README.md

In KellyCahill/rfTSP: robust classification and feature selection for binary/multi outcomes

rfTSP

This is an R package for robust binary/multi classification and feature selection using gene expression profiles.

Required packages

• library(randomForest)
• library(doParallel)
• registerDoParallel()
• library(foreach)
• library(truncnorm)

Install this package from github

• in R console:

library(devtools) 
install_github("KellyCahill/rfTSP") 
library(rfTSP)

Example

``` Gnull=900 Gsig=100 Ntrain=c(70,130) Ntest=c(60,40) DAT = simu.multi(Gnull, Gsig, Ntrain, Ntest, MuShift = U[u], Clow = .2, Cup = .5, seed = 15232, label = c(-1,1)) train<-DAT$Xtrain test<-DAT$Xtest Y<-DAT$Ytrain true<-DAT$Ytest N<-length(Y) controls<-which(Y == "1") case<-which(Y == "2") ind<-t(combn(nrow(train), 2))

p<-c() for(j in 1:nrow(ind)){ p[j]<-binom_2proptest(j, data = train, N = N, controls = controls, case = case, ind = ind) }

k<-length(which(p < .005))

tsp<-getKtsp(train, Y, k)

dichotomized_train<-data_transform(train, tsp) dichotomized_test<-data_transform(test, tsp)

Assumes training and testing have the same genes in the same order

fit<-randomForest(as.factor(Y)~., data = dichotomized_train, ntree = 500, replace = FALSE) predict_label<-predict(fit, newdata = dichotomized_test, type = "response") error<-length(which(predict_label !=true ))/length(true)

}

KellyCahill/rfTSP documentation built on May 21, 2019, 1:41 p.m.