phangorn
Phylogenetic Reconstruction and Analysis
Allows for estimation of phylogenetic trees and networks
using Maximum Likelihood, Maximum Parsimony, distance methods and
Hadamard conjugation. Offers methods for tree comparison, model
selection and visualization of phylogenetic networks as described in
Schliep et al. (2017)
phangorn is a package for phylogenetic reconstruction and analysis in the R language. phangorn offers the possibility of reconstructing phylogenies with distance based methods, maximum parsimony or maximum likelihood (ML) and performing Hadamard conjugation. Extending the general ML framework, this package provides the possibility of estimating mixture and partition models. Furthermore, phangorn offers several functions for comparing trees, phylogenetic models or splits, simulating character data and performing congruence analyses.
You can install
- the latest released version
install.packages("phangorn") - the latest development version
devtools::install_github("KlausVigo/phangorn")
To install the development version you may need to install the Biostrings and seqLogo package from bioconductor first:
source("https://bioconductor.org/biocLite.R")
biocLite("Biostrings", "seqLogo")
Also the development version usually depends on tha latest ape development version and information to download can be found here
If you use phangorn please cite:
Schliep K.P. 2011. phangorn: phylogenetic analysis in R. Bioinformatics, 27(4) 592-593
License
phangorn is licensed under the GPLv2.
News
CHANGES in PHANGORN VERSION 2.4.0
NEW FEATURES
o new function add.tips
BUG FIXES
o densiTree, bug report by Richel Bilderbeek
o tree rearrangement in optim.pml return tree with double edge matrix
o phyDat should work now for named vectors
OTHER CHANGES
o many more unit tests
o more consistent coding style
o flashClust and seqLogo are not suggested packages any more
CHANGES in PHANGORN VERSION 2.3.0
NEW FEATURES
o new function mast to compute the maximum agreement subtree
o identify.networx to identify splits in a network
o densiTree got more attributes
o unique.dist, to filter duplicate distances
o ancestral.pml and ancestral.pars are now may return proper phyDat objects
and in for DNA may return ambiguous states
BUG FIXES
o densiTree may if the consensus tree had different ordering of labels
OTHER CHANGES
o acctran allows multiPhylo objects as input
o rNNI has been completely rewritten and is faster
CHANGES in PHANGORN VERSION 2.2.0
NEW FEATURES
o midpoint in now generic and works for multiPhylo objects
o as.bitsplits.splits to better interact with ape
BUG FIXES
o bug fix in RI and CI for ambiguous data
OTHER CHANGES
o added roxygen2 documentation, so R code is now much better documented,
reorganisation of some documentation
o many more unit tests
CHANGES in PHANGORN VERSION 2.1.0
NEW FEATURES
o new functions to compute the (approximate) SPR distance (sprdist, SPR.dist)
contributed by Leonardo de Oliveira Martins.
o super tree methods based on NNI and SPR distances
BUG FIXES
o fixed bug in KF.dist
OTHER CHANGES
o improvements to as.networx. It often now produces networks with less edges
resulting in much nicer plots
o plot.networx does take a different layout algorithm
o as.data.frame.phyDat and as.character.phyDat return amino acids
now in upper cases
o more unit tests
o improved cbind.phyDat, faster and more flexible
o phangorn now requires ape 4.0
o phangorn now imports Rcpp, but not nnls any more
CHANGES in PHANGORN VERSION 2.0.4
NEW FEATURES
o new weighted Robinson-Foulds (1979) distance wRF, suggested by Michelle Kendall
and normalized Robinson-Foulds suggested by Sereina Rutschmann
o codon model "YN98" can be called directly, improved documentation
o bootstrap.phyDat got an new argument jumble,
which randomizes input order of sequences
OTHER CHANGES
o more unit tests
CHANGES in PHANGORN VERSION 2.0.3
NEW FEATURES
o new function maxCladeCred computing the maximum clade credibility tree
from a sample of trees
o function read.nexus.networx and write.nexus.networx to import / export
phylogenetic networx to SplitsTree
o function as.AAbin.phyDat to exchange data with the new AAbin class from ape
o likelihood ratchet (Vos 2003) topology optimisation in optim.pml
o with KF.dist (Kuhner & Felsenstein) and path.dist can be used like RF.dist
BUG FIXES
o improvements to optim.pml to avoid numerical problems,
can now handle much more taxa (still experimantal)
OTHER CHANGES
o mrca.phylo can be used like mrca from ape (faster for large trees)
o individual splits can be colored in lento plots (suggested by France Thouzé)
o plot.networx now (silently) returns a networx object with some of the graphics
parameters, more plot options
o lots of small changes to make handling and identifying splits, edges in
trees and networks easier
o plotBS has a new argument frame and arguments BStrees can be empty
o new vignette IntertwiningTreesAndNetworks
CHANGES in PHANGORN VERSION 2.0.2
OTHER CHANGES
o phangorn now suggests only the Biostrings and not depends on it
o some improvements to bab (branch and bound), may be faster
CHANGES in PHANGORN VERSION 2.0.0
NEW FEATURES
o as.phyDat.MultipleAlignment to excange data with Biostrings
o dist.ml can now compute distances using a discrete gamma model
and the "F81" model.
o optim.pml got a new rearrangement argument and can now
use a stochastic rearrangements similar to Nguyen et al. (2015)
BUG FIXES
o plotBS may double counted edges when rooted trees were in the sample
OTHER CHANGES
o optim.pml uses more C code and should be more robust
o more unit tests
o baseFreq got additional parameter 'all' similar to base.freq in ape
o lots of little improvements in the manual and help files
o modelTest now also shows AIC weights and model="all" will test
all available models
CHANGES in PHANGORN VERSION 1.99.14
NEW FEATURES
o phyDat2alignment to exports files to seqinr
o readDist, writeDist to import / export distance matrices
o cophenetic distance function for splits and networx
o added unit tests
BUG FIXES
o as.splits.networx did not work properly for 4 taxa
(reported by Laurélène Faye)
o RF.dist returned sometimes wrong values
(reported by Andres Dajles)
o plotBS did sometimes not work if the tree had no edge lengths
o plotBS did not work propoerly if input trees were rooted
(reported by Quynh Quach)
o plot.networx ignored cex argument in "2D" plots
o Siblings ignored include.self argument if node is a vector
OTHER CHANGES
o plotBS got an additional argument p to plot only support values
greater than p
o pml and optim.pml now uses more C-code (and is a bit faster)
o defaults in modelTest changed
o discrete.gamma is now exported in the NAMESPACE
CHANGES in PHANGORN VERSION 1.99-13
OTHER CHANGES
o improved importing and conversion of data
o improved stability of pml and optim.pml
CHANGES in PHANGORN VERSION 1.99-12
NEW FEATURES
o added neighborNet algorithm (Bryant and Moulton 2004)
very experimental at the moment
BUG FIXES
o plotBS was not working correctly if bootstraped trees are in
compressed form (bug report by Tobias Müller)
OTHER CHANGES
o many splits and networx methods have been improved
and a vignette was added
o phangorn now suggests only the rgl and not depends on it
to avoid problems on different platforms (suggestion by Matt Pennell)
o new package dependencies knitr for html vignettes and nnls
CHANGES in PHANGORN VERSION 1.99-10
BUG FIXES
o reorder.networx may not work as expected
o Gamma model was not working properly in simSeq.pml
CHANGES in PHANGORN VERSION 1.99-9
BUG FIXES
o bug fixes for clang environment
o midpoint takes care of node labels
CHANGES in PHANGORN VERSION 1.99-8
NEW FEATURES
o pmlPart got an argument rooted to handle rooted trees
o simSeq is now a generic function. This simplifies the
construcion of parametric bootstrap test
o SOWH.test (very experimental)
o as.networx and plot.networx improved considerably
(often generate networks less edges)
and planar graphs are now plotted nicely
BUG FIXES
o fixed some bugs in ancestral.pars
o amino acid model "Blosum62" was not working
OTHER CHANGES
o improvements to read.nexus.splits, write.nexus.splits to
be more consistant with splitstree
o splitsNetwork got an additional argument splits
o help for consensusNet, as.splits, as.networx have been reorganised
and improved
o treedist is much faster for larger trees
o several changes to keep R CMD check happy
o a development version phangorn is now available on github
https://github.com/KlausVigo/phangorn.git
CHANGES in PHANGORN VERSION 1.99-6
NEW FEATURES
o cladePar helps coloring trees
o treedist is faster for larger trees, better documentation and examples
BUG FIXES
o the plot of consensusNet shows now the proper bootstrap values
OTHER CHANGES
o phangorn does not depend only suggest rgl
(should build on OS X now)
o default rearrangement for parsimony is now "SPR"
CHANGES in PHANGORN VERSION 1.99-5
NEW FEATURES
o RF.dist works also on "multiPhylo" objects and is quite fast
o optim.pml can now handle NNI tree arrangements for rooted trees,
still experimental but useful for dating etc.
BUG FIXES
o rNNI did return sometimes trees without tip labels
o SH.test did not work for pmlCluster objects
o df for rooted rooted/ultrametric trees are correctly computed
OTHER CHANGES
o lots of internal code C-code changed
o exports of some of the internal ML function, this should speed up in
future other packages e.g. the colescentMCMC package, which use them
considerably (interface may changes in the future)
o registered C routines
CHANGES in PHANGORN VERSION 1.99-0
NEW FEATURES
o new function dist.p to estimate pairwise polymorphism
p-distances from DNA sequences
BUG FIXES
o as.data.frame.phyDat returned only site patterns and so did
write.phyDat for nexus files
o some of the recently introduced (1.7-4) amino acid models were not known
by all functions and contained NAs
OTHER CHANGES
o changed package imports and depends structure to aviod error messages,
thanks to Brian Ripley
o a lot of the internal C-code has changed
CHANGES in PHANGORN VERSION 1.7-4
NEW FEATURES
o densiTree plots are available now
o new species tree and super tree methods
o more amino acid models
BUG FIXES
o phangorn now depends on rgl instead of suggests rgl,
rgl wants to be loaded before igraph, otherwise a compiling error
on some platforms occured!
o fixed a bug that sometimes caused in pratched to crash
o fixed a bug when using negative indices in subset.phyDat
o the search heuristic SPR in optim.parsimony evaluates now more trees
and is more likely to find better ones
OTHER CHANGES
o underlying C-code for several functions has changed.
less memory reallocations and potentially time savings
hopefully I included not too many bugs
o optimising edge length changed from Jacobi to Gauss-Seidel method
and will hopefully be more robust in the long term!
o Descendants is much faster for option type="all"
o plotAnc gives user more control and produces nicer plots
CHANGES in PHANGORN VERSION 1.7-1
NEW FEATURES
o pmlPart got additional argument model
(request from Santiago Claramunt)
BUG FIXES
o pmlPart should be more robust
OTHER CHANGES
o started reorganising the code
o underlying code of several parsimony functions has changed and
these are now considerably faster
o some examples are changed to allow faster checking on CRAN
CHANGES in PHANGORN VERSION 1.6-5
NEW FEATURES
o dist.hamming handles ambigious states now as dist.ml
(request from Arne Mooers)
BUG FIXES
o phangorn links properly to ape
CHANGES in PHANGORN VERSION 1.6-3
NEW FEATURES
o optim.parsimony has a new search heuristic (SPR)
BUG FIXES
o changed package to work with igraph >= 0.6
OTHER CHANGES
o arguments of pratchet changed
CHANGES in PHANGORN VERSION 1.6-0
NEW FEATURES
o dist.ml has more options and is faster (ca. 5 times for nucleotides and 20 times for amino acids)
BUG FIXES
o plotBS did not work properly with ape version 3.0
OTHER CHANGES
o vignettes changed for a faster compilation of the package
o Ancestors allows a vector of nodes as input
o midpoint uses less memory and works for larger trees (10000 of tips)
o ancestral.pars gives better formated output
CHANGES in PHANGORN VERSION 1.5-1
OTHER CHANGES
o several examples changed for a faster compilation of the package
CHANGES in PHANGORN VERSION 1.5-0
NEW FEATURES
o codon models can be used directly
(dn/ds ratio can be computed)
o modelTest works now also for amino acids
BUG FIXES
o the code to compute RI and CI changed and should be more robust
OTHER CHANGES
o package parallel is used instead of multicore
o vignettes, examples, help improved
o ChangeLog is called NEWS
CHANGES in PHANGORN VERSION 1.4-1
NEW FEATURES
o parsimony branch-and-bould algorithms bab (so far pretty slow and memory intensive)
o more amino acid models
o function nnls.tree to compute non-negative edge weights for
a given tree and a distance matrix
BUG FIXES
o allTrees returns now an integer edge matrix,
this could have caused some problems previously
o CI and RI now take better care of ambiguous states
o dist.ml has default value for amino acids
o as.splits.multiPhylo produces more sensible bipartitions
and so lento and consensusNet produce more useful plots
(thanks to Emmanuel Paradis)
OTHER CHANGES
o several changes to the networx classes and methods
o modelTest now also returns the function calls of the estimated models,
which can be used in downstream analyses
o vignette "Trees" has a few more examples
o dist.ml is more general (base frequencies and rate matrix can be supplied)
o pml objects are more compact, thanks to the Matrix package
o xtable is now a suggested package (needed for vignettes)
CHANGES in PHANGORN VERSION 1.4-0
NEW FEATURES
o plot.network to plot split networks in 3D (requires rgl) and 2D
(still very experimantal)
o consensusNet computes consensus networks
o Lento plot allows to take multiPhylo objects as input
BUG FIXES
o CI and RI did not work with only one site pattern present
o pratchet returned only one, not all of the best trees found
OTHER CHANGES
o phangorn now requires the Matrix, igraph and rgl packages
o designTree returns a sparse Matrix and this can save a lot of memory
o internal code for computing bipartitions is much faster for large trees,
and so are several functions depending on it, e.g. RF.dist, treedist, Descendants
CHANGES in PHANGORN VERSION 1.3-1
BUG FIXES
o the multicore package may failed, if executed in a GUI environment,
more error checks included
o optim.pml, in rare cases may failed to optimize edge length
(should be more robust now)
OTHER CHANGES
o some changes to keep R CMD check happy
o modelTest, pratchet, bootstrap.pml, bootstrap.phyDat
got an additional argument multicore option to switch
between serial and parallel execution
CHANGES in PHANGORN VERSION 1.3-0
NEW FEATURES
o acctran to assign edge length to parsimony trees
OTHER CHANGES
o phangorn can now be cited
o additional and improved ancestral reconstructions methods
(ACCTRAN, MPR)
o new vignette describing ancestral sequence reconstruction
CHANGES in PHANGORN VERSION 1.2-0
NEW FEATURES
o new function pratchet (parsimony ratchet)
o new function midpoint for rooting trees
o new function pruneTree to build concensus trees from node labels
(e.g. bootstrap values)
o multicore support for modelTest
BUG FIXES
o ancestral.pars sometimes did not show all possible states
o the call-attributes did not get proper changed in update.pml and
optim.pml
OTHER CHANGES
o there is now a general help page displayed with '?phangorn'
o dist.hamming is faster
o getClans, getSlices and getDiverstity can now
handle multifurcating trees
CHANGES in PHANGORN VERSION 1.1-2
NEW FEATURES
o more generic methods for class splits (print, as.matrix)
o plotBS can plot now cladograms and phylograms
BUG FIXES
o read.phyDat sometimes did not work properly for amino acids
CHANGES in PHANGORN VERSION 1.1-1
NEW FEATURES
o optim.pml allows to optimise rooted trees
OTHER CHANGES
o description of getClans improved
CHANGES in PHANGORN VERSION 1.1-0
NEW FEATURES
o Consistency Index (CI) and and Rentention Index (RI)
o clanistic tools
o new generic function cbind.phyDat
o optim.parsimony works now also with the fitch algorithm,
faster than the sankoff version
BUG FIXES
o treedist and RF.dist now check whether trees are binary and
try to handle multifurcations (thanks to Jeremy Beaulieu for bug fixes)
OTHER CHANGES
o second vignette describing some special features
o allTrees is faster
o trace and pml.control are now more consistent
o optim.pml uses less memory and can be faster
for data with lots of characters
CHANGES in PHANGORN VERSION 1.0-2
BUG FIXES
o pml.control did not work properly
OTHER CHANGES
o pmlCluster, pmlMix and pmlPart gained an attribute control,
which controls the outermost loop
o some more error checking for pml and parsimony classes
(thanks to Emmanuel and Liat)
CHANGES in PHANGORN VERSION 1.0-1
NEW FEATURES
o ancestral sequence reconstruction
(parsimony and likelihood based)
o a small convenience function acgt2ry for ry-coding
o as.phylo.splits computes a tree from compatible splits
BUG FIXES
o a small error in pmlCluster was fixed
OTHER CHANGES
o upgma changed to accommodate change in as.phylo.hclust
o lento plots are looking nicer
CHANGES IN PHANGORN VERSION 1.0-0
NEW FEATURES
o implementation of many nucleotide substitution models
(additional general transition models can be defined)
o new function modelTest, comparison of different phylogenetic model
with AIC or BIC
o Lento plot
o subset functions for phyDat objects
BUG FIXES
o an error in pace is fixed
OTHER CHANGES
o parsimony (fitch and sankoff) can now handle multiPhylo objects
o splits structure (which is a list of bipartitions), used by lento
and hadamard conjugation
o phyDat objects can be more general generated using a contrast
matrix
CHANGES IN PHANGORN VERSION 0.99-6
NEW FEATURES
o pace, extracts the ancestral states of the root of a tree
using the sankoff algorithm
BUG FIXES
o fixed a bug in dist.ml (thanks to Emmanuel)
o fixed a bug introduced to SH.test in 0.99-5
OTHER CHANGES
o fixed several spelling mistakes in the documentation
CHANGES IN PHANGORN VERSION 0.99-5
NEW FEATURES
o parallel computing via multicore
(so far bootstrap.pml, bootstrap.pml profit under linux)
o compute edge weights for parsimony trees
BUG FIXES
o optim.pml had problems when
o as.character converted ?,- wrongly to NA
o fitch needed binary trees as input, otherwise pscore is
likely to be wrong (returns now a warning)
o optim.pml had a problem with identical sequences
OTHER CHANGES
o optim.parsimony returns now a tree with edge weights
o vignette is enhanced, I fixed some spelling mistakes and added
some more examples.
CHANGES IN PHANGORN VERSION 0.99-4
NEW FEATURES
o new generic function unique.phyDat
OTHER CHANGES
o internal data format phyDat changed
and data are stored more memory efficient
(optim.pml and friends use less memory and may be faster)
CHANGES IN PHANGORN VERSION 0.99-3
BUG FIXES
o RF.dist sometimes returned wrong distances
o rate parameter is now properly normalized in pml.Part
and pmlCluster
o simSeq had problems simulating a single character
NEW FEATURES
o rSPR and rNNI to simulate tree rearrangements
CHANGES IN PHANGORN VERSION 0.99-2
NEW FEATURES
o bootstrap.pml and bootstrap.phyDat: parametric
bootstrap methods
o simSeq: A new function to simulate sequence data
o read.phyDat: simplifies reading in alignments
o SH.test: Shimodaira-Hasegawa test
o RF.dist: Robinson-Foulds distance as replacement
for treedist (uses less memory and is much faster)
BUG FIXES
o dist.ml returned wrong variances.
o as.character.phyDat, as.data.frame caused an error
for alignments with only one site.
OTHER CHANGES
o added vignette describing how to perform some
standard phylogenetic analysis with phangorn.
o more functions to convert between different data formats.
o NNI tree search is now general possible for
partition models (pmlPart, pmlCluster)
CHANGES IN PHANGORN VERSION 0.0-5
BUG FIXES
o Solved a namespace problem with ape (>=2.2-3).
CHANGES IN PHANGORN VERSION 0.0-4
NEW FEATURES
o splitsNetwork fits a phylogenetic network using a L1 penalty.
(High memory consumption)
o pmlPen: A new function to estimate penalized likelihood models
for sets of edge weights in mixtures or partition models.
BUG FIXES
o dist.ml should be more forgiving for different inputs.
OTHER CHANGES
o a new dataset.
CHANGES IN PHANGORN VERSION 0.0-3
NEW FEATURES
o amino acid models
o several new maximum likelihood models:
mixture models (pmlMix), and some model for
phylogenomic data partition models (pmlPart),
and clustering of partitions / genes (pmlCluster)
(still experimental, feed back wellcome)
o design matrices for phylogenetic distance methods
o added some functions useful for simulations
(nni, allTrees)
OTHER CHANGES
o the data object phyDat changed slightly internally
o a new dataset
o read.aa to read amino acid data in phylip format
based on read.dna from the ape package
CHANGES IN PHANGORN VERSION 0.0-2
NEW FEATURES
o more generic functions (plot.pml, update.pml)
BUG FIXES
o the "Fitch" algorithm in parsimony contained a bug
OTHER CHANGES
o pml has a cleaner interface (less parameter)
o new faster parsimony analysis (more compiled C-Code)
o added NAMESPACE