A Randomized String Kernel and Its Application to RNA Interference

Abstract

String kernels directly model sequence similarities without the necessity of extracting numerical features in a vector space. Since they better capture complex traits in the sequences, string kernels often achieve better pre-diction performance. RNA interference is an important biological mechanism with many therapeutical applica-tions, where strings can be used to represent target mes-senger RNAs and initiating short RNAs and string ker-nels can be applied for learning and prediction. How-ever, existing string kernels are not particularly devel-oped for RNA applications. Moreover, most existing string kernels are n-gram based and suffer from high dimensionality and inability of preserving subsequence orderings. We propose a randomized string kernel for use with support vector regression with a purpose of better predicting silencing efficacy scores for the candi-date sequences and eventually improving the efficiency of biological experiments. We show the positive defi-niteness of this kernel and give an analysis of random-ization error rates. Empirical results on biological data demonstrate that the proposed kernel performed better than existing string kernels and achieved significant im-provements over kernels computed from numerical de-scriptors extracted according to structural and thermo-dynamic rules. In addition, it is computationally more efficient.