Search Results

A few classes of RNA-RNA interaction (RRI) with complex roles in cellular functions, such as miRNA-target and lncRNAs, have already been studied. Accordingly, RRI bioinformatics tools proposed in the last decade are tailored for those specific classes. Interestingly, there are somewhat unnoticed mRNA-mRNA interactions in the literature with potentially drastic biological roles. Hence, there is a need for high-throughput generic RRI bioinformatics tools that can be used in more comprehensive settings. In this work, we revisit two of the RRI partition function algorithms, piRNA and rip. These are equivalent methods that implement the most comprehensive and computationally intensive thermodynamic model for RRI. We propose simpler models that are shown to retain the vast majority of the thermodynamic information that the more complex models capture. Specifically, we simplify the energy model by ignoring the system’s entropy and show its equivalency to a base-pair counting model. We allow different weights for base-pairs to maximize the correlations with the full thermodynamic model. Our newly developed algorithm, BPPart, is 225× faster than piRNA and is more expressive and easier to analyze due to its simplicity and order of magnitude reduction in the number of dynamic programming tables. Still, based on our analysis of both the real and randomly generated data, its scores achieve a correlation of 0.855 with piRNA at 37^{∘}C. Finally, we illustrate one use-case of such simpler models to generate hypotheses about the roles of specific RNAs in various diseases. We have made our tool publicly available and believe that this faster and more expressive model will make the incorporation of physics-guided information in complex RRI analysis and prediction models more accessible.