"phylo"Applying various transformation to the branches of a phylogenetic tree.
# S3 method for phylo
rescale(x, model = c("BM", "OU", "EB", "nrate", "lrate",
"trend", "lambda", "kappa", "delta", "white", "depth"), ...)A transformation function (or rescaled phylogenetic tree of class 'phylo' (ape format) is returned. Possible transforms include the following:
is the Brownian motion model, which fits a random walk with variance sigsq
is the Ornstein-Uhlenbeck model (Butler and King 2004), which fits a random walk with a central tendency with an attraction strength proportional to the parameter alpha. The OU model is called the hansen model in ouch, although the way the parameters are fit is slightly different here. The parameter used for transformation is alpha
is the Early-burst model (Harmon et al. 2010) and also called the ACDC model (accelerating-decelerating; Blomberg et al. 2003). Set by the a rate parameter, EB fits a model where the rate of evolution increases or decreases exponentially through time, under the model r[t] = r[0] * exp(a * t), where r[0] is the initial rate, a is the rate change parameter, and t is time. The parameter used for transformation is a.
is the multiple-rates model where time slices have independent rates of evolution. The parameters used for transformation are time and rate, both of which may be vectors. If rates for several time slices are given, the vectors time and rate must match in length. The time argument is expected in relative branching heights (where the root is 0 and the tips of an ultrametric tree terminate at a relative time of 1). Note that the default behavior is for the multirate transformation to rescale the tree to its original height, while preserving the relative rates across the tree. Note also that the default initial relative rate (sigsq, for the rootmost time slice) is 1.
is the multiple-rates model where local clades have independent rates of evolution. The parameters used for transformation are node and rate, both of which may be vectors. If rates for several lineages are given, the vectors node and rate must match in length. The node argument is expected to have node identifiers consistent with ape labeling (see read.tree). This node argument defines where breakpoints occur in the tree (at which a transition to the associated relative rate occurs). Note that the stem branch associated with the node is included as part of the defined rate partition. Rates given are relative to a default rootmost partition with a rate scalar (sigsq) of 1.
is a diffusion model with linear trend in rates through time. The parameter used for transformation is slope.
is one of the Pagel (1999) models that fits the extent to which the phylogeny predicts covariance among trait values for species. The model effectively transforms the tree as follows: values of lambda near 0 cause the phylogeny to become more star-like, and a lambda value of 1 recovers the BM model. The parameter used for transformation is lambda.
is a punctuational (speciational) model of trait evolution (Pagel 1999), where character divergence is related to the number of speciation events between two species. Note that if there are missing speciation events in the given phylogeny (due to extinction or incomplete sampling), interpretation under the kappa model may be difficult. Considered as a tree transformation, the model raises all branch lengths to an estimated power (kappa). The parameter used for transformation is kappa.
is a time-dependent model of trait evolution (Pagel 1999). The delta model is similar to ACDC insofar as the delta model fits the relative contributions of early versus late evolution in the tree to the covariance of species trait values. Where delta is greater than 1, recent evolution has been relatively fast; if delta is less than 1, recent evolution has been comparatively slow. Intrepreted as a tree transformation, the model raises all node depths to an estimated power (delta). The parameter used for transformation is delta. Note that the default behavior is for the delta transformation to rescale the tree to its original height.
is a white-noise (non-phylogenetic) model, which assumes data come from a single normal distribution with no covariance structure among species
is simply a transformation of the total depth of the tree; stretching the tree has an effect of increasing rates of evolution under Brownian motion (relative to characters evolved on the unstretched tree), and compressing the tree has the opposite effect. The parameter used for transformation is depth.
an object of class "phylo"/
a model used to transform the tree (see Details).
argument(s) to be passed to the transformation function (see Examples).
LJ Harmon and JM Eastman
This function takes a tree and returns either a transformed tree if ... is not empty and gives the parameter value(s) for the tree transformation. If ... is left empty, a function is returned to the user that can be efficiently iterated over many parameter values for transformation. The available models are meant to correspond with changing the model of phenotypic evolution for discrete or continuous characters.
Pagel, M. 1999. Inferring the historical patterns of biological evolution. Nature 401:877-884.
Butler, M.A. and A.A. King, 2004. Phylogenetic comparative analysis: a modeling approach for adaptive evolution. American Naturalist 164:683-695.
Various papers in prep., L. J. Harmon and J. T. Weir.
geo <- get(data(geospiza))
## returning a function
ltrns <- rescale(geo$phy, "lambda")
plot(ltrns(0))
title("lambda: 0.0")
plot(ltrns(0.5))
title("lambda: 0.5")
plot(ltrns(1))
title("lambda: 1")
## transforming the tree
lphy <- rescale(geo$phy, "lambda", 0.5) # transform tree in one fell swoop
plot(lphy)
title("lambda: 0.5")
## multirate tree -- time
rtrns <- rescale(geo$phy, "nrate")
rphy <- rtrns(time=c(0.2, 0.4, 0.6, 0.8), rate=c(2, 4, 8, 16))
plot(rphy)
title("5-rate tree: by time")
## multirate tree -- lineages
mtrns <- rescale(geo$phy, "lrate")
mphy <- mtrns(node=c(25, 20), rate=c(4, 8))
plot(mphy)
title("3-rate tree: by lineages")
Run the code above in your browser using DataLab