Supertrees by Flipping

Increasing Data Transparency and Estimating Phylogenetic Uncertainty in Supertrees: Approaches Using Nonparametric Bootstrapping

“…Second, the relative performance of the conventional supertree methods in these experiments was somewhat surprising. For example, the poor performance of MRF in our empirical experiments contradicts previous simulation studies in which MRF consistently outperformed other supertree approaches (e.g., Chen et al, 2002Chen et al, , 2003Eulenstein et al, 2004). This apparent anomaly may stem from our combination of strongly conflicting trees with identical terminal taxa.…”

“…Our rational for choosing these four methods from the growing pool of available supertree approaches (currently comprising more than a dozen alternatives; e.g., Wilkinson et al, 2005b;BinindaEmonds, 2004a) deserves comment: MRP is the most frequently used method (e.g., Bininda-Emonds, 2004b), MinCut and modified MinCut methods share the unique (and highly desirable) property of running in polynomial time on species number (e.g., Semple and Steel, 2000;Page, 2002), and MRF has typically outperformed other methods under simulation (e.g., Bininda-Emonds and Sanderson, 2001;Chen et al, 2002;Eulenstein et al, 2004).…”

“…First, the performance of the bootstrap procedures is likely to scale with the attendant computational burden, such that a greater computational investment is likely to pay dividends in the form of more accurate supertree estimates. Second, although MRP (like many parsimony-based phylogenetic inference algorithms) is NP-complete (Graham and Foulds, 1982), recent developments in supertree estimation have led both to heuristic parsimony-based algorithms that are fixed-parameter tractable (Chen et al, 2002) and to more efficient algorithms with polynomial running times (e.g., Semple and Steel, 2000;Page, 2002); use of these algorithms could greatly ameliorate the computational burden of the proposed bootstrapping approaches. Finally, the supertree bootstrapping approaches are inherently suited to distribution in a parallel processing environment.…”

“…Combination of component phytogenies using conventional supertree methods.-For the purpose of comparison, we also combined the set of source trees for each partition set using several conventional supertree algorithms, including standard MRP (Baum, 1992;Ragan, 1992), matrix representation with flipping (MRF; Chen et al, 2002Chen et al, , 2003Burleigh et al, 2004), MinCut (MC; Semple and Steel, 2000), and modified MinCut (MC*; Page, 2002). Our rational for choosing these four methods from the growing pool of available supertree approaches (currently comprising more than a dozen alternatives; e.g., Wilkinson et al, 2005b;BinindaEmonds, 2004a) deserves comment: MRP is the most frequently used method (e.g., Bininda-Emonds, 2004b), MinCut and modified MinCut methods share the unique (and highly desirable) property of running in polynomial time on species number (e.g., Semple and Steel, 2000;Page, 2002), and MRF has typically outperformed other methods under simulation (e.g., Bininda-Emonds and Sanderson, 2001;Chen et al, 2002;Eulenstein et al, 2004).…”

Edit distance measure for graphs

“…The edit distance problem began with the following question of Chen, Eulenstein, Fernández-Baca and Sanderson [6]: given a bipartite graph G, how many edgedeletions plus edge-additions are necessary to ensure that G has no copy of M as an induced subgraph. The study of the edit distance in graphs was originated independently by Axenovich, Kézdy and Martin [4] and Alon and Stav [3].…”