Molecular Dynamics Simulations Accurately Recreate the RNA Folding Process

Ribonucleic acid (RNA) is one of life's most versatile molecules, with roles going far beyond being a messenger of genetic code, as it is fundamentally involved in gene regulation, processing, and maintenance across all living systems. This versatility is deeply tied to RNA's ability to adopt complex three-dimensional shapes, known as secondary and tertiary structures. With the global rise of RNA-based therapeutics, understanding and precisely predicting secondary and tertiary structures is essential to fully harness the RNA's potential in biotechnology.

Despite the need to model these structures, simulating the complete folding process of an RNA molecule from an initially unfolded state remains a challenge in computational biophysics. These molecular dynamics (MD) simulations involve many physics-based computational potentials (or 'force fields') and special methods to adjust to different folding timescales. Historically, success in this area has been limited. Although many different modeling approaches have been tested, prior MD simulation studies only reported the accurate folding of a few, simple stem-loop structures.

Against this backdrop, Associate Professor Tadashi Ando from the Department of Applied Electronics, Tokyo University of Science, Japan, conducted a large-scale, rigorous evaluation of modern simulation methods. In his latest paper published in the journal ACS Omega on October 26, 2025, he explored whether a combination of state-of-the-art computational tools could reliably model the fundamental folding process of a significantly larger and more structurally diverse library of RNA stem loops than ever tested before.

The study employed conventional MD simulations using two cutting-edge computational components: the DESRES-RNA atomistic force field (a potential function refined for highly accurate RNA simulations) and the GB-neck2 implicit solvent model. This implicit solvent model is the key, as it approximates the surrounding liquid as a continuous medium instead of actual molecules. This dramatically accelerated the rate of conformational sampling compared to explicit solvent models, which are much more computationally demanding. Dr. Ando applied this methodology to a varied set of 26 RNA stem-loops, ranging from 10 to 36 nucleotides in length and structures featuring bulges and internal loops. Worth noting, all simulations started from fully extended, unfolded conformations.

The results demonstrated a high degree of folding stability and accuracy, validating the predictive power of the combined DESRES-RNA and GB-neck2 tools. A total of 23 out of 26 RNA molecules successfully folded into the expected structures. For the 18 simpler stem loops, this folding was achieved with exceptional accuracy, showing root mean square deviation (RMSD) values-a measure of deviation from the known experimental structure-of less than 2 Å for the stem and less than 5 Å for the molecules.

Even among the more challenging motifs, five of the eight containing bulges or internal loops were successfully achieved. For these complex structures, the simulations also revealed a distinct folding pathway. "Previous studies have reported only two or three folding examples of simple stem-loop structures of approximately 10 residues, whereas this study deals with structures of much greater scale and complexity," highlights Dr. Ando.

The successful simulation of folding across such a broad and complex library represents a significant milestone in computational biology. It rigorously validates the chosen combination of force field and solvent models as a reliable starting point for investigating large-scale conformational changes in RNA. The study also unveiled directions for future refinement of these tools. While stem regions were highly accurate, the loop regions showed less accuracy, with RMSD values of approximately 4 Å. This indicates that the implicit solvent model parameters need optimization, particularly for modeling non-canonical base pairs and incorporating the critical effects of divalent cations like magnesium.

Overall, this study represents an essential step to improve available RNA-based medical tools. Understanding how RNA folds is crucial when designing drugs that can target it, which could lead to new treatments for genetic disorders, viral infections such as COVID-19 or influenza, and certain cancers.

"The ability to reproduce the overall folding of the basic stem-loop structure is an important milestone in understanding and predicting the structure, dynamics, and function of RNA using highly accurate and reliable computational models," concludes Dr. Ando. "I expect this computational technique to lead to applications in RNA molecule design and drug discovery."