Optimisation of NMR dynamic models II. A new methodology for the dual optimisation of the model-free parameters and the Brownian rotational diffusion tensor

Finding the dynamics of an entire macromolecule is a complex problem as the model-free parameter values are intricately linked to the Brownian rotational diffusion of the molecule, mathematically through the autocorrelation function of the motion and statistically through model selection. The solution to this problem was formulated using set theory as an element of the universal set —the union of all model-free spaces (d’Auvergne EJ and Gooley PR (2007) Mol BioSyst 3(7), 483–494). The current procedure commonly used to find the universal solution is to initially estimate the diffusion tensor parameters, to optimise the model-free parameters of numerous models, and then to choose the best model via model selection. The global model is then optimised and the procedure repeated until convergence. In this paper a new methodology is presented which takes a different approach to this diffusion seeded model-free paradigm. Rather than starting with the diffusion tensor this iterative protocol begins by optimising the model-free parameters in the absence of any global model parameters, selecting between all the model-free models, and finally optimising the diffusion tensor. The new model-free optimisation protocol will be validated using synthetic data from Schurr JM et al. (1994) J Magn Reson B 105(3), 211–224 and the relaxation data of the bacteriorhodopsin (1–36)BR fragment from Orekhov VY (1999) J Biomol NMR 14(4), 345–356. To demonstrate the importance of this new procedure the NMR relaxation data of the Olfactory Marker Protein (OMP) of Gitti R et al. (2005) Biochem 44(28), 9673–9679 is reanalysed. The result is that the dynamics for certain secondary structural elements is very different from those originally reported. Electronic supplementary material The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

Model-free model elimination: A new step in the model-free dynamic analysis of NMR relaxation data

Model-free analysis is a technique commonly used within the field of NMR spectroscopy to extract atomic resolution, interpretable dynamic information on multiple timescales from the R ₁, R ₂, and steady state NOE. Model-free approaches employ two disparate areas of data analysis, the discipline of mathematical optimisation, specifically the minimisation of a χ² function, and the statistical field of model selection. By searching through a large number of model-free minimisations, which were setup using synthetic relaxation data whereby the true underlying dynamics is known, certain model-free models have been identified to, at times, fail. This has been characterised as either the internal correlation times, τ _e , τ _f , or τ _s , or the global correlation time parameter, local τ _m , heading towards infinity, the result being that the final parameter values are far from the true values. In a number of cases the minimised χ² value of the failed model is significantly lower than that of all other models and, hence, will be the model which is chosen by model selection techniques. If these models are not removed prior to model selection the final model-free results could be far from the truth. By implementing a series of empirical rules involving inequalities these models can be specifically isolated and removed. Model-free analysis should therefore consist of three distinct steps: model-free minimisation, model-free model elimination, and finally model-free model selection. Failure has also been identified to affect the individual Monte Carlo simulations used within error analysis. Each simulation involves an independent randomised relaxation data set and model-free minimisation, thus simulations suffer from exactly the same types of failure as model-free models. Therefore, to prevent these outliers from causing a significant overestimation of the errors the failed Monte Carlo simulations need to be culled prior to calculating the parameter standard deviations.     

Fast-time scale dynamics of outer membrane protein A by extended model-free analysis of NMR relaxation data

In order to better understand the dynamics of an integral membrane protein, backbone amide ¹⁵N NMR dynamics measurements of the β-barrel membrane protein OmpA have been performed at three magnetic fields. A total of nine relaxation data sets were globally analyzed using an extended model-free formalism. The diffusion tensor was found to be prolate axially symmetric with an axial ratio of 5.75, indicating a possible rotation of the protein within the micelle. The generalized order parameters gradually decreased from the mid-plane towards the two ends of the barrel, counteracting the dynamic gradient of the lipids in a matching bilayer, and were dramatically reduced in the extracellular loops. Large-scale internal motions on the ns time scale indicate that entire loops most likely undergo concerted (“sea anemone”-like) motions emanating from their anchoring points on the barrel. The case of OmpA in DPC micelles also illustrates inherent limitations of analyzing the data with even the most sophisticated current models of the model-free formalism. It is likely that conformational exchange processes on the ms-μs also play a role in describing the motions of some residues, but their analysis did not produce unique results that could be independently verified.     

Anisotropic rotational diffusion in model-free analysis for a ternary DHFR complex

Model-free analysis has been extensively used to extract information on motions in proteins over a wide range of timescales from NMR relaxation data. We present a detailed analysis of the effects of rotational anisotropy on the model-free analysis of a ternary complex for dihydrofolate reductase (DHFR). Our findings show that the small degree of anisotropy exhibited by DHFR (D_||/D=1.18) introduces erroneous motional models, mostly exchange terms, to over 50% of the NH spins analyzed when isotropic tumbling is assumed. Moreover, there is a systematic change in S², as large as 0.08 for some residues. The significant effects of anisotropic rotational diffusion on model-free motional parameters are in marked contrast to previous studies and are accentuated by lowering of the effective correlation time using isotropic tumbling methods. This is caused by the preponderance of NH vectors aligned perpendicular to the principal diffusion tensor axis and is readily detected because of the high quality of the relaxation data. A novel procedure, COPED (COmparison of Predicted and Experimental Diffusion tensors) is presented for distinguishing genuine motions from the effects of anisotropy by comparing experimental relaxation data and data predicted from hydrodynamic analyses. The procedure shows excellent agreement with the slow motions detected from the axially symmetric model-free analysis and represents an independent procedure for determining rotational diffusion and slow motions that can confirm or refute established procedures that rely on relaxation data. Our findings show that neglect of even small degrees of rotational diffusion anisotropy can introduce significant errors in model-free analysis when the data is of high quality. These errors can hinder our understanding of the role of internal motions in protein function.     

New modification of model-free approach to analysis of nuclear magnetic relaxation data in proteins

A formal approach to the analysis of 13C magnetic relaxation data in proteins has been developed. It is based on the concepts of one of the authors on the internal motions in solid polymers (Fedotov, V.D., Pulse NMR in bulk polymers, Doctoral dissertation, Kazan, USSR, 1981). According to this approach the intermolecular motions in proteins are considered as anisotropic ones and described in terms of a spectrum of correlation times. To characterize the motions a set of formal microdynamic parameters has been introduced. They are: the anisotropy parameter (a measure of spatial restriction of motion), the most probable correlation time, the parameter of the correlation time distribution width. The analysis of protonated carbon relaxation in globular proteins (bovine pancreatic trypsin inhibitor and ribonuclease S) and polymers has been carried out by the model-free approach. Microdynamic parameters of CH3-, CH2-, aromatic CH-groups have been considered within the framework of the diffusional rotation-oscillation models. To explain the backbone CH-group relaxation the model of the defect diffusion has been applied. The distinctive feature of the results obtained is the broad correlation time distribution for all groups of any type. The causes of nonexponential correlation function of local motion have been discussed. To elucidate the nature of the correlation time the carbon magnetization decays in the wide range of microdynamic parameter values imitating various experimental conditions have been calculated.     

Application of the quasi-spectral density function of 15N nuclei to the selection of a motional model for model-free analysis

Summary Parameters used in model-free analysis were related to simulated spectral density functions in a frequency region experimentally obtained by quasi-spectral density function analysis of ¹⁵N nuclei. Five kinds of motional models used in recent model-free analyses were characterized by a simple classification of the experimental spectral density function. We demonstrate advantages and limitations of each of the motional models. To verify the character of the models, model selection using experimental spectral density functions was examined.     

Model-free Analysis of Protein Dynamics: Assessment of Accuracy and Model Selection Protocols Based on Molecular Dynamics Simulation

The popular model-free approach to analyze NMR relaxation measurements has been examined using artificial amide (15)N relaxation data sets generated from a 10 nanosecond molecular dynamics trajectory of a dihydrofolate reductase ternary complex in explicit water. With access to a detailed picture of the underlying internal motions, the efficacy of model-free analysis and impact of model selection protocols on the interpretation of NMR data can be studied. In the limit of uncorrelated global tumbling and internal motions, fitting the relaxation data to the model-free models can recover a significant amount of quantitative information on the internal dynamics. Despite a slight overestimation, the generalized order parameter is quite accurately determined. However, the model-free analysis appears to be insensitive to the presence of nanosecond time scale motions with relatively small magnitude. For such cases, the effective correlation time can be significantly underestimated. As a result, proteins appear to be more rigid than they really are. The model selection protocols have a major impact on the information one can reliably obtain. The commonly employed protocol based on step-up hypothesis testing has severe drawbacks of oversimplification and underfitting. The consequences are that the order parameter is more severely overestimated and the correlation time more severely underestimated. Instead, model selection based on Bayesian Information Criteria (BIC), recently introduced to the model-free analysis by d'Auvergne and Gooley (2003), provides a better balance between bias and variance. More appropriate models can be selected, leading to improved estimate of both the order parameter and correlation time. In addition, the computational cost is significantly reduced and subjective parameters such as the significance level are unnecessary.     

Probing the flexibility of the DsbA oxidoreductase from Vibrio cholerae--a 15N - 1H heteronuclear NMR relaxation analysis of oxidized and reduced forms of DsbA

We have determined the structure of the reduced form of the DsbA oxidoreductase from Vibrio cholerae. The reduced structure shows a high level of similarity to the crystal structure of the oxidized form and is typical of this class of enzyme containing a thioredoxin domain with an inserted alpha-helical domain. Proteolytic and thermal stability measurements show that the reduced form of DsbA is considerably more stable than the oxidized form. NMR relaxation data have been collected and analyzed using a model-free approach to probe the dynamics of the reduced and oxidized states of DsbA. Akaike's information criteria have been applied both in the selection of the model-free models and the diffusion tensors that describe the global motions of each redox form. Analysis of the dynamics reveals that the oxidized protein shows increased disorder on the pico- to nanosecond and micro- to millisecond timescale. Many significant changes in dynamics are located either close to the active site or at the insertion points between the domains. In addition, analysis of the diffusion data shows there is a clear difference in the degree of interdomain movement between oxidized and reduced DsbA with the oxidized form being the more rigid. Principal components analysis has been employed to indicate possible concerted movements in the DsbA structure, which suggests that the modeled interdomain motions affect the catalytic cleft of the enzyme. Taken together, these data provide compelling evidence of a role for dynamics in the catalytic cycle of DsbA.     

Multiple loci mapping via model-free variable selection

Despite recent flourish of proposals on variable selection, genome-wide multiple loci mapping remains to be challenging. The majority of existing variable selection methods impose a model, and often the homoscedastic linear model, prior to selection. However, the true association between the phenotypical trait and the genetic markers is rarely known a priori, and the presence of epistatic interactions makes the association more complex than a linear relation. Model-free variable selection offers a useful alternative in this context, but the fact that the number of markers p often far exceeds the number of experimental units n renders all the existing model-free solutions that require n > p inapplicable. In this article, we examine a number of model-free variable selection methods for small-n-large-p regressions in the context of genome-wide multiple loci mapping. We propose and advocate a multivariate group-wise adaptive penalization solution, which requires no model prespecification and thus works for complex trait-marker association, and handles one variable at a time so that works for n < p. Effectiveness of the new method is demonstrated through both intensive simulations and a comprehensive real data analysis across 6100 gene expression traits.     

Base dynamics in a UUCG tetraloop RNA hairpin characterized by 15N spin relaxation: correlations with structure and stability.

Intramolecular dynamics of guanine and uracil bases in a 14-nt RNA hairpin including the extraordinarily stable UUCG tetraloop were studied by 15N spin relaxation experiments that are sensitive to structural fluctuations occurring on a time scale of picoseconds to nanoseconds. The relaxation data were interpreted in the framework of the anisotropic model-free formalism, using assumed values for the chemical shift anisotropies of the 15N spins. The rotational diffusion tensor was determined to be symmetric with an axial ratio of 1.34 +/- 0.12, in agreement with estimates based on the ratio of the principal moments of the inertia tensor. The model-free results indicate that the bases of the G x U pair in the tetraloop are at least as rigid as the interior base pairs in the stem, whereas the 5'-terminal guanine is more flexible. The observed range of order parameters corresponds to base fluctuations of 19-22 degrees about the chi torsion angle. The results reveal dynamical consequences of the unusual structural features in the UUCG tetraloop and offer insights into the configurational entropy of hairpin formation.     

Reverse smoothing: a model-free data smoothing algorithm

Biophysical chemistry experiments, such as sedimentation-equilibrium analyses, require computational techniques to reduce the effects of random errors of the measurement process. The existing approaches have primarily relied on assumption of polynomial models and least-squares approximation. Such models by constraining the data to remove random fluctuations may distort the data and cause loss of information. The better the removal of random errors the greater is the likely introduction of systematic errors through the constraining fit itself. An alternative technique, reverse smoothing, is suggested that makes use of a more model-free approach of exponential smoothing of the first derivative. Exponential smoothing approaches have been generally unsatisfactory because they introduce significant data lag. The approaches given here compensates for the lag defect and appears promising for the smoothing of many experimental data sequences, including the macromolecular concentration data generated by sedimentation-equilibria experiments. Test results on simulated sedimentation-equilibrium data indicate that a 4-fold reduction in error may be typical over standard analyses techniques.     

Quantitative cell biology with the Virtual Cell

Cell biological processes are controlled by an interacting set of biochemical and electrophysiological events that are distributed within complex cellular structures. Computational models, comprising quantitative data on the interacting molecular participants in these events, provide a means for applying the scientific method to these complex systems. The Virtual Cell is a computational environment designed for cell biologists, to facilitate the construction of models and the generation of predictive simulations from them. This review summarizes how a Virtual Cell model is assembled and describes the physical principles underlying the calculations that are performed. Applications to problems in nucleocytoplasmic transport and intracellular calcium dynamics will illustrate the power of this paradigm for elucidating cell biology.     

Evaluating the Accuracy of Diffusion MRI Models in White Matter

Models of diffusion MRI within a voxel are useful for making inferences about the properties of the tissue and inferring fiber orientation distribution used by tractography algorithms. A useful model must fit the data accurately. However, evaluations of model-accuracy of commonly used models have not been published before. Here, we evaluate model-accuracy of the two main classes of diffusion MRI models. The diffusion tensor model (DTM) summarizes diffusion as a 3-dimensional Gaussian distribution. Sparse fascicle models (SFM) summarize the signal as a sum of signals originating from a collection of fascicles oriented in different directions. We use cross-validation to assess model-accuracy at different gradient amplitudes (b-values) throughout the white matter. Specifically, we fit each model to all the white matter voxels in one data set and then use the model to predict a second, independent data set. This is the first evaluation of model-accuracy of these models. In most of the white matter the DTM predicts the data more accurately than test-retest reliability; SFM model-accuracy is higher than test-retest reliability and also higher than the DTM model-accuracy, particularly for measurements with (a) a b-value above 1000 in locations containing fiber crossings, and (b) in the regions of the brain surrounding the optic radiations. The SFM also has better parameter-validity: it more accurately estimates the fiber orientation distribution function (fODF) in each voxel, which is useful for fiber tracking.     

An efficient method of modeling material properties using a thermal diffusion analogy: an example based on craniofacial bone

The ability to incorporate detailed geometry into finite element models has allowed researchers to investigate the influence of morphology on performance aspects of skeletal components. This advance has also allowed researchers to explore the effect of different material models, ranging from simple (e.g., isotropic) to complex (e.g., orthotropic), on the response of bone. However, bone's complicated geometry makes it difficult to incorporate complex material models into finite element models of bone. This difficulty is due to variation in the spatial orientation of material properties throughout bone. Our analysis addresses this problem by taking full advantage of a finite element program's ability to solve thermal-structural problems. Using a linear relationship between temperature and modulus, we seeded specific nodes of the finite element model with temperatures. We then used thermal diffusion to propagate the modulus throughout the finite element model. Finally, we solved for the mechanical response of the finite element model to the applied loads and constraints. We found that using the thermal diffusion analogy to control the modulus of bone throughout its structure provides a simple and effective method of spatially varying modulus. Results compare favorably against both experimental data and results from an FE model that incorporated a complex (orthotropic) material model. This method presented will allow researchers the ability to easily incorporate more material property data into their finite element models in an effort to improve the model's accuracy.     

Estimation of organ transport function: model-free deconvolution by recursive quadratic programming optimization.

A model-free deconvolution method is proposed for evaluating the frequency distribution function of organ transit times. The deconvolution is treated as a nonlinear constrained optimization problem and it is solved by using a modified constrained variable metric approach. The only constraint implemented in the solution is that all the discrete transport function values are not allowed to become negative. The method is tested on model mathematical systems of known analytical transport functions. The tests are performed on systems that included noise in both the input and output functions. The criteria of successful deconvolution are the reconvolution error and, most importantly, the deviation of the computed transport function from the known analytical one. The proposed method is then applied, as a pilot experiment, to biological data obtained from an isolated, perfused rabbit lung preparation contained within a plethysmograph. The results indicate that this type of deconvolution produces stable estimates which faithfully follow the analytical function while negating the need to assume either any functional form for the behavior of the transport function or any educated initial guess of its values.     

NMR analysis of carbohydrates with model-free spectral densities: the dispersion range revisited

Over the past decade molecular mechanics and molecular dynamics studies have demonstrated considerable flexibility for carbohydrates. In order to interpret the corresponding NMR parameters, which correspond to a time-averaged or 'virtual' conformer, it is necessary to simulate the experimental data using the averaged geometrical representation obtained with molecular modelling methods. This structural information can be transformed into theoretical NMR data using empirical Karplus-type equations for the scalar coupling constants and the appropriate formalism for the relaxation parameters. In the case of relaxation data, the 'model-free' spectral densities have been widely used in order to account for the internal motions in sugars. Several studies have been conducted with truncated model-free spectral densities based on the assumption that internal motion is very fast with respect to overall tumbling. In this report we present experimental and theoretical evidence that suggests that this approach is not justified. Indeed, recent results show that even in the case of moderate-sized carbohydrates internal motions are occurring on the same timescale as molecular reorientation. Simulations of relaxation parameters (NOESY volumes, proton cross-relaxation rates, carbon T1 and nOe values) in the dispersion range (0.1<Tc<5 ns) show that rates of internal motion can be fairly precisely defined with respect to overall tumbling. Experimental data for a variety of oligosaccharides clearly indicate similar timescales for internal and overall motion. This revised version was published online in August 2006 with corrections to the Cover Date.     

Assessing potential bias in the determination of rotational correlation times of proteins by NMR relaxation

The various factors that influence the reliable and efficient determination of the correlation time describing molecular reorientation of proteins by NMR relaxation methods are examined. Nuclear Overhauser effects, spin-lattice, and spin-spin relaxation parameters of 15N NMR relaxation in ubiquitin have been determined at 17.6, 14.1, 11.7 and 9.4 Tesla. This unusually broad set of relaxation parameters has allowed the examination of the influence of chemical shift anisotropy, the functional form of the model-free spectral density, and the reliability of determined spin- spin relaxation parameters on the characterization of global tumbling of the protein. Treating the 15N chemical shift anisotropy (CSA) as an adjustable parameter, a consensus value of –170 ± 15ppm for the breadth of the chemical shift tensor and a global isotropic correlation time of 4.1ns are found when using the model-free spectral density to fit T1 and NOE data from all fields. The inclusion of T2 relaxation parameters in the determination of the global correlation time results in its increase to 4.6ns. This apparent inconsistency may explain a large portion of the discrepancy often found between NMR- and fluorescence-derived m values for proteins. The near identity of observed T2 and T1 values suggests that contributions from slow motions are not the origin of the apparent inconsistency with obtained T1 and NOE data. Various considerations suggest that the origin of this apparent discrepancy may reside in a contribution to the spectral density at zero frequency that is not represented by the simple model-free formalism in addition to the usual experimental difficulties associated with the measurement of these relaxation parameters. Finally, an axially symmetric diffusion tensor for ubiquitin is obtained using exclusively T1 and NOE data. A recommendation is reached on the types and combinations of relaxation data that can be used to reliably determine m values. It is also noted that the reliable determination of m values from 15N T1 and NOE relaxation parameters will become increasingly difficult as m increases.