oncotree {Oncotree}

Build and display an oncogenetic tree

Description

Build a directed tree structure to model the process of occurrence of genetic alterations (events) in carcinogenesis. The model is described in more detail in Oncotree-package. Methods for printing a short summary, displaying the tree on an R plot, and producing latex code for drawing the tree (using the ‘pstricks’ and ‘pst-tree’ LaTeX packages) are provided.

Usage

  oncotree.fit(dataset, error.fun = function(x, y){sum((x - y)^2)})
  ## S3 method for class 'oncotree'
print(x, ...)
  ## S3 method for class 'oncotree'
plot(x, edge.weights = c("none", "observed", "estimated"),
    edge.digits=2, node.coords=NULL, plot=TRUE, cex = par("cex"), 
		col.edge=par("col"), col.text=par("col"), col.weight=par("col"),...)
            
  pstree.oncotree(x, edge.weights=c("none","observed","estimated"), edge.digits=2,
                  shape=c("none","oval", "circle", "triangle", "diamond"),
                  pstree.options=list(arrows="->", treefit="loose", 
                                      arrowscale="1.5 0.8", nodesep="3pt"))

Arguments

dataset

A data frame or a matrix with variable names as a listing of genetic events taking on binary values indicating missing (0) or present (1). Each row is an independent sample.

error.fun

A function of two variables that measures the deviation of the observed marginal frequencies of the events (which will be the first argument in the call) from the estimated ones. The false positive and negative error rates are obtained by minimizing error.fun. If error.fun=NULL is used, the error rates are not estimated.

x

An object of class oncotree.

edge.weights

Choice of edge weights to show on the plot.

edge.digits

The number of significant digits to use when displaying edge weights.

node.coords

A matrix with node-coordinates or NULL if the coordinates should be computed automatically (default).

plot

Logical; indicates whether the tree should be plotted.

cex

Scaling factor for the text in the nodes.

col.edge

color of the tree edges.

col.text

color of the node label.

col.weight

color of the edge weights.

...

Ignored for print. For plot these can be graphical parameters passed to lines when the edges are drawn

shape

The shape of the node in the pst-tree representation.

pstree.options

Additional options for pst-tree. See the pstricks documentation for possible values.

Details

‘pst-tree’ is a very flexible package, and very detailed formatting of the tree is possible. pstree.oncotree provides some default settings for drawing trees, but they can be easily overridden: most options can be set in pstree.options, while the appearance of the tree nodes can be controlled by defining a one-parameter \lab command that gives the desired appearance. For example, if red, non-mathematical test is desired in an oval, you could use \newcommand{\lab}[1]{\Toval[name=#1]{{\red #1}}}.

Value

For oncotree.fit: an object of class oncotree which has components

data

data frame used, after dropping events with zero observed frequency, and adding a column for the artificial ‘Root’ node

nmut

number of tree nodes: the number genetic events present in data +1 for the ‘Root’ node

parent

a list containing information about the tree structure with the following components

• childa character vector of the event names starting with ‘Root’

• parenta character vector of the names of the parents of child

• parent.numa numeric vector with column indices corresponding to parent

• obs.weightsraw edge transition probabilities P(child|parent)

• est.weightsedge transition probabilities adjusted for the error rates eps

level

a numeric vector of the depth of each node in the tree (1 for the root, 2 for its children, etc.)

numchild

a numeric vector giving the number of children for each node

levelnodes

a numeric vector of the number of nodes found at each level of the tree

levelgrp

a character matrix with its rows giving the ordered nodes at each level

eps

a numeric vector of length two showing the estimated false positive and negative error rates (if error.fun is not NULL). Do not modify directly, but rather through error.rates<-.

For print.oncotree:

the original object is returned invisibly. It prints a summary showing the number of nodes, the parent-child relationships, and the false positive and negative error rates.

For plot.oncotree:

a matrix with node-coordinates is returned invisibly. The column names of the matrix are the names of the nodes/events (including 'Root'), the rows gives the x- and y-coordinates, respectively. This matrix provides a valid input for node.coords. If plot=TRUE, a plot of the tree is produced.

For pstree.oncotree:

a character string with the LaTeX code needed to draw a tree. \usepackage{pstricks,pst-tree} is required in the preamble of the LaTeX file, and it should be processed through a PostScript intermediary (DVIPS or similar) and not through PDFLaTeX.

Author(s)

Lisa Pappas

References

Szabo, A. and Boucher, K. (2002) Estimating an oncogenetic tree when false negative and positives are present. Mathematical Biosciences, 176/2, 219-236.

See Also

bootstrap.oncotree,error.rates<-, generate.data,ancestors,distribution.oncotree

Examples

  data(ov.cgh)
  ov.tree <- oncotree.fit(ov.cgh, error.fun=function(x,y){max(abs(x-y))})
  ov.tree
  nodes <- plot(ov.tree, edge.weights="est")
  #move the Root node to the left
  nodes["x","Root"] <- nodes["x","8q+"]
  plot(ov.tree, node.coords=nodes)
  #output for pstricks+pst-tree
  pstree.oncotree(ov.tree, edge.weights="obs", shape="oval")