G. J. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. Fastdnaml: A tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Computer Applications in the Biosciences, 10:41--48, 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

...., iii) branch lengths scaling factor (1, 4, 10) and iv) transversion transition ratio (2, 4) e ectively generating 24 sequence sets per tree topology. Using these simulated sequences as input, we ran both treepuzzle [16] and HC where the witness quartet set was inferred using fastDNAml [10]. With both fastDNAml and treepuzzle, we ran the software on default settings, such that the assumptions on the model of evolution is incorrect with respect to the generated sequences. We then proceed to test the accuracy and robustness of the inference methods under such a circumstance when our ....

Olsen, G., Matsuda, H., Hagstrom, R., and Overbeek, R. Fastdnaml: a tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Current Applications in Biosciences 10 (1994), 41-48.

...., iii) branch lengths scaling factor (1, 4, 10) and iv) transversion transition ratio (2, 4) e ectively generating 24 sequence sets per tree topology. Using these simulated sequences as input, we ran both treepuzzle [16] and HC where the witness quartet set was inferred using fastDNAml [10]. With both fastDNAml and treepuzzle, we ran the software on default settings, such that the assumptions on the model of evolution is incorrect with respect to the generated sequences. We then proceed to test the accuracy and robustness of the inference methods under such a circumstance when our ....

Olsen, G., Matsuda, H., Hagstrom, R., and Overbeek, R. Fastdnaml: a tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Current Applications in Biosciences 10 (1994), 41-48.

....from T in polynomial time [7] Quartet based methods operate in two phases. In the first phase, they construct a set Q of quartets on the different sets of four taxa. A popular approach is to use maximum likelihood (ML) a computationally intensive but statistically sophisticated method [8, 21]. In the second phase, they combine these quartets into a tree on the entire set of taxa. In practice, not all quartets are accurately inferred, so it is necessary for quartet methods to handle incorrect quartets. Most optimization problems related to tree reconstruction from quartets are NP hard. ....

Olsen, G. J., Matsuda, H., Hagstrom, R., and Overbeek, R., "fastDNAml: A tool for construction of phylogenetic trees of DNA sequences using maximum likelihood," Comput. Appl. Biosci. 10 (1994): 41-48.

....of input size. 80 To test the accuracy of HC # , we designed an experiment using real dataset by taking a random sample of subtrees of various sizes of the RDP database. Next we ran our implementation of HC # against paupNJ, paupPars and paupPuzPars. We also used fastDNAml implementation [37] of ML on the smaller datasets, which became computationally intractable for larger datasets with greater than 30 leaves. Tables (4) and (5) lists the accuracy of the methods on 30 taxon and 45 taxon trees. The witness quartet set Q that HC # used was generated with fastDNAml. Tree Edges ....

Olsen, G., Matsuda, H., Hagstrom, R., and Overbeek, R. Fastdnaml: a tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Current Applications in Biosciences 10 (1994), 41-- 48.

....be classi ed into two categories: constructive approach and non constructive one. The constructive approach, which is more popular, builds an evolutionary tree by adding one taxon at a time, starting at an empty tree, with some heuristic information. DNAml [9] and its improved variant, fastDNAml [26], are the representative of them. Although fastDNAml is one of the most widely used method in the phylogenetics literature, its performance is limited due to its incremental nature in constructing trees. In particular, the performance of fastDNAml is notably a ected by the order of taxon ....

....observed sequence is de ned by: ln L = k i=1 ln L i where k is the number of sites and L i is the likelihood of obtaining the nucleotide, one of fA; C; G; Tg, at site i. Based on the maximum likelihood, trees with higher log likelihoods are considered better. 2. 2 fastDNAml The fastDNAml [26] is one of the most popular programs with reasonable performance and running time. It is an improved version of its predecessor, DNAml [9] in terms of both performance and running time. The main motivation for the fastDNAml was to reduce the computational cost of DNAml. The DNAml was e ective in ....

[Article contains additional citation context not shown here]

G. J. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. fastDNAml: a tool for construction of phylogenetic trees of DNA sequences using maximum likelihood. Computer Applications in the Biosciences, 10(1):41-48, 1994.

....means to reduce topology evaluation time significantly, which represents the by far most cost intensive part of every phylogenetic tree inference process based on the maximum likelihood method irrespective of the tree building algorithm deployed. We implemented our concept in (parallel) fastDNAml [7, 11] and named the resulting program Parallel A(x)ccelerated Maximum Likelihood (PAxML) In tests with alignments of 150 up to 500 sequences, we achieved global run time improvements of 26 up to 65 for both the sequential and the parallel version on various platforms. An important result of this ....

....In Table 2 we list the global run time (secs) of AxML (v2.5) AxML (v1.7) and fastDNAml (v1.2.2) for alignments containing 150, 200, 250 and 500 sequences. The tree inference was conducted without global and local rearrangements (for details on program options see fastDNAml documentation and [7]) # sequences v2.5 v1.7 fastDNAml 150 632 748 1603 200 1227 1443 3186 250 2055 2403 5431 500 10476 12861 26270 Table 2. AxML (v1.7) v2.5) vs. fastDNAml In Table 3 we describe results obtained by comparing the output of the de novo tree building function with and without distance based ....

G. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. fastdnaml: A tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Comput. Appl. Biosci., 10:41--48, 1994.

....the computation of the topology evaluation function used by maximum likelihood based programs. These optimizations are applicable to most existing sequential and parallel programs for phylogenetic tree inference based on the maximum likelihood method, especially derivatives of fastDNAml [3, 10] and the phylip [4, 15] package, and are independent from the specific search space strategy. We implemented the optimizations proposed in this paper in A(x)cce lerated Maxi mum Likelihood (AxML)andParallel AxML (PAxML) based on the latest sequential and parallel releases of fastDNAml ....

....the coarse grained parallelization scheme, even on the hybrid architecture of the Hitachi supercomputer, since it ensures optimal e#ciency and scalability of our algorithmic optimization. 3 Implementation We integrated subtree equality vectors into three existing phylogeny programs: fastDNAml [10], parallel fastDNAml [12] and TrExML [17] Wenametheoptimizedversions AxML, PAxML and ATrExML respectively. About 300 lines of code (# 5 in the sequential code) have been added to those programs, thus demonstrating the e#ciency, simplicity and applicability of our approach. A simple analysis ....

[Article contains additional citation context not shown here]

Olsen, G.J., Matsuda, H., Hagstrom, R., and Overbeek, R. 1994. fastDNAml: A tool for construction of phylogenetic trees of DNA sequences using maximum likelihood. Comput.Appl.Biosci.10: 41-48.

....explain. ML reconstruction consists of two tasks. The first task involves branch length estimation: Given a topology, find branch lengths to maximize the likelihood. This task is accomplished by iterative methods suchasExpectation Maximization (EM) 8,9] or using Newton Raphson optimizations [25]. Each iteration of these methods requires traversing all tree branches. In addition, these methods are only guaranteed to find local maxima, although in practice they often recover the global maximum [4] The second, more challenging, ML reconstruction task is to find a tree topology that ....

G. J. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. fastDNAmL: a tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Computer Applications in the Biosciences, 10(1):41--48, 1994.

....that is often valuable in phylogenetic inference. 3 Branch swapping Methods For large data sets, the vast majority of phylogenetic tree estimates are obtained by using the uphill searching algorithms employed in popular phylogenetic analysis packages such as PAUP [25] PHYLIP [7] and fastDNAML [17]. These packages all employ similar search strategies, which begin by building an initial tree through stepwise addition of taxa to the tree. The search begins with the unique three taxon tree formed from the first three taxa in the input data set. A four taxon tree is then constructed by ....

....tree. Although algorithms of this type are the most commonly used among biologists, their performance is not well understood. Empirical studies indicate that the time required by such algorithms is O(n 3 ) 7, 23] and that the algorithms may often return estimates that are only locally optimal [17, 23]. While it is clear that the estimates obtained by such algorithms are dependent on the starting point (i.e. the order of the sequences in the input data set) little attention has been paid to the selection of good starting points. It is generally agreed that such algorithms must be run numerous ....

G. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. FastDNAml: A tool for construction of phylogenetic trees of DNA sequences using maximum likelihood. Computations in Applied Biosciences, 10(1):41-- 48, 1994.

....the tree should be xyjzw; and site 5 also suggests that the tree should be xyjzw. Therefore, the preferred tree should be xyjzw. 2.3 Programs and More The well known Phylip [18] package contains a program called dnaml that implements Maximum Likelihood. However, the program fastDNAml [50] is an improved version of dnaml . It runs significantly faster than dnaml . Therefore, fastDNAml will be used in Chapter 4 when we conduct simulation studies. Phylip also contains a program called dnpars which we will be using. The program dnapars implements Maximum CHAPTER 2. QUARTET ....

....really represent real situations. Furthermore, we would like to compare our program with all the other popular phylogenetic methods. In these comparison studies, we decide to pick real trees and real data. The programs will be used here are: ffl fastDNAml, Maximum Likelihood, Package: fastDNAml [50] ffl dnapars, Maximum Parsimony, Package: Phylip [18] ffl neighbor, Neighborhood Joining, Package: Phylip [18] ffl fitch, Least Square Distance Method, Package: Phylip [18] ffl puzzle, Quartet Puzzling, Quartet Puzzle [55] CHAPTER 4. SIMULATION STUDY AND PERFORMANCE COMPARISON 58 ....

Olsen, G.J., H. Matsuda, R. Hagstrom, R. Overbeek. 1994. fastDNAml: a tool for construction of phylogenetic trees of DNA sequences using maximum likelihood. Computer Applications in the Biosciences 10(1):41-48, 1994.

....explain. ML reconstruction consists of two tasks. The first task involves branch length estimation: Given a topology, find branch lengths to maximize the likelihood. This task is accomplished by iterative methods such as Expectation Maximization (EM) 8, 9] or using Newton Raphson optimizations [25]. Each iteration of these methods requires traversing all tree branches. In addition, these methods are only guaranteed to find local maxima, although in practice they often recover the global maximum [4] The second, more challenging, ML reconstruction task is to find a tree topology that ....

G. J. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. fastDNAmL: a tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Computer Applications in the Biosciences, 10(1):41--48, 1994.

....explain. ML reconstruction consists of two tasks. The rst task involves branch length estimation: Given a topology, nd branch lengths to maximize the likelihood. This task is accomplished by iterative methods such as Expectation Maximization (EM) 8, 9] or using Newton Raphson optimization [24]. Each iteration of these methods requires computations that take on the order of the number of taxa times the number of training positions. In addition, these methods are only guaranteed to nd local maxima, although in practice they often recover the global maximum [4] The second, more ....

G. J. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. fastDNAmL: a tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Comp. App. Biosciences, 10(1):41-48, 1994.

....are several methods for inferring phylogenetic trees from genetic sequence data. Approaches using maximum likelihood (ML) based techniques belong to the most respected. Since Felsenstein (1981) has first introduced this approach several further improvements and methods have been published (e.g. Olsen et al. 1994; Strimmer von Haeseler 1996) Here we used the ML approach of Olsen et al. 1994) and present a way to use it for larger datasets. 1 The Algorithm 1.1 Computing the ML Value of a Tree The algorithm introduced by Felsenstein (1981) for computing the maximum likelihood (ML) value for a ....

Olsen, G. J., Matsuda, H., Hagstrom, R., & Overbeek, R. (1994). fastDNAml: A tool for construction of phylogenetic trees of DNA sequences using maximum likelihood.

....tree in turn. The program can also save trees in a variety of file formats, so that it can be used to move trees between programs that use different file formats. Tree file formats TREEVIEW can read tree files produced by the CLUSTAL W (Thompson et al. 1994) COMPONENT (Page, 1993) fastDNAml (Olsen et al. 1994), Hennig86 (Farris, 1988) MacClade (Maddison and Maddison, 1992) PAUP (Swofford, 1993) PHYLIP (Felsenstein, 1993) the Ribosomal Database Project (Maidak et al. 1994) among others. These programs use a variant of the nested parentheses format for describing trees, variously referred to as the ....

Olsen, G.J., Matsuda, H., Hagstrom, R. and Overbeck, R. (1994) fastDNAml: a tool for construction of phylogenetic trees of DNA sequences using maximum likelihood. Comput. Applic. Biosci., 10, 41-48.

No context found.

G. J. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek, "fastDNAml: A Tool for Construction of Phylogenetic Trees of DNA Sequences Using Maximum Likelihood," Comp. Appli. Bio. Sci., 10(1), 41--48 (1994).

No context found.

G. J. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. Fastdnaml: A tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Computer Applications in the Biosciences, 10:41--48, 1994.

No context found.

G. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. fastDNAml: a tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Computer Applications in the Biosciences, 10:41--48, 1994.

No context found.

OLSEN, G., H. MATSUDA,R.HAGSTROM, AND R. OVERBEEK. 1994. FastDNAml: A tool for construction of phylogenetic trees of DNA sequences using maximum likelihood. Comput. Appl. Biosci. 10:41-- 48.

No context found.

G.J. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. FastDNAml: A tool for construction of phylogenetic trees of DNA sequences using maximum likelihood. Computer Applications in the Biosciences, 10:41--48, 1994.

No context found.

G. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. fastDNAml: a tool for construction of phylogenetic trees 12 of DNA sequences using maximum likelihood. Computer Applications in the Biosciences (CABIOS), 10:41{ 48, 1994.

No context found.

G. Olsen, H. Matsuda, R. Hagstrom, and R. Overbeek. fastdnaml: A Tool for Construction of Phylogenetic Trees of DNA Sequences using Maximum Likelihood. Comput. Appl. Biosci., 10:41--48, 1994.

No context found.

Olsen, G. J., H. Matsuda, R. Hagstrom, and R. Overbeek. 1994. fastDNAml: A tool for construction of phylogenetic trees of DNA sequences using maximum likelihood. Comput. Appl. Biosci. 10: 41-48.

No context found.

Olsen, G., Matsuda, H., Hagstrom, R., and Overbeek, R. Fastdnaml: a tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Current Applications in Biosciences 10 (1994), 41--48.

No context found.

Olsen GJ, Natsuda H, Hagstrom R, Overbeek R: FastDNAML: A Tool for construction of phylogenetic trees of DNA sequences using maximumlikelihood. Comput. Applic. Biosci. 1994, 10:41--48.

No context found.

Olsen, G., Matsuda, H., Hagstrom, R., and Overbeek, R. Fastdnaml: a tool for construction of phylogenetic trees of dna sequences using maximum likelihood. Current Applications in Biosciences 10 (1994), 41{ 48.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC