plasma (version 1.1.4)

plasma-class: Class "plasma"

Description

The plasma object class is returned after running the plasma function. The plasma function uses the PLSRCox components from one dataset as the predictor variables and the PLSRCox components of another dataset as the response variables to fit a partial least squares regression (plsr) model. Then, we take the mean of the predictions to create a final matrix of samples versus components.

The matrix of components described earlier is then used to fit a Cox Proportional Hazards (coxph) model with AIC stepwise variable selection to return a final object of class plasma which includes a coxph model with a reduced number of predictors.

Usage

plasma(object, multi)
# S4 method for plasma,missing
plot(x, y, ...)
# S4 method for plasma
barplot(height, source, n, direction = c("both", "up","down"),
                lhcol = c("cyan", "red"), wt = c("raw", "std"),  ...)
# S4 method for plasma
predict(object, newdata = NULL, type = c("components", "risk",
                 "split"), ...)

Value

The plasma function returns a newly constructed object of the plasma class. The plot method invisibly returns the object on which it was invoked. The predict method returns an object of the plasmaPredictions class.

Objects from the Class

Objects should be defined using the plasma function.

Arguments

multi

an object of the MultiplePLSCoxModels class.

object

an object of the plasma class.

height

an object of the plasma class for the barplot method.

x

an object of class plasma for plotting the Kaplan-Meier curves.

y

An ignored argrument for the plot method.

source

A length-one character vector; the name of a data set in a plasma object.

n

A length-one integer vector; the number of high-weight features to display.

direction

A length-one character vector; show features with positive weights (up), negative (down), or both.

lhcol

A chaacter vector of length 2, indicating the preferred colors for low (negative) or high (positive) weights.

wt

A character string indicating whether to plot raw weights or standardized weights.

newdata

A MultiOmics object with the same structure as the training data.

type

An enumerated character value.

...

Additional graphical parameters.

Slots

traindata:

An object of class MultiOmics used for training the model.

compModels:

A list containing objects in the form of plsr.

fullModel:

A coxph object with variables (components) selected via AIC stepwise selection.

Methods

plot:

Plots a Kaplan-Meier curve of the final coxph model that has been categorized into “low risk” and “high risk” based whether it is higher or lower, respectively, than the median value of risk.

predict:

creates an object of class plasmaPredictions.

barplot:

Produces a barplot of the n largest weights assigned to features from the appropriate data source.

Author

Kevin R. Coombes , Kyoko Yamaguchi

See Also

plasmaPredictions, plsr

Examples

fls <- try(loadESCAdata())
if (inherits(fls, "try-error")) {
  stop("Unable to load data from remote server.")
}
# restrict data set size
MO <- with(plasmaEnv, prepareMultiOmics(
    assemble[c("ClinicalBin", "ClinicalCont", "RPPA")], Outcome) )

splitVec <- with(plasmaEnv, rbinom(nrow(Outcome), 1, 0.6))
trainD <- MO[, splitVec == 1]
testD <- MO[, splitVec == 0]

firstPass <- fitCoxModels(trainD, "Days", "vital_status", "dead")
pl <- plasma(object = trainD, multi = firstPass)

plot(pl, legloc = "topright", main = "Training Data")
barplot(pl, "RPPA", 6)
barplot(pl, "RPPA", 10, "up")