Rtreemix (version 1.34.0)

RtreemixModel-class: Class "RtreemixModel"

Description

This class contains all the data needed for characterizing the mutagenetic trees mixture model (mixture parameters, mixture components, ...). The tree components of the model are given as a list of directed graphNEL objects.

Objects from the Class

Objects can be created by calls of the form new("RtreemixModel", ParentData, Weights, WeightsCI, Resp, CompleteMat, Star, Trees). The RtreemixModel class extends the RtreemixData class and specifies the mutagenetic trees mixture model. If the model is not randomly generated the parent class gives the RtreemixData used for learning the mixture model. The directed trees that build up the model are represented as a list of directed graphNEL objects, and their weights (the mixture parameters) are given as a numeric vector. This class can also contain other useful information connected with the mixture model like confidence intervals for the mixture parameters and the edge weights (resulting from a bootstrap analysis), an indicator for the presence of the star component, etc. They are all listed in the text below with brief descriptions. The ParentData is an RtreemixData object that specifies the data used for estimating the mutagenetic trees mixture model. It is not specified for random mixture models, since they are not estimated from a given dataset but generated randomly. The Weights is a numeric vector that contains the mixture parameters of the model. Its length equals the length of the list of tree components Trees. The WeightsCI is a named list with length equal to the length of the Weights. Each list element is a numeric vector of length two specifying the lower and upper bound of the confidence interval for the corresponding mixture parametar. The confidence intervals are derived using the bootstrap method. The Resp is a numeric matrix that specifies the responsibility of each tree component to generate each of the patterns in the ParentData. The number of rows in Resp is equal to the number of trees in the mixture (the length of the list Trees) and the number of columns equals the number of patients in ParentData. For random mixture models it is an empty matrix, since they are not estimated from a given dataset. The CompleteMat is a binary matrix that specifies the complete data in case some measurements for some patients are missing in the data used for learning the model (the ParentData). It has the same size as the matrix specifying the data in ParentData. The missing data are estimated in the EM-algorithm used for fitting the mixture model. When there are no missing data in ParentData, or the model is randomly generated the CompleteMat is an empty matrix. The Star is an indicator of the presence of a noise (star) component and is mostly relevant for models with a single tree component, since it is assumed that mixture models with at least two components always have the noise as a first component. It is of type logical. The Trees is a list of directed graphNEL objects, each for every tree component in the mixture model. The genetic events are represented as nodes in the graphs. The edgeData of each tree can have two attributes: "weight" and "ci". Please see the help page on graph-class and graphNEL-class in the package graph. The "weight" attribute is for edge weight, i.e. the conditional probability that the child event of the edge occured given that the parent event already occured. The "ci" attribute is for the bootstrap confidence intervals for the edge weight (a numeric vector with length two).

Slots

Weights:

Object of class "numeric". The length of the Weights must be equal to the length of Trees.

WeightsCI:

Object of class "list". The length of the WeightsCI must be equal to the length of Weights.

Resp:

Object of class "matrix". The number of rows of Resp must be identical to the length of Trees, and its number of columns to the number of patients in the dataset used for estimating the mixture model (ParentData).

CompleteMat:

Object of class "matrix". When specified (when there are missing data) the size of the CompleteMat must be equal to the size of the matrix used to estimate the model.

Star:

Object of class "logical".

Trees:

Object of class "list". The length of Trees equals the length of Weights.

Extends

Class "RtreemixData", directly.

Methods

CompleteMat

signature(object = "RtreemixModel"): A method used for obtaining the complete dataset, in case there were any missing measurements for some patients in the dataset used to learn the mixture model.

Resp

signature(object = "RtreemixModel"): A method for obtaining the matrix of responsibilities for the trees to generate each of the samples in the dataset used for learning the model (ParentData).

Star

signature(object = "RtreemixModel"): A method for checking the presence of a noise component in the mixture model (informative only for models with one tree component).

Trees

signature(object = "RtreemixModel"): A method for obtaining the tree components of the mixture model as a list of directed graphNEL objects.

Weights

signature(object = "RtreemixModel"): A method for obtaining the mixture parameters (the weights of the trees in the model).

Weights<-

signature(object = "RtreemixModel"): A method for replacing the value of the slot Weights with a specified numeric vector. The components of this vector have to sum up to one.

WeightsCI

signature(object = "RtreemixModel"): A method for obtaining the weights of the mixture parameters.

getData

signature(object = "RtreemixModel"): A method for obtaining the ParentData of the mixture model, i.e. the data used for learning the model.

getTree

signature(object = "RtreemixModel", k = "numeric"): A method for obtaining the k-th tree component of the mixture model as a directed graphNEL object.

numTrees

signature(object = "RtreemixModel"): A method for obtaining the number of tree components building up the mixture model.

plot

signature(x = "RtreemixModel", y = "missing"): A method for visualizing the tree components comprising a mutagenetic trees mixture model. The user can also specify the fontSize (the default value is 8) used for the text labels of the nodes and the edges of the plotted trees. Additionally, one can use the parameter k to plot a certain tree component from the mixture model.

References

Learning multiple evolutionary pathways from cross-sectional data, N. Beerenwinkel et al.

See Also

RtreemixGPS-class, RtreemixStats-class, RtreemixData-class, RtreemixSim-class, fit-methods, bootstrap-methods, generate-methods, comp.models, comp.trees

Examples

## Generate a random RtreemixModel object with 2 components.
rand.mod <- generate(K = 2, no.events = 9, noise.tree = TRUE, prob = c(0.2, 0.8))
show(rand.mod)
plot(rand.mod) ## plot the tree components of the model
plot(rand.mod, k = 2) ## plot the second component of the model

## Draw data from a specified mixture model.
draws <- sim(model = rand.mod, no.draws = 200)
show(draws)

## Create an RtreemixModel object by fitting model to the drawn data.
mod <- fit(data = draws, K = 2, equal.edgeweights = TRUE, noise = TRUE)
show(mod)

## See the values of the slots of the RtreemixModel object.
Weights(mod)
Resp(mod)
CompleteMat(mod)
Star(mod)
Trees(mod)
## See data used for learning the model.
getData(mod)
## See the number of tree components in the mixture model.
numTrees(mod)
## See a specific tree component k.
getTree(object = mod, k = 2)
## See the conditional probabilities assigned to edges of the second tree component.
edgeData(getTree(object = mod, k = 2), attr = "weight")
## See the probability distribution encoded by the model on the set of all possible patterns.
distr <- distribution(model = mod)
distr
## Get the probabilities.
distr$probability
## See the probability distribution encoded by the model on the set of all possible patterns
## calculated for given sampling mode, and input and output parameters.
distr1 <- distribution(model = mod, sampling.mode = "exponential", sampling.param = 1, output.param = 1)
distr1

## Create a RtreemixModel and analyze its variance with the bootstrap method.
mod.boot <- bootstrap(data = draws, K = 2, equal.edgeweights = TRUE, B = 100)

## See the confidence intervals for the mixture parameters (the weights).
WeightsCI(mod.boot)
## See the confidence intervals of the conditional probabilities assigned to the edges.
edgeData(getTree(mod.boot, 2), attr = "ci")