzling [35], Short Quartet methods =-=[10]-=-, and Quartet Cleaning [2]. Quartet methods are at one extreme of divide-and-conquer methods, since they decompose the datasets into the smallest possible subsets for which nontrivial trees exist—subs... |

93 |
BR: Combining trees as a way of combining data sets for phylogenetic inference, and the desirability of combining gene trees. Taxon
- Baum
- 1992
(Show Context)
Citation Context ...mbine existing, published phylogenies, on which the community agrees, into a tree leaf-labeled by the entire set of species. The most popular supertree method is Matrix Representation Parsimony (MRP) =-=[1, 29]-=-, which ∗ Department of Computer Science, U. of Texas at Austin, usman,tandy@cs.utexas.edu † Department of Computer Science, U. of New Mexico, moret,tlw@cs.unm.edu 1shas been used in a number of phylo... |

85 |
The winds of (evolutionary) change: Breathing new life into microbiology
- Olsen, Woese, et al.
- 1994
(Show Context)
Citation Context ...s is not profitable. Other divide-and-conquer methods include Compartmentalization [23], which is not fully described and so cannot be implemented, and a strategy used to analyze a biological dataset =-=[26]-=-, where again the decomposition and merging steps are not well enough describe to enable one to implement and test the strategy. 2 Divide-and-Conquer Reconstruction Methods Each of the divide-and-conq... |

84 |
A characterization of rigid circuit graphs
- Buneman
- 1974
(Show Context)
Citation Context ...j ≤ q. The graph is then minimally triangulated, i.e., edges are added to the graph until every cycle of length at least four has a chord (an edge connecting two nonconsecutive vertices on the cycle) =-=[9, 14]-=-, while attempting to minimize the weight of the largest edge added. Obtaining an optimal triangulation of a graph is in general NP-hard [7], but threshold graphs are usually triangulated or close to ... |

78 | Disk-covering, a fast-converging method for phylogenetic tree reconstruction
- Huson, Nettles, et al.
- 1999
(Show Context)
Citation Context ...vestigate the first two questions, we compare methods that differ explicitly in how they decompose the dataset and how they merge subtrees into a supertree. The family of Disk-Covering Methods (DCMs) =-=[17, 18]-=- plays a major role in addressing these questions. DCMs are metamethods for phylogenetic reconstruction that use a sophisticated dataset decomposition technique (based on cliques in a triangulated gra... |

65 |
The Parsimony Ratchet, a new method for rapid parsimony analysis
- Nixon
- 1999
(Show Context)
Citation Context ...=1 chuckscore=1 swap=tbr; set maxtrees=1000 increase=no; filter best=yes; hsearch start=current swap=tbr hold=1 nchuck=1000 timelimit=3600; contree all/ strict=yes; We also used the parsimony ratchet =-=[25]-=-, in a PAUP*4.0b10 implementation written by Bininda-Emonds [3]. The ratchet is a simple and effective heuristic for general optimizing search and works iteratively as follows: 1. Run Fast HS for MP. ... |

61 |
Optimal algorithms for comparing trees with labelled leaves
- Day
- 1985
(Show Context)
Citation Context ...for DCM1, the order of merging is irrelevant. We call this the Strict Consensus Subtree Merger because the definition of the tree that will be induced on X after the merger is the "strict consens=-=us" (Day, 1995-=-) of the two initially induced subtrees. This is defined to be the maximally resolved tree that is a common contraction of the two subtrees. We will call this subtree on X the backbone. Merging the tw... |

60 |
Phylogenetic inference based on matrix representation of trees
- Ragan
- 1992
(Show Context)
Citation Context ...mbine existing, published phylogenies, on which the community agrees, into a tree leaf-labeled by the entire set of species. The most popular supertree method is Matrix Representation Parsimony (MRP) =-=[1, 29]-=-, which ∗ Department of Computer Science, U. of Texas at Austin, usman,tandy@cs.utexas.edu † Department of Computer Science, U. of New Mexico, moret,tlw@cs.unm.edu 1shas been used in a number of phylo... |

56 |
A composite estimate of primate phylogeny
- Purvis
- 1995
(Show Context)
Citation Context ...ment of Computer Science, U. of Texas at Austin, usman,tandy@cs.utexas.edu † Department of Computer Science, U. of New Mexico, moret,tlw@cs.unm.edu 1shas been used in a number of phylogenetic studies =-=[5, 6, 19, 21, 28]-=-. Bininda-Emonds and colleagues [4, 6] have evaluated the behavior of several variants of MRP on small simulated datasets with respect to topological accuracy. We study the application of supertree me... |

46 |
Modified mincut supertrees
- Page
(Show Context)
Citation Context ...accuracy of reconstructions made with several variants of MRP in simulations (Bininda-Emonds, 2002; Bininda-Emonds & Sanderson, 2001), while Page compared it to a different method (based on mincuts) (=-=Page, 2002-=-). Beyond these three studies, little has been done to investigate the relative performance of supertree methods in an experimental setting. In this study, we use two different supertree methods, MRP ... |

42 | Quartet cleaning: improved algorithms and simulations
- Berry, Jiang, et al.
- 1999
(Show Context)
Citation Context ...er methodology will have the largest impact. Some of the earliest divide-and-conquer methods are quartet-based methods, such as Quartet Puzzling [35], Short Quartet methods [10], and Quartet Cleaning =-=[2]-=-. Quartet methods are at one extreme of divide-and-conquer methods, since they decompose the datasets into the smallest possible subsets for which nontrivial trees exist—subsets of just four taxa each... |

42 |
Analyzing large data sets in reasonable times: Solutions for composite optima
- Goloboff
- 1999
(Show Context)
Citation Context ...acaeae obtained from the Gutell Lab at the Institute for Cellular and Molecular Biology, The University of Texas at Austin; max p-distance = 0.524. 2. A set of 439 aligned DNA sequences (2,461 sites) =-=[13]-=-; max p-distance = 0.649. 3. A set of 476 aligned DNA sequences (1,008 sites) [13]; max p-distance = 0.445. 4. A set of 500 aligned rbcL DNA sequences (1,398 sites) [30]; max p-distance = 0.184. 5. A ... |

42 | The RDP-II (Ribosomal Database Project
- Maidak, Cole, et al.
(Show Context)
Citation Context ...istance = 0.445. 4. A set of 500 aligned rbcL DNA sequences (1,398 sites) [30]; max p-distance = 0.184. 5. A set of 556 aligned 16S rRNA sequences (2,402 sites) for the Spirochaetes class of Bacteria =-=[22]-=-; max p-distance = 0.31. 6. A set of 567 “three gene: rbcL, atpB, and 18s” aligned DNA sequences (4,592 sites) [32]; max p-distance = 0.15. 7. A set of 590 aligned small subunit Archaea rRNA sequences... |

36 | Towards a discipline of experimental algorithmics
- Moret
- 2002
(Show Context)
Citation Context ...other). On the other hand, biological datasets offer data with all the biases and peculiarities that are so hard to produce in simulations. Thus the main use of “real-world” data is in spot-checking (=-=Moret, 2002-=-)—confirming that predictions made on the basis of simulation results hold for biological data or pinpointing problems with models when the datasets yield incompatible results. In this study, we choos... |

35 |
Consensus supertrees: the synthesis of rooted trees containing overlapping sets of labeled leaves
- Gordon
- 1986
(Show Context)
Citation Context ...edge, we identify all pieces of T that should attach to that edge and attach them directly to ve. The Strict Consensus Merger of two trees is very similar to Gordon’s Strict Consensus Supertree (SCS) =-=[15]-=-: the only difference is how collisions are handled: in case of collisions, the SCS tree is a strict contraction of the SCM tree, because it contracts additional edges located within pieces involved i... |

33 | Performance study of phylogenetic methods: (unweighted) quartet methods and neighbor-joining
- John, Moret, et al.
(Show Context)
Citation Context ...econstruction techniques we study here)—MRP is too expensive given the tiny trees and SCM will usually return a totally unresolved tree because too many quartets will be in error. In an earlier study =-=[33]-=-, we compared various quartet-based methods and the fast and simple neighbor-joining (NJ) [31] method on simulated data. Quartet Puzzling, which merges quartet trees using a greedy heuristic, clearly ... |

29 |
The European database on small subunit ribosomal RNA
- Wuyts, Peer, et al.
- 2002
(Show Context)
Citation Context ...e = 0.31. 6. A set of 567 “three gene: rbcL, atpB, and 18s” aligned DNA sequences (4,592 sites) [32]; max p-distance = 0.15. 7. A set of 590 aligned small subunit Archaea rRNA sequences (1,962 sites) =-=[38]-=-; max p-distance = 0.382. 8. A set of 695 aligned 16S rRNA sequences (2,550 sites) for the Cyanobacteria class of Bacteria [22]; max p-distance = 0.219. 9. A set of 816 aligned small subunit Mitochond... |

29 | Scaling up accurate phylogenetic reconstruction from gene-order data
- Tang, Moret
- 2003
(Show Context)
Citation Context ...ods, of which Quartet Puzzling (Strimmer and von Haeseler, 1996) is the most popular, and the family of Disk-Covering Methods (DCMs) (Huson et al., 1999a,b; Nakhleh et al., 2001; Warnow et al., 2001; =-=Tang and Moret, 2003-=-). In each of these methods, a supertree method (Step 3) is used to combine subtrees into a tree on the entire dataset. Supertree methods are thus an integral aspect of a divide-and-conquer strategy, ... |

24 |
Designing fast converging phylogenetic methods
- John, Warnow
(Show Context)
Citation Context ...nal values and run Fast HS starting with the tree from the previous search. 4. Repeat steps two and three as desired. 2.2 Data Decomposition 2.2.1 DCM-based decomposition Disk-Covering Methods (DCMs) =-=[17, 18, 24, 37]-=- are meta-methods for phylogenetic reconstruction: they operate in conjunction with a “base method” such as an MP heuristic or NJ. DCMs decompose the input set into smaller overlapping sets on which s... |

24 |
The maximum likelihood point for phylogenetic tree is not unique
- Steel
- 1994
(Show Context)
Citation Context ...of supertree methods in a different context: as part of divide-and-conquer methods that can be used to solve difficult optimization problems such as Maximum Parsimony (MP) and Maximum Likelihood (ML) =-=[11, 12, 16, 34]-=-. These two problems are sufficiently hard that a biologically acceptable phylogenetic analysis can take a very long time (months, perhaps) to derive. Our conjecture, which we study in this paper, is ... |

23 |
Angiosperm phylogeny inferred from 18S rDNA, rbcL, and atpB sequences
- Soltis, Soltis, et al.
- 2000
(Show Context)
Citation Context ... 556 aligned 16S rRNA sequences (2,402 sites) for the Spirochaetes class of Bacteria [22]; max p-distance = 0.31. 6. A set of 567 “three gene: rbcL, atpB, and 18s” aligned DNA sequences (4,592 sites) =-=[32]-=-; max p-distance = 0.15. 7. A set of 590 aligned small subunit Archaea rRNA sequences (1,962 sites) [38]; max p-distance = 0.382. 8. A set of 695 aligned 16S rRNA sequences (2,550 sites) for the Cyano... |

22 |
Solving large scale phylogenetic problems using DCM2
- Huson, Vawter, et al.
- 1999
(Show Context)
Citation Context ...a of Figures 8 and 9) show that DCM2+SCM(d4) outperforms DCM2+SCM(d0) on most of the ten datasets. This improvement in MP scores as we increase the threshold value is consistent with previous studies =-=[18]-=- and our recent simulation studies (not shown). Note that, as we increase the threshold, the number of subproblems decreases, so that, although the maximum subproblem size increases, the total time de... |

17 |
Assessment of the accuracy of matrix representation with parsimony analysis supertree construction
- Bininda-Emonds, Sanderson
- 2001
(Show Context)
Citation Context ...urvis, 1995; ¡ Department of Computer Science, U. of Texas at Austin, usman,tandy@cs.utexas.edu † Department of Computer Science, U. of New Mexico, moret,tlw@cs.unm.edu 1sBininda-Emonds et al., 1999; =-=Bininda-Emonds and Sanderson, 2001-=-; Liu et al., 2001; Jones et al., 2002). Bininda-Emonds and colleagues (Bininda-Emonds and Sanderson, 2001; Bininda-Emonds, 2003a) have evaluated the behavior of several variants of MRP on small simul... |

15 |
Analyzing large data sets: rbcL 500 revisited. Syst. Biol
- RICE, DONOGHUE, et al.
- 1997
(Show Context)
Citation Context ...ed DNA sequences (2,461 sites) [13]; max p-distance = 0.649. 3. A set of 476 aligned DNA sequences (1,008 sites) [13]; max p-distance = 0.445. 4. A set of 500 aligned rbcL DNA sequences (1,398 sites) =-=[30]-=-; max p-distance = 0.184. 5. A set of 556 aligned 16S rRNA sequences (2,402 sites) for the Spirochaetes class of Bacteria [22]; max p-distance = 0.31. 6. A set of 567 “three gene: rbcL, atpB, and 18s”... |

14 | Better methods for solving parsimony and compatibility
- Bonet, Steel, et al.
- 1998
(Show Context)
Citation Context ...dentify the best-supported edges. Resolving the remaining polytomies (by adding edges) so as to minimize the parsimony score of the resulting tree is the NP-hard Optimal Tree Refinement (OTR) problem =-=[8]-=-. To “solve” it, we pass unresolved trees as constraint trees to PAUP*4.0b10 and use a fast MP heuristic search for a resolved tree, using the following command: constraints c1 (monophyly) = <the unre... |

14 |
Absolute convergence: true trees from short sequences
- John
(Show Context)
Citation Context ...nal values and run Fast HS starting with the tree from the previous search. 4. Repeat steps two and three as desired. 2.2 Data Decomposition 2.2.1 DCM-based decomposition Disk-Covering Methods (DCMs) =-=[17, 18, 24, 37]-=- are meta-methods for phylogenetic reconstruction: they operate in conjunction with a “base method” such as an MP heuristic or NJ. DCMs decompose the input set into smaller overlapping sets on which s... |

13 |
Molecular and morphological supertrees for eutherian (placental) mammals
- –G, Miyamoto, et al.
- 2001
(Show Context)
Citation Context ...ment of Computer Science, U. of Texas at Austin, usman,tandy@cs.utexas.edu † Department of Computer Science, U. of New Mexico, moret,tlw@cs.unm.edu 1shas been used in a number of phylogenetic studies =-=[5, 6, 19, 21, 28]-=-. Bininda-Emonds and colleagues [4, 6] have evaluated the behavior of several variants of MRP on small simulated datasets with respect to topological accuracy. We study the application of supertree me... |

13 |
Sequence length requirements for phylogenetic methods
- Moret, Roshan, et al.
- 2002
(Show Context)
Citation Context ...s for two reasons: (i) we need them for spot-checking our conjectures, which we derived from largescale simulation studies that we have already conducted (St. John et al., 2001; Nakhleh et al., 2002; =-=Moret et al., 2002-=-); and (ii) no comparable experimental study has been conducted—existing reports are limited to small biological datasets or to just one larger dataset. Since our conjectures may hold in significant p... |

9 | The accuracy of fast phylogenetic methods for large datasets
- Nakhleh, Moret, et al.
- 2002
(Show Context)
Citation Context ...ose biological datasets for two reasons: (i) we need them for spot-checking our conjectures, which we derived from largescale simulation studies that we have already conducted (St. John et al., 2001; =-=Nakhleh et al., 2002-=-; Moret et al., 2002); and (ii) no comparable experimental study has been conducted—existing reports are limited to small biological datasets or to just one larger dataset. Since our conjectures may h... |

8 |
A phylogenetic supertree of the bats (Mammalia: Chiroptera
- Jones, Purvis, et al.
- 2002
(Show Context)
Citation Context ...ment of Computer Science, U. of Texas at Austin, usman,tandy@cs.utexas.edu † Department of Computer Science, U. of New Mexico, moret,tlw@cs.unm.edu 1shas been used in a number of phylogenetic studies =-=[5, 6, 19, 21, 28]-=-. Bininda-Emonds and colleagues [4, 6] have evaluated the behavior of several variants of MRP on small simulated datasets with respect to topological accuracy. We study the application of supertree me... |

7 | Absolute convergence: true trees from short sequences - Warnow, Moret, et al. - 2001 |

6 |
MJ: Assessment of the accuracy of matrix representation with parsimony supertree construction
- ORP, Sanderson
(Show Context)
Citation Context |

6 |
Cladistic analysis of molecular and morphological data
- Mishler
- 1994
(Show Context)
Citation Context ...ds, but was much slower and clearly less accurate than NJ. These results suggest that decompositions into tiny subsets is not profitable. Other divide-and-conquer methods include Compartmentalization =-=[23]-=-, which is not fully described and so cannot be implemented, and a strategy used to analyze a biological dataset [26], where again the decomposition and merging steps are not well enough describe to e... |

5 |
MRP Supertree Construction in the Consensus Setting
- Bininda-Emonds
- 2003
(Show Context)
Citation Context ...sman,tandy@cs.utexas.edu † Department of Computer Science, U. of New Mexico, moret,tlw@cs.unm.edu 1shas been used in a number of phylogenetic studies [5, 6, 19, 21, 28]. Bininda-Emonds and colleagues =-=[4, 6]-=- have evaluated the behavior of several variants of MRP on small simulated datasets with respect to topological accuracy. We study the application of supertree methods in a different context: as part ... |

3 |
Ratchet implementation in PAUP*4.0b10
- Bininda-Emonds
- 2003
(Show Context)
Citation Context ...best=yes; hsearch start=current swap=tbr hold=1 nchuck=1000 timelimit=3600; contree all/ strict=yes; We also used the parsimony ratchet [25], in a PAUP*4.0b10 implementation written by Bininda-Emonds =-=[3]-=-. The ratchet is a simple and effective heuristic for general optimizing search and works iteratively as follows: 1. Run Fast HS for MP. 2. Randomly select 25% of the sites, set their weights to 2 and... |

3 |
Ratchet implementation in PAUP*4.0b10. available from http://www.tierzucht.tum.de:8080/WWW/Homepages/Bininda-Emonds
- Bininda-Emonds
- 2003
(Show Context)
Citation Context ...xico, moret,tlw@cs.unm.edu 1sBininda-Emonds et al., 1999; Bininda-Emonds and Sanderson, 2001; Liu et al., 2001; Jones et al., 2002). Bininda-Emonds and colleagues (Bininda-Emonds and Sanderson, 2001; =-=Bininda-Emonds, 2003-=-a) have evaluated the behavior of several variants of MRP on small simulated datasets with respect to topological accuracy. We study the application of supertree methods in a different context: as par... |

2 |
Building trees by combining phylogenetic information: a complete phylogeny of the extant Carnivora (Mammalia
- Bininda-Emonds, Gittleman, et al.
- 1999
(Show Context)
Citation Context |

2 |
Molecular Systematics. Sinauer Pub
- Hillis, Moritz, et al.
- 1996
(Show Context)
Citation Context |

2 |
Phylogenetics of sees plants: An analysis of nucleotide sequences from the plastid gene rbcL
- Chase, Soltis, et al.
- 1993
(Show Context)
Citation Context ...ically acceptable phylogenetic analysis can take years to complete on only a few hundred taxa. One popular example is the 500 taxon rbcL dataset, which has been analyzed for years without resolution (=-=Chase et al., 199-=-3; Rice et al., 1997). The computational requirements of large-scale phylogenetics have motivated systematists to develop alternative techniques for reconstructing evolutionary Department of Computer... |

2 | The neighbor-joining method: A new method for reconstructing phylogenetic trees - Sautou, Nei - 1987 |