Efficient Methods for Selecting Sirna Sequences by Using the Average Silencing Probability and a Hidden Markov Model

Shigeru Takasaki

doi:10.4172/jcsb.1000137

Abstract

Short interfering RNA (siRNA) has been widely used for studying gene functions in mammalian cells but varies markedly in its gene silencing efficacy. Although many design rules/guidelines for effective siRNAs based on various criteria have been reported recently, there are only a few consistencies among them. This makes it difficult to select effective siRNA sequences in mammalian genes. This paper first clarifies problems of the recently reported siRNA design guidelines and then proposes a new method for selecting effective siRNA sequences from many possible candidates by using the average silencing probability on the basis of large number of known effective siRNAs. It is different from the previous score-based siRNA design techniques and can predict the probability that a candidate siRNA sequence will be effective. The results of evaluating it by applying it to recently reported effective and ineffective siRNA sequences for various genes indicate that it would be useful for many other genes. The evaluation results indicate that the proposed method would be useful for many other genes. It should therefore be useful for selecting siRNA sequences effective for mammalian genes. The paper also describes another method using a Hidden Markov Model (HMM) to select the optimal functional siRNAs.

Highlights

RNA interference (RNAi) silences gene expression by introducing double-stranded RNA homologous to the target mRNA
The effectiveness of the short interfering RNA responsible for RNA interference varies widely depending on the target sequence positions selected from the target gene [7,8]
As the key to Support Vector Machines (SVMs) success is to collect many useful features of effective short interfering RNA (siRNA) sequences, the usefulness of methods using SVMs may depend on the selected siRNAs

Summary

Introduction

RNA interference (RNAi) silences gene expression by introducing double-stranded RNA homologous to the target mRNA. As the key to SVM success is to collect many useful features of effective siRNA sequences, the usefulness of methods using SVMs may depend on the selected siRNAs. Holen [27] recently reported siRNA rules based on apparent overrepresentation or underrepresentation of certain nucleotides in certain positions of Novartis data set. Still another problem is that the previously reported methods cannot estimate the probability that a candidate siRNA will silence the target gene. Even if a high-scored siRNA were obtained using the reported methods, it would be difficult to estimate the probability that it would accomplish the expected gene degradation

Methods and Materials

Evaluation and model generation data

Results and Discussion

Evaluation for the HMM

Conclusions