test_predictor: Test accuracy of predictor on known phenotypes

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

Description

This function takes the expression data input to build_predictor() and the coefficient estimates from build_predictor() for phenotype prediction. The known phenotypes are also input for comparison and asseessment of predictor accuracy.

Usage

test_predictor(inputdata = NULL, phenodata = NULL, phenotype = NULL,
  covariates = NULL, type = "factor", predictordata = NULL)

Arguments

inputdata	output from filter_regions() inputdata
phenodata	data set with phenotype information; samples in rows, variables in columns phenodata
phenotype	phenotype of interest phenotype
covariates	Which covariates to include in model covariates
type	The class of the phenotype of interest (numeric, binary, factor) type
predictordata	object output from build_predictor predictordata

Value

list of actual and predicted phenotype, and summarization of output

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")

## select 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)

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