RNA Structure Prediction

Abstract

It is important to determine the structures of RNA molecules to understand their functions. An RNA secondary structure, which is the set of complementary base pairs, is often modeled by the Boltzmann distribution of free energy and by stochastic context-free grammar. RNA secondary structures are predicted as structures with minimum free energy structures, which are the maximum likelihood estimators, or as structures with maximum expected accuracy structures. Point estimation of the RNA secondary structure is often unreliable, but marginal probabilities, such as base-pairing probabilities, are useful. Recently, high-throughput experimental methods have been implemented to identify RNA base pairs, and computational methods that utilize those experimental results in predicting RNA secondary structures have been developed. The common structure prediction and the prediction of the joint structures of interacting RNAs are computationally more difficult, but several methods are available to approximately solve them. Computation for RNA secondary structures usually utilizes dynamic programming, but stochastic sampling and integer programming are used for computationally expensive problems. Predicting three dimensional structures is challenging, but modeling of non-Watson-Click interactions and fragment assembly have been applied.