Using matching substructures as an optimization objective for RNA design

RNA design is a problem that has been shown to be NP-Hard. It is best described as the inverse of RNA folding. RNA folding describes the process of calculating the most likely secondary structure that a strand of nucleotides will fold into. Inversely, RNA design describes the process of designing a strand of nucleotides that will fold into a given secondary structure. The problem is made more difficult by the presence of a second objective, structural stability. Free energy is a measure of structural stability. In previous research, we have attempted to solve this problem using SIMARD (Simulated Annealing RNA Design). SIMARD employs a simulated annealing framework alongside a preselection strategy to design high-quality sequences in a reasonable amount of time. In this paper, we introduce the integration of BEAR (Brand nEw Alphabet for RNAs) to SIMARD as a way of notating secondary structures for quality evaluation. We attempt to design sequences with four different experimental configurations across two data sets. We find that representing our sequences with the BEAR grammar allows us to improve the average structural similarity of our generated sequences. We also find that SIMARD outperforms six other algorithms when running on the Eterna100 benchmark in terms of successfully designed structures.