Improved prediction of RNA secondary structure by integrating the free energy model with restraints derived from experimental probing data

Abstract

Recently, several experimental techniques have emerged for probing RNA structures based on high-throughput sequencing. However, most secondary structure prediction tools that incorporate probing data are designed and optimized for particular types of experiments. For example, RNAstructure-Fold is optimized for SHAPE data, while SeqFold is optimized for PARS data. Here, we report a new RNA secondary structure prediction method, restrained MaxExpect (RME), which can incorporate multiple types of experimental probing data and is based on a free energy model and an MEA (maximizing expected accuracy) algorithm. We first demonstrated that RME substantially improved secondary structure prediction with perfect restraints (base pair information of known structures). Next, we collected structure-probing data from diverse experiments (e.g. SHAPE, PARS and DMS-seq) and transformed them into a unified set of pairing probabilities with a posterior probabilistic model. By using the probability scores as restraints in RME, we compared its secondary structure prediction performance with two other well-known tools, RNAstructure-Fold (based on a free energy minimization algorithm) and SeqFold (based on a sampling algorithm). For SHAPE data, RME and RNAstructure-Fold performed better than SeqFold, because they markedly altered the energy model with the experimental restraints. For high-throughput data (e.g. PARS and DMS-seq) with lower probing efficiency, the secondary structure prediction performances of the tested tools were comparable, with performance improvements for only a portion of the tested RNAs. However, when the effects of tertiary structure and protein interactions were removed, RME showed the highest prediction accuracy in the DMS-accessible regions by incorporating in vivo DMS-seq data.

Characterizing RNA structures in vitro and in vivo with selective 2′-hydroxyl acylation analyzed by primer extension sequencing (SHAPE-Seq)

“…In the following steps we outline the process for restraining three different RNA folding algorithms with SHAPE-Seq reactivity data: RNAStructure [39,54], restrained MaxExpect (RME) [46], and the Washietl et al method (as part of the RNAprobing webserver) [47]. As discussed in Section 2.5, RNAStructure (containing Fold and ShapeKnots ) can calculate the MFE structure directly as well as generate an ensemble of structures (with the and commands).…”

“…RME uses SHAPE reactivities to modify the partition function and selects a structure from it that best matches the experimental data [46]. First, RME calculates a partition function after adding a pseudo-free energy term for each nucleotide i using: …”

“…Then, a predicted base pairing probability matrix is derived from the partition function that is then adjusted by the experimental data to account for discrepancies between the predicted base pairing matrix and the measured reactivities. This newly modified base pairing matrix is finally used to predict a structure that maximizes expected accuracy [46]. …”

Data-directed RNA secondary structure prediction using probabilistic modeling

“…This, although feasible, is a nontrivial task because there is no principled way to choose an optimal formula. Moreover, if one tries to combine information sources, such as data from different probes, RNAlin would either require multiple linear models or a more complex model, thus increasing the number of parameters Wu et al 2015). This would not only pose extra computational challenges for parameter optimization, but would also carry the risk of converging to local optima that are not globally optimal, as the multidimensional search space is not guaranteed to be convex.…”

“…), these techniques differ in the types of structural information they extract and in the statistical properties of the data they generate. Propelled by these experimental advances, prediction algorithms that incorporate probing data as soft constraints to direct predictions have recently emerged (Deigan et al 2009;Cordero et al 2012;Sükösd et al 2012;Washietl et al 2012;Zarringhalam et al 2012;Hajdin et al 2013;Ouyang et al 2013;Luntzer et al 2015;Wu et al 2015). While structure-probing data do not directly report pairing states of nucleotides (Sloma and Mathews 2015), they proved useful in improving the accuracy of MFE predictions (Deigan et al 2009;Hajdin et al 2013;Luntzer et al 2015).…”