predict_pheno: Makes predictions for selected phenotype

In leekgroup/phenopredict: Predicting phenotypes from RNA-seq expression data

Description

This function outputs predictions for phenotype of interest using selected regions upon which predictor has been built

Usage

predict_pheno(inputdata_test = NULL, phenodata = NULL, phenotype = NULL,
  type = c("factor", "numeric"), covariates = NULL, predictordata = NULL)

Arguments

inputdata_test	Object containing expression data from test set and corresponding GenomicRanges object expression
phenodata	data set with phenotype information; samples in rows, variables in columns phenodata
phenotype	phenotype of interest phenotype
type	The class of the phenotype of interest (numeric, factor)
covariates	Which covariates to include in model covariates
predictordata	object output from build_predictor predictordata

Value

A vector of predicted phenoyptes

Examples

library('GenomicRanges')
library('dplyr')

## Make up some some region data
regions <- GRanges(seqnames = 'chr2', IRanges(
     start = c(28971710:28971712, 29555081:29555083, 29754982:29754984),
     end = c(29462417:29462419, 29923338:29923340, 29917714:29917716)))

## make up some expression data for 9 rows and 30 people
data(sysdata, package='phenopredict')
## includes R object 'cm'
exp= cm[1:length(regions),1:30]

## generate some phenotype information
sex = as.data.frame(rep(c("male","female"),each=15))
age = as.data.frame(sample(1:100,30))
pheno = dplyr::bind_cols(sex,age)
colnames(pheno) <- c("sex","age")

## filter regions to be used to build the predictor
inputdata <- filter_regions(expression=exp, regiondata=regions,
	phenodata=pheno, phenotype="sex",
	covariates=NULL,type="factor", numRegions=2)

## build phenotype predictor
predictor<-build_predictor(inputdata=inputdata ,phenodata=pheno,
	phenotype="sex", covariates=NULL,type="factor", numRegions=2)

## determine resubstitution error
## carry out prediction in training data set
predictions_test<-test_predictor(inputdata=inputdata ,phenodata=pheno,
	phenotype="sex", covariates=NULL,type="factor",predictordata=predictor)

## generate new expressiondata set for prediction
exp_new= cm_new[1:length(regions),1:30]
## extract test data
test_data<-extract_data(newexpression=exp_new,
newregiondata=predictor$regiondata, predictordata=predictor)

## predict phenotype in test data
predictions <- predict_pheno(inputdata_test=test_data, phenodata=pheno,
phenotype="sex", type="factor" ,covariates=NULL, predictordata = predictor)

leekgroup/phenopredict documentation built on May 14, 2019, 11:27 a.m.