2D projections of RNA folding landscapes

Abstract

The analysis of RNA folding landscapes yields insights into the kinetic folding behavior not available from classical structure prediction methods. This is especially important for multi-stable RNAs whose function is related to structural changes, as in the case of riboswitches. However, exact methods such as barrier tree analysis scale exponentially with sequence length. Here we present an algorithm that computes a projection of the energy landscape into two dimensions, namely the distances to two reference structures. This yields an abstraction of the high-dimensional energy landscape that can be conveniently visualized, and can serve as the basis for estimating energy barriers and refolding pathways. With an asymptotic time complexity of O(n7) the algorithm is computationally demanding. However, by exploiting the sparsity of the dynamic programming matrices and parallelization for multi-core processors, our implementation is practical for sequences of up to 400 nt, which includes most RNAs of biological interest.