Prediction of siRNA Efficacy Using BP Neural Network and Support Vector Machine

Abstract

RNA interference (RNAi) is a mechanism for sequence-specific, post-transcriptional down-regulation of gene expression. The success of RNAi gene silencing depends on siRNA feature design. The shortcoming of previously reported methods which design siRNA sequences based on limited rules is that they are difficult to accurately predict the efficacy that a candidate siRNA sequence will silence the target gene. With validated siRNA databases have been developed in recent years, machine learning methods can be applied to predict siRNA accuracy and optimize design. This paper proposed a combined prediction method of BP neural network and support vector machine (SVM) for selecting effective siRNA sequences. With SVM, siRNA sequences were classified into effective or ineffective siRNAs. Subsequently, BP neural network model with great learning ability selected highly effective candidate sequences from effective siRNAs. We applied this method to published siRNAs datasets, and the experimental results confirmed good prediction capability.