Utilizing Selected Di- and Trinucleotides of siRNA to Predict RNAi Activity

Algorithm 1

The calculation process of threshold .

Input: A data set , where is the feature set extracted from siRNA

sequence and is the experimentally determined siRNA activities. The features of are first sorted

by the variable importance in descending order. The initial value of and are 1 and ,

respectively.

Output: optimal features .

  The dataset is divided into ten parts. Nine parts are used as the training set and the rest are used as

a testing set. We build a Random Forest model using the feature set and the training set and then

predict the testing siRNAs using the model. The correlation coefficient between the observed and predicted

siRNA activities is .

while and   do

  Calculate the prediction accuracy using according to the first step.

  If then

  else

  end if

 end while

  .