Program T-REX (Tree and reticulogram reconstruction)
What does the T-REX web server do?
The T-REX web server allows users to carry out several popular algorithms for inferring and validating phylogenetic trees and networks, and performing other related bioinformatics tasks. A number of new algorithms developed in our laboratory have been also added to T-REX. The up-to-date web server version of T-REX includes:
1) Methods for the visualization, drawing and interactive manipulation of phylogenetic trees (using Hierarchical, Radial and Axial types of drawing). For instance, the Newick Viewer application allows users to visualize a phylogenetic tree coded by its Newick string and the Newick Builder application allows users to draw a phylogenetic tree and then save it as a Newick string.
2) Inference of phylogenetic trees using distances (NJ, NINJA, BioNJ, UNJ, ADDTREE, MW, FITCH, Circular order reconstruction), maximum likelihood (PHYML, RAxML and ML methods from PHYLIP) and maximum parsimony (MP methods from PHYLIP). T-Rex also carries out bootstrap and jackknife resampling to assess strength of support of the tree or network branches.
With tree distance methods a tree metric (i.e. additive distance) is fitted to the given dissimilarity matrix that include evolutionary distances between the considered species (i.e. taxa, objects). For instance, the program carries out six distance-based methods for fitting a tree metric (distance representable by a tree with non-negative branch lengths) to the given dissimilarity matrix. The following distance methods are available:
· ADDTREE by Sattath and Tversky (1977);
· Neighbor-joining (NJ) method by Saitou and Nei (1987);
· NINJA large-scale neighbor-joining by Wheeler (2009);
· BioNeighbor-joining (BioNJ) method by Gascuel (1997);
· Unweighted neighbor-joining method (UNJ) by Gascuel (1997);
· FITCH method from the PHYLIP package by Felsenstein (1989);
The first three methods, NJ (including NINJA) and ADDTREE, are among the most frequently used methods for inferring phylogenetic trees. The third and forth method, called BioNJ and UNJ, use the same that NJ selection criterion, but different estimation and reduction formulae to infer the tree. The fifth method reconstructs a tree using circular orders of elements associated with a given dissimilarity. This fitting method, presented in Makarenkov and Leclerc (1997), was inspired by Yushmanov's (1984) paper which introduced the concept of circular orders of elements corresponding to the circular (say, clockwise) scanning of leaves of a tree drawing on the plane. The sixth method, called MW, looks for the best additive tree with respect to the given dissimilarity and weight matrices. This method allows for arbitrary weights which may be chosen according to one of the classical weighting models proposed in the literature. The tree obtained by any of the six above-mentioned methods is then polished by means of the procedure of quadratic approximation (see Barthélemy and Guénoche, 1991 in the unweighted case or Makarenkov and Leclerc, 1999 in the weighted case) of branch lengths in order to improve the value of the least-squares criterion and to avoid negative branch lengths.
3) T-REX also allows one to reconstruct phylogenetic trees from a distance matrix containing missing values (i.e. incomplete matrix). The following four fitting methods are available:
4) As far as reticulogram inferring (i.e. reticulated cladogram or network) is concerned, the program first builds a supporting phylogenetic tree using one of the available tree inferring methods. Following that step, a reticulation (a new branch) that minimizes the least-squares or weighted least-squares loss function is added to the network at each step of the inferring procedure. Two statistical criteria (Q1 and Q2) have been proposed in order to measure the gain in fit when reticulations are added. The minimum of each of these criteria may suggest a stopping rule for the addition of reticulations. Thus, the user can either select an appropriate criterion to stop the procedure of adding reticulations or indicate an exact number of reticulations to be added to the reticulogram. For a detailed description of the reticulogram reconstruction method, the user is referred to the papers by Legendre and Makarenkov (2002) and Makarenkov and Legendre (2004).
5) When HGT (horizontal gene transfer) detection option is selected, the program infers an optimal (i.e. minimum-cost) scenario of horizontal (i.e. lateral) gene transfers reconciling a given pair of species and gene trees (for more detail, see Boc, Philippe and Makarenkov, 2010). Statistical validation of the obtained gene transfers by bootstrapping can be also performed. HGT Consensus, Parallel and Interactive versions of the HGT-Detection algorithm are also available. A consol version of the HGT-Detection program allowing for detecting partial horizontal gene transfer events (when only a part of the gene is transferred and acquired by the host allele through intragenic recombination) is also provided.
6) MAFFT and MUSCLE algorithm, which is one of the most widely used multiple sequence alignment tools, is available.
7) The following popular Sequence to Distance transformations:
Uncorrected, Jukes-Cantor, Tajima-Nei, Kimura 2-parameters, Tamura, Jin-Nei gamma, Kimura protein, LogDet, F84, WAG, JTT and LG.
are made available.
8) Computation of the Robison and Foulds topological distance between two (or more) phylogenetic trees can be carried out.
9) Tree format conversion - from Newick to Distance matrix format and vice versa - is provided.
10) Generation of a specified number of random phylogenetic trees with a specified number of leaves (i.e., species, taxa) and a predefined average branch length according the procedure described by Kuhner and Felsenstein (1994).
As results the program provides:
Windows 9x/ME/NT/2000/XP executable, documentation and sample files.
The complete documentation is included in the T-REX package.
documentation and sample files.
Read the documentation (also included with the above package)
32-bit DOS Console application, suitable for DOS sessions under Windows 95/98/NT. Minimal user interface, provides text-based results (fitted distance matrix, tree edge lengths and statistics).
So far, does not include the algorithm for detection of Horizontal Gene Transfers. Read the documentation (also included with the above package). A complete graphical application is in preparation.
C++ source code of the 32-bit DOS version above, so you can compile it for your favorite operating system. Make sure you also read the documentation.
Last updated on Thursday, May 3, 2012 by Vladimir Makarenkov