Fitting Corrections to an RNA Force Field Using Experimental Data

Cesari, Andrea; Bottaro, Sandro; Lindorff‐Larsen, Kresten; Banáš, Pavel; Šponer, Jiřı́; Bussi, Giovanni

doi:10.1021/acs.jctc.9b00206

Cited by 66 publications

(119 citation statements)

References 52 publications

Supporting

Mentioning

116

Contrasting

Unclassified

Order By: Relevance

“…12 However, the simulations do indicate that in some cases finer optimization could have visible influence on the results. We made an initial attempt to refine simultaneously all the parameters of the gHBfix potential by using the reweighting scheme proposed in Ref 55 . We used the set of fourteen available REST2 simulations of r(GACC) TN from our previous work, where different bias to base-base and SPh interactions was applied.…”

Section: Intercalated States Can Be Suppressed By Penalizing Sph Intementioning

confidence: 99%

Fine-tuning of the AMBER RNA Force Field with a New Term Adjusting Interactions of Terminal Nucleotides

Mlýnský

Kührová

Kuhr

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Determination of RNA structural-dynamic properties is challenging for experimental methods. Thus atomistic molecular dynamics (MD) simulations represent a helpful technique complementary to experiments. However, contemporary MD methods still suffer from limitations of force fields (ffs), including imbalances in the non-bonded ff terms. We have recently demonstrated that some improvement of state-of-the-art AMBER RNA ff can be achieved by adding a new term for H-bonding called gHBfix, which increases tuning flexibility and reduces the risk of side-effects. Still, the first gHBfix version did not fully correct simulations of short RNA tetranucleotides (TNs). TNs are key benchmark systems due to availability of unique NMR data, although giving too much weight on improving TN simulations can easily lead to over-fitting to A-form RNA. Here we combine the gHBfix version with another term called tHBfix, which separately treats H-bond interactions formed by terminal nucleotides. This allows to refine simulations of RNA TNs without affecting simulations of other RNAs. The approach is in line with adopted strategy of current RNA ffs, where the terminal nucleotides possess different parameters for the terminal atoms than the internal nucleotides. The combination of gHBfix with tHBfix significantly improves the behavior of RNA TNs during well-converged enhanced-sampling simulations. TNs mostly populate canonical A-form like states while spurious intercalated structures are largely suppressed. Still, simulations of r(AAAA) and r(UUUU) TNs show some residual discrepancies with the primary NMR data which suggests that future tuning of some other ff terms might be useful.

show abstract

Section: Intercalated States Can Be Suppressed By Penalizing Sph Intementioning

confidence: 99%

Fine-tuning of the AMBER RNA Force Field with a New Term Adjusting Interactions of Terminal Nucleotides

Mlýnský

Kührová

Kuhr

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…determine an optimal value of θ. A similar idea was used by Cesari et al to find the optimal balance between the distribution arising from the force field and the experimental data [53]. Here, we put forward the following procedure:…”

Section: Resultsmentioning

confidence: 99%

“…We would like to end this discussion by reminding the reader that the BME and related approaches are particularly suited to avoid overfitting. Maximization of the entropy implies that the reweighted distribution is minimally perturbed, unlike in other reweighting approaches [45,46,[46][47][48][49][49][50][51][52][53]. One may think that still, small values of θ, which means a strong confidence on the experimental data, would lead to an overfit of the finite number of computed structures.…”

Section: Discussionmentioning

confidence: 99%

Bayesian-Maximum-Entropy reweighting of IDP ensembles based on NMR chemical shifts

Crehuet¹,

Jorro

Lindorff‐Larsen

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

Bayesian and Maximum Entropy approaches allow for a statistically sound and systematic fitting of experimental and computational data. Unfortunately, assessing the relative confidence in these two types of data remains difficult as several steps add unknown error. Here we propose the use of a validation-set method to determine the balance, and thus the amount of fitting. We apply the method to synthetic NMR chemical shift data of an intrinsically disordered protein. We show that the method gives consistent results even when other methods to assess the amount of fitting cannot be applied. Finally, we also describe how the errors in the chemical shift predictor can lead to an incorrect fitting and how using secondary chemical shifts could alleviate this problem.

show abstract

“…NMR data have been widely and successfully used in combination with ensemble reweighting strategies of proteins [29,31,47,55,56,59,60,62,[121][122][123][124][125][126][127][128][129][130][131][132]; other approaches employed various sources of experimental data, e.g. for small molecules.…”

Section: Refining Protein and Rna Force Fieldsmentioning

confidence: 99%

“…An alternative to the χ 2 r = 1 method is cross-validation, i.e. fitting the optimal ensemble to a separate set of data that has not been used in the analysis [62,82,131,192,193]. The reweighting or experiment-biased simulation may then be done for several values of θ to monitor when the goodness of fit to the unused data starts to decrease.…”

Section: Balance Between Simulations and Experimental Datamentioning

confidence: 99%

How to learn from inconsistencies: Integrating molecular simulations with experimental data

Orioli

Larsen

Bottaro

et al. 2020

Progress in Molecular Biology and Translational Science

Self Cite

167

View full text Add to dashboard Cite

Molecular simulations and biophysical experiments can be used to provide independent and complementary insights into the molecular origin of biological processes. A particularly useful strategy is to use molecular simulations as a modelling tool to interpret experimental measurements, and to use experimental data to refine our biophysical models. Thus, explicit integration and synergy between molecular simulations and experiments is fundamental for furthering our understanding of biological processes. This is especially true in the case where discrepancies between measured and simulated observables emerge. In this chapter, we provide an overview of some of the core ideas behind methods that were developed to improve the consistency between experimental information and numerical predictions. We distinguish between situations where experiments are used to refine our understanding and models of specific systems, and situations where experiments are used more generally to refine transferable models. We discuss different philosophies and attempt to unify them in a single framework. Until now, such integration between experiments and simulations have mostly been applied to static data, and we discuss more recent developments aimed to analyse time-dependent or time-resolved data.

show abstract

Fitting Corrections to an RNA Force Field Using Experimental Data

Cited by 66 publications

References 52 publications

Fine-tuning of the AMBER RNA Force Field with a New Term Adjusting Interactions of Terminal Nucleotides

Fine-tuning of the AMBER RNA Force Field with a New Term Adjusting Interactions of Terminal Nucleotides

Bayesian-Maximum-Entropy reweighting of IDP ensembles based on NMR chemical shifts

How to learn from inconsistencies: Integrating molecular simulations with experimental data

Contact Info

Product

Resources

About