DMLpredict: Predict new data from DML

In zwdzwd/sesame: SEnsible Step-wise Analysis of DNA MEthylation BeadChips

Predict new data from DML

Description

This function is also important for investigating factor interactions.

Usage

DMLpredict(betas, fm, pred = NULL, meta = NULL, BPPARAM = SerialParam())

Arguments

betas	beta values, matrix or SummarizedExperiment rows are probes and columns are samples.
fm	formula
pred	new data for prediction, useful for studying effect size. This argument is a data.frame to specify new data. If the argument is NULL, all combinations of all contrasts will be used as input. It might not work if there is a continuous variable input. One may need to explicitly provide the input in a data frame.
meta	data frame for sample information, column names are predictor variables (e.g., sex, age, treatment, tumor/normal etc) and are referenced in formula. Rows are samples. When the betas argument is a SummarizedExperiment object, this is ignored. colData(betas) will be used instead.
BPPARAM	number of cores for parallel processing, default to SerialParam() Use MulticoreParam(mc.cores) for parallel processing. For Windows, try DoparParam or SnowParam.

Value

a SummarizedExperiment of predictions. The colData describes the input of the prediction.

Examples

data <- sesameDataGet('HM450.76.TCGA.matched')

## use all contrasts as new input
res <- DMLpredict(data$betas[1:10,], ~type, meta=data$sampleInfo)

## specify new input
res <- DMLpredict(data$betas[1:10,], ~type, meta=data$sampleInfo,
  pred = data.frame(type=c("Normal","Tumour")))

## note that the prediction needs to be a factor of the same
## level structure as the original training data.
pred = data.frame(type=factor(c("Normal"), levels=c("Normal","Tumour")))
res <- DMLpredict(data$betas[1:10,], ~type,
  meta=data$sampleInfo, pred = pred)

zwdzwd/sesame documentation built on Nov. 11, 2024, 9:33 a.m.