MiDA (version 0.1.2)

MiSelectSignif: Select biological markers with high fold change and classification importance

Description

Choose probes which change is biologically significant based on binary classification feature importance, gene expression fold change and statistical significance.

Usage

MiSelectSignif(probes, mean1, mean2, FC.method, infl, stat.val,
  tresh.FC = 0.75, tresh.infl = 0.75, tresh.stat = 0.05)

Arguments

probes

character vector of probe (gene, transcript) names.

mean1

numeric vector of mean values for probes expression in the first group of comparison.

mean2

numeric vector of mean values for probes expression in the second group of comparison.

FC.method

character specifying the method of fold change counting. Possible values are: "absolute" (mean1-mean2), "percent" ((mean1*100/mean2)-100), "ratio" (mean1/ mean2), "Log2.ratio" (log2(mean1/mean2)).

infl

numeric vector of mean values for probes feature importance (relative influence) from binary classification.

stat.val

numeric vector of statistical significance (p-value, q-value) for testing differences of mean1 and mean2.

tresh.FC

numeric from o to 1 specifying the tresh hold for fold change FC parameters (quantile).The significant fold change is bigger than tresh.FC.

tresh.infl

numeric from o to 1 specifying the tresh hold for feature importance infl parameters (quantile).The significant feature importance is bigger than tresh.infl.

tresh.stat

numeric from o to 1 specifying the tresh hold for statistical significance stat.val.The significant fold change is lesser than tresh.stat.

Value

data frame of probe names, their fold change values, statistical significance values, feature inportance values and marker values.

Details

The order must be the same for all parameters.

This function marks as "markers" probes that statistically significant change their expression in two groups of comparison with high (over tresh hold) fold change and feature importance from binary classification.

Examples

# NOT RUN {
probes<-paste("probe", 1:50, sep="") #probes
mean1<-rnorm(50, mean=0, sd=1) #means
mean2<-rnorm(50, mean=5, sd=1)
infl<-c(1:50) # influence
stat.val<-rep(c(0.05, 0.04), c(20, 30))
Result<-MiSelectSignif(probes, mean1, mean2, FC.method="absolute", infl, stat.val,
                      tresh.FC=0.75, tresh.infl=0.75, tresh.stat=0.05)
Result[1:5,]

# }