JMIR Publications

Background:

The prediction of tertiary RNA structures is significant to the field of medicine (eg, messenger RNA [mRNA] vaccines, genome editing) and the exploration of viral transcripts. Though many RNA folding software programs exist, few studies have condensed their locus of attention solely to viral pseudoknotted RNA. These regulatory pseudoknots play a role in genome replication, gene expression, and protein synthesis.

Objective:

The objective of this study was to explore 5 RNA folding engines that compute either the minimum free energy (MFE) or the maximum expected accuracy (MEA), when applied to a specified suite of viral pseudoknotted RNAs that have been previously confirmed using mutagenesis, sequence comparison, structure probing, or nuclear magnetic resonance (NMR).

Methods:

The folding engines used in this study were tested against 26 experimentally derived short pseudoknotted sequences (20-150 nt) using metrics that are commonplace while testing software prediction accuracy: percentage error, mean squared error (MSE), sensitivity, positive predictive value (PPV), Youden’s index (J), and F₁-score. The data set used in this study was accrued from the Pseudobase++ database containing 398 RNAs, which was assessed using a set of inclusion and exclusion criteria following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines. Base pairings within a given RNA sequence were deemed correct or incorrect following Mathews’ parameters.

Results:

This paper reported RNA prediction engines with greater accuracy, such as pKiss, when compared to previous iterations of the software and when compared to older folding engines. This paper also reported that when assessed using metrics such as the F₁-score and the PPV, MEA folding software does not always outperform MFE folding software in prediction accuracy when applied to viral pseudoknotted RNA. Moreover, the results suggested that thermodynamic model parameters will not ensure accuracy if auxiliary parameters, such as Mg²⁺ binding, dangling end options, and hairpin-type penalties, are not applied.

Conclusions:

This is the first attempt at applying a suite of RNA folding engines to a data set solely comprising viral pseudoknotted RNA. The observations reported in this paper highlight the quality between different ab initio prediction methods, while enforcing the idea that a better understanding of intracellular thermodynamics is necessary for a more efficacious screening of RNAs.