pml: Likelihood of a tree.

Description

pml computes the likelihood of a phylogenetic tree given a sequence alignment and a model. optim.pml optimizes the different model parameters.

Usage

pml(tree, data, bf=NULL, Q=NULL, inv=0, k=1, shape=1, rate=1, model="", ...)     
optim.pml(object, optNni=FALSE, optBf=FALSE, optQ=FALSE, optInv=FALSE, optGamma=FALSE, optEdge=TRUE, optRate=FALSE, optRooted=FALSE,  control = pml.control(epsilon=1e-08, maxit=10, trace=1), model = NULL,  rearrangement = ifelse(optNni, "NNI","none"), subs = NULL,  ratchet.par = list(iter = 20L, maxit = 100L, prop = 1/3),...)  
pml.control(epsilon = 1e-08, maxit = 10, trace = 1)

Arguments

tree

A phylogenetic tree, object of class phylo.

data

An alignment, object of class phyDat.

Base frequencies.

A vector containing the lower triangular part of the rate matrix.

inv

Proportion of invariable sites.

Number of intervals of the discrete gamma distribution.

shape

Shape parameter of the gamma distribution.

rate

Rate.

model

allows to choose an amino acid models or nucleotide model, see details.

object

An object of class pml.

optNni

Logical value indicating whether toplogy gets optimized (NNI).

optBf

Logical value indicating whether base frequencies gets optimized.

optQ

Logical value indicating whether rate matrix gets optimized.

optInv

Logical value indicating whether proportion of variable size gets optimized.

optGamma

Logical value indicating whether gamma rate parameter gets optimized.

optEdge

Logical value indicating the edge lengths gets optimized.

optRate

Logical value indicating the overall rate gets optimized.

optRooted

Logical value indicating if the edge lengths of a rooted tree get optimized.

ratchet.par

search parameter for stochastic search

rearrangement

type of tree tree rearrangements to perform, one of "none", "NNI", "stochastic" or "ratchet"

control

A list of parameters for controlling the fitting process.

subs

A (integer) vector same length as Q to specify the optimization of Q

...

Further arguments passed to or from other methods.

epsilon

Stop criterion for optimisation (see details).

maxit

Maximum number of iterations (see details).

trace

Show output during optimization (see details).

Value

pml or optim.pml return a list of class pml, some are useful for further computations like

Details

The topology search uses a nearest neighbor interchange (NNI) and the implementation is similar to phyML. The option model in pml is only used for amino acid models. The option model defines the nucleotide model which is getting optimised, all models which are included in modeltest can be chosen. Setting this option (e.g. "K81" or "GTR") overrules options optBf and optQ. Here is a overview how to estimate different phylogenetic models with pml:

model	optBf
optQ	Jukes-Cantor
FALSE	FALSE
F81	TRUE
FALSE	symmetric
FALSE	TRUE

Via model in optim.pml the following nucleotide models can be specified: JC, F81, K80, HKY, TrNe, TrN, TPM1, K81, TPM1u, TPM2, TPM2u, TPM3, TPM3u, TIM1e, TIM1, TIM2e, TIM2, TIM3e, TIM3, TVMe, TVM, SYM and GTR. These models are specified as in Posada (2008).

So far 17 amino acid models are supported ("WAG", "JTT", "LG", "Dayhoff", "cpREV", "mtmam", "mtArt", "MtZoa", "mtREV24", "VT","RtREV", "HIVw", "HIVb", "FLU", "Blossum62", "Dayhoff_DCMut" and "JTT_DCMut") and additionally rate matrices and amino acid frequencies can be supplied.

It is also possible to estimate codon models (e.g. YN98), for details see also the chapter in vignette("phangorn-specials").

If the option 'optRooted' is set to TRUE than the edge lengths of rooted tree are optimized. The tree has to be rooted and by now ultrametric! Optimising rooted trees is generally much slower. pml.control controls the fitting process. epsilon and maxit are only defined for the most outer loop, this affects pmlCluster, pmlPart and pmlMix. epsilon is defined as (logLik(k)-logLik(k+1))/logLik(k+1), this seems to be a good heuristics which works reasonably for small and large trees or alignments. If trace is set to zero than no out put is shown, if functions are called internally than the trace is decreased by one, so a higher of trace produces more feedback.

If rearrangement is set to stochastic a stochastic search algorithm similar to Nguyen et al. (2015). and for ratchet the likelihood ratchet as in Vos (2003). This should helps often to find better tree topologies, especially for larger trees.

References

Felsenstein, J. (1981) Evolutionary trees from DNA sequences: a maximum likelihood approach. Journal of Molecular Evolution, 17, 368--376.

Felsenstein, J. (2004). Inferring Phylogenies. Sinauer Associates, Sunderland.

Yang, Z. (2006). Computational Molecular evolution. Oxford University Press, Oxford.

Adachi, J., P. J. Waddell, W. Martin, and M. Hasegawa (2000) Plastid genome phylogeny and a model of amino acid substitution for proteins encoded by chloroplast DNA. Journal of Molecular Evolution, 50, 348--358

Rota-Stabelli, O., Z. Yang, and M. Telford. (2009) MtZoa: a general mitochondrial amino acid substitutions model for animal evolutionary studies. Mol. Phyl. Evol, 52(1), 268--72

Whelan, S. and Goldman, N. (2001) A general empirical model of protein evolution derived from multiple protein families using a maximum-likelihood approach. Molecular Biology and Evolution, 18, 691--699

Le, S.Q. and Gascuel, O. (2008) LG: An Improved, General Amino-Acid Replacement Matrix Molecular Biology and Evolution, 25(7), 1307--1320

Yang, Z., R. Nielsen, and M. Hasegawa (1998) Models of amino acid substitution and applications to Mitochondrial protein evolution. Molecular Biology and Evolution, 15, 1600--1611

Abascal, F., D. Posada, and R. Zardoya (2007) MtArt: A new Model of amino acid replacement for Arthropoda. Molecular Biology and Evolution, 24, 1--5

Kosiol, C, and Goldman, N (2005) Different versions of the Dayhoff rate matrix - Molecular Biology and Evolution, 22, 193--199

L.-T. Nguyen, H.A. Schmidt, A. von Haeseler, and B.Q. Minh (2015) IQ-TREE: A fast and effective stochastic algorithm for estimating maximum likelihood phylogenies. Molecular Biology and Evolution, 32, 268--274.

Vos, R. A. (2003) Accelerated Likelihood Surface Exploration: The Likelihood Ratchet. Systematic Biology, 52(3), 368--373

Yang, Z., and R. Nielsen (1998) Synonymous and nonsynonymous rate variation in nuclear genes of mammals. Journal of Molecular Evolution, 46, 409-418.

Lewis, P.O. (2001) A likelihood approach to estimating phylogeny from discrete morphological character data. Systematic Biology 50, 913--925.

Examples

Run this code

  example(NJ)
# Jukes-Cantor (starting tree from NJ)  
  fitJC <- pml(tree, Laurasiatherian)  
# optimize edge length parameter     
  fitJC <- optim.pml(fitJC)
  fitJC 
  
## Not run:     
# # search for a better tree using NNI rearrangements     
#   fitJC <- optim.pml(fitJC, optNni=TRUE)
#   fitJC   
#   plot(fitJC$tree)
# 
# # JC + Gamma + I - model
#   fitJC_GI <- update(fitJC, k=4, inv=.2)
# # optimize shape parameter + proportion of invariant sites     
#   fitJC_GI <- optim.pml(fitJC_GI, optGamma=TRUE, optInv=TRUE)
# # GTR + Gamma + I - model
#   fitGTR <- optim.pml(fitJC_GI, rearrangement = "stochastic", 
#       optGamma=TRUE, optInv=TRUE, model="GTR") 
# ## End(Not run)


# 2-state data (RY-coded)  
  dat <- acgt2ry(Laurasiatherian) 
  fit2ST <- pml(tree, dat) 
  fit2ST <- optim.pml(fit2ST,optNni=TRUE) 
  fit2ST
# show some of the methods available for class pml
  methods(class="pml")

Run the code above in your browser using DataLab