Modeling, Simulations, and Bioinformatics at the Service of RNA Structure

Abstract

RNAs are the ultimate frontier of structural biology. They are large and complex molecules that can adopt complex structures displaying a wide variety of functions from carriers of genetic information to regulators of gene expression or even catalysis. Understanding 3D structure will be crucial to understanding RNA function and to creating RNAs with new functionalities. A revolution similar to that of protein engineering might be possible if enough structural information on RNA is obtained. Here, we analyze how the latest-generation theoretical models can contribute to such a revolution. Although chemically close to DNA, RNA can adopt a wide range of structures from regular helices to complex globular conformations, showing a complexity similar to that of proteins. The determination of the structure of RNA molecules, crucial for functional understanding, is severely handicapped by their size and flexibility, which makes the systematic use of experimental approaches difficult. Simulation techniques are also suffering from very severe problems related to the accuracy of the methods and their ability to sample a large and complex conformational landscape. Here, we systematically review recent approaches created to reduce the shortcoming of the current generation of simulation methods—from highly accurate models able to deal with small systems to coarse-grained approaches that are less accurate but applicable to dealing with large models. Although chemically close to DNA, RNA can adopt a wide range of structures from regular helices to complex globular conformations, showing a complexity similar to that of proteins. The determination of the structure of RNA molecules, crucial for functional understanding, is severely handicapped by their size and flexibility, which makes the systematic use of experimental approaches difficult. Simulation techniques are also suffering from very severe problems related to the accuracy of the methods and their ability to sample a large and complex conformational landscape. Here, we systematically review recent approaches created to reduce the shortcoming of the current generation of simulation methods—from highly accurate models able to deal with small systems to coarse-grained approaches that are less accurate but applicable to dealing with large models. Physics teaches us that ab initio quantum mechanics (QM) can represent with high accuracy any biomolecular system, among them RNA. Unfortunately, because of their computational cost, the practical application of ab initio QM formalisms to large systems such as RNA is often impossible. In fact, even simpler QM methods such as those based on density functional theory (DFT) fail to treat systems larger than 102 atoms, several orders of magnitude less than the size required to study RNAs in solution.1Smith L.G. Zhao J. Mathews D.H. Turner D.H. Physics-based all-atom modeling of RNA energetics and structure.Wiley Interdiscip. Rev. RNA. 2017; 8: e1422Crossref Scopus (5) Google Scholar Further simplifications of the basic QM formalism, such as those implicit in semiempirical (SE) methods, can extend the range of applicability of QM theory but at the expense of an expected loss of accuracy.2Šponer J. Krepl M. Banáš P. Kührová P. Zgarbová M. Jurečka P. Havrila M. Otyepka M. How to understand atomistic molecular dynamics simulations of RNA and protein-RNA complexes?.Wiley Interdiscip. Rev. RNA. 2017; 8: e1405Crossref Scopus (13) Google Scholar High-level QM and DFT calculations have had a central role in the development and validation of recent RNA force fields (FFs; see below). A recent example is the B97D3/AUG-CC-PVTZ study of the backbone and glycosidic torsions by Aytenfisu et al.,3Aytenfisu A.H. Spasic A. Grossfield A. Stern H.A. Mathews D.H. Revised RNA dihedral parameters for the amber force field improve RNA molecular dynamics.J. Chem. Theor. Comput. 2017; 13: 900-915Crossref PubMed Scopus (0) Google Scholar who highlighted systematic errors in current RNA FFs that might lead to incorrect molecular dynamics (MD) trajectories. Different conclusions were reached by the group led by Šponer, who again used DFT calculations as a reference, namely that the errors in current RNA FFs are related to imbalanced hydration and not to intrinsic errors in the classical gas phase Hamiltonian.4Szabla R. Havrila M. Kruse H. Šponer J. Comparative assessment of different RNA tetranucleotides from the DFT-D3 and force field perspective.J. Phys. Chem. B. 2016; 120: 10635-10648Crossref Scopus (7) Google Scholar The same group recently studied 46 different backbone conformations of the UpU dinucleotide step (see Figure 1) by using a variety of QM methods from the state-of-the-art CCSD(T) to the latest-generation SE algorithms,5Kruse H. Mladek A. Gkionis K. Hansen A. Grimme S. Šponer J. Quantum chemical benchmark study on 46 RNA backbone families using a dinucleotide unit.J. Chem. Theor. Comput. 2015; 11: 4972-4991Crossref PubMed Scopus (35) Google Scholar providing the community with an invaluable dataset for refinement of RNA FFs. Very recently our group used DFT/MM (density functional theory and molecular mechanics) calculations to fit some dihedrals directly for QM calculations in solution, opening a new approach to using DFT calculations in the refinement of RNA FF (see the next section).6Darré L. Ivani I. Dans P.D. Gómez H. Hospital A. Orozco M. Small details matter: the 2′-hydroxyl as a conformational switch in RNA.J. Am. Chem. Soc. 2016; 138: 16355-16363Crossref PubMed Scopus (6) Google Scholar Hobza and co-workers have led the use of high-level QM methods for the description of non-covalent interactions in biomolecules, including nucleobases. The latest contributions from the group include the construction of reference databases of interaction energies computed at very high QM level. These databases are very useful for the parameterization and validation of lower-level methods, including FFs. To obtain a deeper insight into this benchmark calculation, we recommend the last review published by this group.7Riley K.E. Pitoňák M. Jurečka P. Hobza P. Stabilization and structure calculations for noncovalent interactions in extended molecular systems based on wave function and density functional theories.Chem. Rev. 2010; 110: 5023-5063Crossref PubMed Scopus (533) Google Scholar, 8Řezáč J. Hobza P. Benchmark calculations of interaction energies in noncovalent complexes and their applications.Chem. Rev. 2016; 116: 5038-5071Crossref PubMed Scopus (101) Google Scholar On the other hand, focusing on the last couple of years we should cite the DFT study by Rypniewski et al.9Rypniewski W. Banaszak K. Kuliński T. Kiliszek A. Watson-Crick-like pairs in CCUG repeats: evidence for tautomeric shifts or protonation.RNA. 2016; 22: 22-31Crossref PubMed Scopus (6) Google Scholar of the C-U and U-U pairings, wherein the authors suggest that unusual tautomeric forms, or even anionic states of pyrimidines, can play a role in stabilizing certain forms of RNA. Similarly, Preethi et al.10Preethi S.P. Sharma P. Mitra A. Structural landscape of base pairs containing post-transcriptional modifications in RNA.RNA. 2017; 23: 847-859Crossref PubMed Scopus (1) Google Scholar, 11Preethi S.P. Sharma P. Mitra A. Higher order structures involving post transcriptionally modified nucleobases in RNA.RSC Adv. 2017; 7: 35694-35703Crossref Google Scholar used high-level DFT methods to study the impact of post-transcriptional modifications in the base pairings occurring in different RNA motives (interface, rRNA, and the intron-exon complexes). In another elegant study, Wilson et al.12Wilson K.A. Holland D.J. Wetmore S.D. Topology of RNA-protein nucleobase-amino acid π-π interactions and comparison to analogous DNA-protein π-π contacts.RNA. 2016; 22: 696-708Crossref PubMed Scopus (4) Google Scholar analyzed the 154 non-redundant RNA-protein π interactions observed in PDB with the M06-2X13Zhao Y. Truhlar D.G. .The M06 suite of density functionals for main group thermochemistry, thermochemical kinetics, noncovalent interactions, excited states, and transition elements: two new functionals and systematic testing of four M06-class functionals and 12 other functionals.Theor. Chem. Acc. 2008; 120: 215-241Crossref Scopus (0) Google Scholar functional, finding that these π-π interactions provide huge stabilization to the protein-RNA complex. Unusual interactions affecting RNA nucleobases have been also studied by means of high-level QM theory. For example, Chawla et al.14Chawla M. Chermak E. Zhang Q. Bujnicki J.M. Oliva R. Cavallo L. Occurrence and stability of lone pair-π stacking interactions between ribose and nucleobases in functional RNAs.Nucleic Acids Res. 2017; 45: 11019-11032Crossref PubMed Scopus (5) Google Scholar used high-level QM theory (up to CCSD(T)) to explore the interactions between the O4′ atom and the π cloud of the nucleobase, finding that this apparently irrelevant interaction can, in fact, significantly affect the packing of RNA. Cation-RNA interaction has been another traditional field for the use of QM theory. A recent example of this family of studies was published by Casalino et al.,15Casalino L. Palermo G. Abdurakhmonova N. Rothlisberger U. Magistrato A. Development of site-specific Mg2+-RNA force field parameters: a dream or reality? Guidelines from combined molecular dynamics and quantum mechanics simulations.J. Chem. Theor. Comput. 2017; 13: 340-352Crossref PubMed Scopus (10) Google Scholar who used DFT calculations to study typical Mg2+-RNA binding motifs, providing a useful benchmark set to develop next-generation Mg2+-adapted FFs. Catalysis is another field where the use of QM is strictly necessary, since it typically implies the restructuration of bonds and electronic effects that classical FFs cannot handle. In that sense, QM/MM (quantum mechanics and molecular mechanics) calculations have become the standard approach for the study of ribozymes.16Bertran J. Oliva A. Ribozymes.in: Tuñón I. Moliner V. Simulating Enzyme Reactivity: Computational Methods in Enzyme Catalysis. The Royal Society of Chemistry, 2017: 404-435Google Scholar, 17Dubecký M. Walter N.G. Šponer J. Otyepka M. Banáš P. Chemical feasibility of the general acid/base mechanism of glmS ribozyme self-cleavage.Biopolymers. 2015; 103: 550-562Crossref PubMed Scopus (4) Google Scholar, 18Świderek K. Marti S. Tuñón I. Moliner V. Bertran J. Peptide bond formation mechanism catalyzed by ribosome.J. Am. Chem. Soc. 2015; 137: 12024-12034Crossref PubMed Scopus (13) Google Scholar, 19Zhang S. Ganguly A. Goyal P. Bingaman J.L. Bevilacqua P.C. Hammes-Schiffer S. Role of the active site guanine in the glmS ribozyme self-cleavage mechanism: quantum mechanical/molecular mechanical free energy simulations.J. Am. Chem. Soc. 2015; 137: 784-798Crossref PubMed Scopus (36) Google Scholar, 20Casalino L. Palermo G. Rothlisberger U. Magistrato A. Who activates the nucleophile in ribozyme catalysis? An answer from the splicing mechanism of group II introns.J. Am. Chem. Soc. 2016; 138: 10374-10377Crossref PubMed Scopus (14) Google Scholar, 21Chen H. Giese T.J. Golden B.L. York D.M. Divalent metal ion activation of a guanine general base in the hammerhead ribozyme: insights from molecular simulations.Biochemistry. 2017; 56: 2985-2994Crossref PubMed Scopus (9) Google Scholar, 22Lee T.-S. Radak B.K. Harris M.E. York D.M. A two-metal-ion-mediated conformational switching pathway for HDV ribozyme activation.ACS Catal. 2016; 6: 1853-1869Crossref PubMed Scopus (11) Google Scholar, 23Mlýnský V. Walter N.G. Šponer J. Otyepka M. Banáš P. The role of an active site Mg2+ in HDV ribozyme self-cleavage: insights from QM/MM calculations.Phys. Chem. Chem. Phys. 2015; 17: 670-679Crossref PubMed Google Scholar, 24Radak B.K. Lee T.-S. Harris M.E. York D.M. Assessment of metal-assisted nucleophile activation in the hepatitis delta virus ribozyme from molecular simulation and 3D-RISM.RNA. 2015; 21: 1566-1577Crossref PubMed Scopus (10) Google Scholar, 25Thaplyal P. Ganguly A. Hammes-Schiffer S. Bevilacqua P.C. Inverse thio effects in the hepatitis delta virus ribozyme reveal that the reaction pathway is controlled by metal ion charge density.Biochemistry. 2015; 54: 2160-2175Crossref PubMed Google Scholar, 26Zhang S. Stevens D.R. Goyal P. Bingaman J.L. Bevilacqua P.C. Hammes-Schiffer S. Assessing the potential effects of active site Mg2+ ions in the glmS ribozyme-cofactor complex.J. Phys. Chem. Lett. 2016; 7: 3984-3988Crossref PubMed Scopus (3) Google Scholar, 27Bingaman J.L. Zhang S. Stevens D.R. Yennawar N.H. Hammes-Schiffer S. Bevilacqua P.C. The GlcN6P cofactor plays multiple catalytic roles in the glmS ribozyme.Nat. Chem. Biol. 2017; 13: 439-445Crossref PubMed Scopus (14) Google Scholar Several of these studies focused on the role of metallic ions or cofactors in catalysis.21Chen H. Giese T.J. Golden B.L. York D.M. Divalent metal ion activation of a guanine general base in the hammerhead ribozyme: insights from molecular simulations.Biochemistry. 2017; 56: 2985-2994Crossref PubMed Scopus (9) Google Scholar, 22Lee T.-S. Radak B.K. Harris M.E. York D.M. A two-metal-ion-mediated conformational switching pathway for HDV ribozyme activation.ACS Catal. 2016; 6: 1853-1869Crossref PubMed Scopus (11) Google Scholar, 23Mlýnský V. Walter N.G. Šponer J. Otyepka M. Banáš P. The role of an active site Mg2+ in HDV ribozyme self-cleavage: insights from QM/MM calculations.Phys. Chem. Chem. Phys. 2015; 17: 670-679Crossref PubMed Google Scholar, 24Radak B.K. Lee T.-S. Harris M.E. York D.M. Assessment of metal-assisted nucleophile activation in the hepatitis delta virus ribozyme from molecular simulation and 3D-RISM.RNA. 2015; 21: 1566-1577Crossref PubMed Scopus (10) Google Scholar, 25Thaplyal P. Ganguly A. Hammes-Schiffer S. Bevilacqua P.C. Inverse thio effects in the hepatitis delta virus ribozyme reveal that the reaction pathway is controlled by metal ion charge density.Biochemistry. 2015; 54: 2160-2175Crossref PubMed Google Scholar, 26Zhang S. Stevens D.R. Goyal P. Bingaman J.L. Bevilacqua P.C. Hammes-Schiffer S. Assessing the potential effects of active site Mg2+ ions in the glmS ribozyme-cofactor complex.J. Phys. Chem. Lett. 2016; 7: 3984-3988Crossref PubMed Scopus (3) Google Scholar, 27Bingaman J.L. Zhang S. Stevens D.R. Yennawar N.H. Hammes-Schiffer S. Bevilacqua P.C. The GlcN6P cofactor plays multiple catalytic roles in the glmS ribozyme.Nat. Chem. Biol. 2017; 13: 439-445Crossref PubMed Scopus (14) Google Scholar Other studies focused on the use of QM methods to understand complex experiments such as those based on the measurement of kinetic isotope effects.28Chen H. Piccirilli J.A. Harris M.E. York D.M. Effect of Zn2+ binding and enzyme active site on the transition state for RNA 2′-O-transphosphorylation interpreted through kinetic isotope effects.Biochim. Biophys. Acta. 2015; 1854: 1795-1800Crossref PubMed Scopus (10) Google Scholar Very interestingly, other non-ribozyme RNAs can also exhibit catalytic properties, as has been described for unpaired nucleotides in non-catalytic RNAs.29Mlýnský V. Bussi G. Understanding in-line probing experiments by modeling cleavage of nonreactive RNA nucleotides.RNA. 2017; 23: 712-720Crossref PubMed Scopus (4) Google Scholar Mlýnský and Bussi29Mlýnský V. Bussi G. Understanding in-line probing experiments by modeling cleavage of nonreactive RNA nucleotides.RNA. 2017; 23: 712-720Crossref PubMed Scopus (4) Google Scholar published a study whereby they used QM(DFTB330Gaus M. Cui Q. Elstner M. DFTB3: extension of the self-consistent-charge density-functional tight-binding method (SCC-DFTB).J. Chem. Theor. Comput. 2012; 7: 931-948Crossref PubMed Scopus (0) Google Scholar)/MM calculations in the context of enhanced sampling methods characterized the pattern of reactivity of specific RNA motifs such as the uGAAAg tetraloop. It is difficult to predict the impact of QM calculations on RNA in the next decade, but the expectation exists that a new generation of QM methods using new partitioning schemes would allow the representation of more realistic segments of RNA. For example, Jin et al.31Jin X. Zhang J.Z.H. He X. Full QM calculation of RNA energy using electrostatically embedded generalized molecular fractionation with conjugate caps method.J. Phys. Chem. A. 2017; 121: 2503-2514Crossref PubMed Scopus (5) Google Scholar have shown good representations of 15-mer RNAs using an electrostatically embedded generalized molecular fractionation with the conjugated caps (i.e., EE-GMFCC) method. Additional expectations arise from the generation of simplified QM approaches, such as density functional tight binding (DFTB) or RNA-adapted SE approaches. As an example, Huang et al.32Huang M. Dissanayake T. Kuechler E. Radak B.K. Lee T.-S. Giese T.J. York D.M. A multidimensional B-spline correction for accurate modeling sugar puckering in QM/MM simulations.J. Chem. Theor. Comput. 2017; 13: 3975-3984Crossref PubMed Scopus (0) Google Scholar recently introduced a multidimensional B-spline correction map (BMAP) to the sugar puckering in the AM1/d-PhoT33Nam K. Cui Q. Gao J. York D.M. .Specific reaction parameterization of the AM1/d Hamiltonian for phosphoryl transfer reactions: H, O, and P atoms.J. Chem. Theor. Comput. 2007; 3: 486-504Crossref PubMed Scopus (96) Google Scholar semiempirical Hamiltonian, which was successfully applied to reproduce different RNA transesterification reactions. Similarly, York's group has presented exciting results on RNA systems using quantum mechanical force fields (QMFFs),34Giese T.J. York D.M. Quantum mechanical force fields for condensed phase molecular simulations.J. Phys. Condens. Matter. 2017; 29: 383002Crossref PubMed Scopus (3) Google Scholar which scale linearly with system size and are then much faster than fully coupled QM methods. Despite recent advances in computers and simulation tools, there is no expectation that QM methods will be able to deal (at least in the next decades) with even medium-size (102 nt) RNAs in solution. This has fueled the development of MM methods (such as MD) based on atomistic classical FFs. The most popular RNA FFs are those originating from the AMBER (Assisted Model Building with Energy Refinement) and CHARMM (Chemistry at Harvard Macromolecular Mechanics) communities. Although both rely grossly on the same formulation of the potential energy functional, they differ in the parameterization strategy (see Vangaveti et al.35Vangaveti S. Ranganathan S.V. Chen A.A. Advances in RNA molecular dynamics: a simulator’s guide to RNA force fields: advances in RNA molecular dynamics.Wiley Interdiscip. Rev. RNA. 2017; 8: e1396Crossref Scopus (8) Google Scholar and references therein). The four decades of healthy competition between CHARMM and AMBER developers have promoted a dramatic advance in the field. Nevertheless, it would be highly desirable that other communities also join the race, which seems to be the case for the OPLS (Optimized Potentials for Liquid Simulations)—one of which has recently published a careful calibration of the torsions of nucleosides and nucleotides from high-level DFT calculations and nuclear magnetic resonance (NMR) observables.36Robertson M.J. Tirado-Rives J. Jorgensen W.L. Improved treatment of nucleosides and nucleotides in the OPLS-AA force field.Chem. Phys. Lett. 2017; 683: 276-280Crossref PubMed Scopus (2) Google Scholar During the last decades, the evolution of both AMBER and CHARMM RNA FFs has been fueled by the ever-growing computational power, pushing the time-length boundaries of MD trajectories. The extension of trajectories has made evident errors not visible in shorter simulations. For example, in the case of the AMBER community (Figure 2) the 94 and 99 FFs (AMBER-ff94 and AMBER-ff99) were used for almost two decades, until significant errors emerged in long-scale simulations. This forced the development of new parameters aimed mainly at refining specific torsion angles (AMBER-ff99-BSC0-χOL3,37Pérez A. Marchán I. Svozil D. Sponer J. Cheatham T.E. Laughton C.A. Orozco M. Refinement of the AMBER force field for nucleic acids: improving the description of α/γ conformers.Biophys. J. 2007; 92: 3817-3829Abstract Full Text Full Text PDF PubMed Scopus (1295) Google Scholar, 38Zgarbová M. Otyepka M. Šponer J. Mládek A. Banáš P. Cheatham T.E. Jurečka P. Refinement of the Cornell et al. nucleic acids force field based on reference quantum chemical calculations of glycosidic torsion profiles.J. Chem. Theor. Comput. 2011; 7: 2886-2902Crossref PubMed Scopus (0) Google Scholar AMBER-ff99-χYIL,39Yildirim I. Stern H.A. Kennedy S.D. Tubbs J.D. Turner D.H. Reparameterization of RNA χ torsion parameters for the AMBER force field and comparison to NMR spectra for cytidine and uridine.J. Chem. Theor. Comput. 2010; 6: 1520-1531Crossref PubMed Scopus (0) Google Scholar and AMBER-ff99-TOR40Yildirim I. Kennedy S.D. Stern H.A. Hart J.M. Kierzek R. Turner D.H. Revision of AMBER torsional parameters for RNA improves free energy predictions for tetramer duplexes with GC and iGiC base pairs.J. Chem. Theor. Comput. 2012; 8: 172-181Crossref PubMed Scopus (0) Google Scholar), and certain non-bonded terms.41Chen A.A. Garcia A.E. High-resolution reversible folding of hyperstable RNA tetraloops using molecular dynamics simulations.Proc. Natl. Acad. Sci. U S A. 2013; 110: 16820-16825Crossref PubMed Scopus (0) Google Scholar The CHARMM community has mainly focused on the representation of proteins and lipids for many years, and the evolution of the nucleic acids version has been slower (Figure 2). CHARMM36 RNA parameters42Denning E.J. Priyakumar U.D. Nilsson L. Mackerell A.D. Impact of 2′-hydroxyl sampling on the conformational properties of RNA: update of the CHARMM all-atom additive force field for RNA.J. Comput. Chem. 2011; 32: 1929-1943Crossref PubMed Scopus (168) Google Scholar were a major advance for this community, as it corrected major caveats of previous versions (CHARMM2243MacKerell A.D. Wiorkiewicz-Kuczera J. Karplus M. An all-atom empirical energy function for the simulation of nucleic acids.J. Am. Chem. Soc. 1995; 117: 11946-11975Crossref Google Scholar and CHARMM2744MacKerell A.D. Banavali N. Foloppe N. Development and current status of the CHARMM force field for nucleic acids.Biopolymers. 2000; 56: 257-265Crossref PubMed Scopus (654) Google Scholar), allowing more reliable simulations of different types of RNAs. Worth mentioning is that both AMBER and CHARMM FFs have been extended to account for non-coding nucleotides (more than 100 variants are available),45Aduri R. Psciuk B.T. Saro P. Taniga H. Schlegel H.B. SantaLucia J. AMBER force field parameters for the naturally occurring modified nucleosides in RNA.J. Chem. Theor. Comput. 2007; 3: 1464-1475Crossref PubMed Scopus (0) Google Scholar, 46Xu Y. Vanommeslaeghe K. Aleksandrov A. MacKerell A.D. Nilsson L. Additive CHARMM force field for naturally occurring modified ribonucleotides: CHARMM potential energy function.J. Comput. Chem. 2016; 37: 896-912Crossref PubMed Scopus (14) Google Scholar opening the possibility to study epigenetic changes in RNA and extending FF-based calculations to the study of non-natural nucleic acids. Despite the titanic efforts of the CHARMM and AMBER communities, several errors in FF-based simulations still persist, for example, high populations of non-native stacking conformations47Bergonzo C. Henriksen N.M. Roe D.R. Cheatham T.E. Highly sampled tetranucleotide and tetraloop motifs enable evaluation of common RNA force fields.RNA. 2015; 21: 1578-1590Crossref PubMed Scopus (57) Google Scholar or significant thermodynamic imbalances between the folded and unfolded states.48Bottaro S. Banáš P. Šponer J. Bussi G. Free energy landscape of GAGA and UUCG RNA tetraloops.J. Phys. Chem. Lett. 2016; 7: 4032-4038Crossref PubMed Scopus (29) Google Scholar These difficulties might point toward fundamental problems in current RNA FFs, for example, unbalanced π stacking49Schrodt M.V. Andrews C.T. Elcock A.H. Large-scale analysis of 48 DNA and 48 RNA tetranucleotides studied by 1 μs explicit-solvent molecular dynamics simulations.J. Chem. Theor. Comput. 2015; 11: 5906-5917Crossref PubMed Scopus (10) Google Scholar and/or hydrogen bond interactions,47Bergonzo C. Henriksen N.M. Roe D.R. Cheatham T.E. Highly sampled tetranucleotide and tetraloop motifs enable evaluation of common RNA force fields.RNA. 2015; 21: 1578-1590Crossref PubMed Scopus (57) Google Scholar, 50Bergonzo C. Cheatham T.E. Improved force field parameters lead to a better description of RNA structure.J. Chem. Theor. Comput. 2015; 11: 3969-3972Crossref PubMed Scopus (0) Google Scholar improper hydration of RNA functional groups,50Bergonzo C. Cheatham T.E. Improved force field parameters lead to a better description of RNA structure.J. Chem. Theor. Comput. 2015; 11: 3969-3972Crossref PubMed Scopus (0) Google Scholar or even fundamental problems in the pairwise additive potential formalism. This has boosted yet another round of parameterization efforts (highlighted in gray on the timeline of Figure 2), using in some cases renewed methodological approaches. The field is now in an exciting, but also turbulent phase, and it is not trivial, even for an expert, to decide the combination of patches to add to the default FFs. We will use the next paragraphs to provide the reader with some clues for selecting the best FF for his or her particular problem. Improving the water model is one possible way to change the balance between hydrogen bonding and stacking of the bases and to correct some of the known caveats of current RNA FFs. Advances in this direction were presented in 2015 by Bergonzo and Cheatham,50Bergonzo C. Cheatham T.E. Improved force field parameters lead to a better description of RNA structure.J. Chem. Theor. Comput. 2015; 11: 3969-3972Crossref PubMed Scopus (0) Google Scholar who combined AMBER-ff99-BSC0-χOL3 FF with four different water models (TIP3P,51Jorgensen W.L. Chandrasekhar J. Madura J.D. Impey R.W. Klein M.L. Comparison of simple potential functions for simulating liquid water.J. Chem. Phys. 1983; 79: 926-935Crossref Google Scholar SPC/E,52Berendsen H.J.C. Grigera J.R. Straatsma T.P. The missing term in effective pair potentials.J. Phys. Chem. 1987; 91: 6269-6271Crossref Google Scholar TIP4P-Ew,53Horn H.W. Swope W.C. Pitera J.W. Madura J.D. Dick T.J. Hura G.L. Head-Gordon T. Development of an improved four-site water model for biomolecular simulations: TIP4P-Ew.J. Chem. Phys. 2004; 120: 9665-9678Crossref PubMed Scopus (1038) Google Scholar and OPC54Izadi S. Anandakrishnan R. Onufriev A.V. Building water models: a different approach.J. Phys. Chem. Lett. 2014; 5: 3863-3871Crossref PubMed Scopus (291) Google Scholar) to study the conformational landscape of the rGACC tetramer. The standard water model in CHARMM is TIP3PCHARMM (TIP3P modified to include non-zero LJ terms for hydrogen atoms), which has been explicitly used in the FF charge parameterization. As a consequence, CHARMM-based simulations are, in principle, restricted to the TIP3PCHARMM model. However, recent efforts in protein FF development55Huang J. Rauscher S. Nawrocki G. Ran T. Feig M. de Groot B.L. Grubmüller H. MacKerell A.D. CHARMM36m: an improved force field for folded and intrinsically disordered proteins.Nat. Methods. 2017; 14: 71-73Crossref PubMed Scopus (241) Google Scholar suggest that changing off-diagonal LJ interactions (mainly the dispersion component) between water hydrogen atoms and the protein (without affecting water-water interactions) improves the protein compaction and folding compared with experimental data. These results suggest a possible pathway for RNA FF improvement focused on the water model. In any case, the modification of the water model is always a risky decision, as current water models have been validated in thousands of studies (just the TIP3P model collects more than 24,000 citations). An RNA-tuned water model might lead to very strong links between RNA FF and water model and to potential problems in the transferability of the resulting water model. A further step to improve nucleic acid interactions was the modification of the LJ parameters of charged phosphates. On the basis of the work of Steinbrecher et al.56Steinbrecher T. Latzer J. Case D.A. Revised AMBER parameters for bioorganic phosphates.J. Chem. Theor. Comput. 2012; 8: 4405-4412Crossref PubMed Scopus (0) Google Scholar on bio-organic phosphates, a ∼5% increase of the van der Waals (vdW) radii in RNA phosphate oxygen atoms (OP1/2, O5′, and O3′) was proposed, which (when combined with the OPC model) improved the representation of the conformational ensemble of the rGACC tetramer.50Bergonzo C. Cheatham T.E. Improved force field parameters lead to a better description of RNA structure.J. Chem. Theor. Comput. 2015; 11: 3969-3972Crossref PubMed Scopus (0) Google Scholar Unfortunately, the improvement was not transferable to rCCCC.50Bergonzo C. Cheatham T.E. Improved force field parameters lead to a better description of RNA structure.J. Chem. Theor. Comput. 2015; 11: 3969-3972Crossref PubMed Scopus (0) Google Scholar Pak and co-workers suggested that the previous correction (named vdWbb) could artifactually weaken phosphate hydration and developed an alternative LJ correction (vdWYP) based on differential pair-dependent Lorentz-Berthelot combination rules.57Yang C. Lim M. Kim E. Pak Y. Predicting RNA structures via a simple van der Waals correction to an all-atom force field.J. Chem. Theor. Comput. 2017; 13: 395-399Crossref PubMed Scopus (0) Google Scholar More precisely, the vdW radii of the OP1/2 phosphate atoms and the O2′ were scaled up by 5%, but only for intramolecular interactions, whereas the original unscaled vdW radii were used for the interaction with water. This parameterization, combined with the OPC water model, was tested in four tetramers: rGA