systematic assessment of the effects of an all-atom force field...

1944 Bull. Korean Chem. Soc. 2014, Vol. 35, No. 7 Jun-Goo Jee

http://dx.doi.org/10.5012/bkcs.2014.35.7.1944

Systematic Assessment of the Effects of an All-Atom Force Field and the

Implicit Solvent Model on the Refinement of NMR Structures with Subsets

of Distance Restraints

Jun-Goo Jee

Research Institute of Pharmaceutical Sciences, College of Pharmacy, Kyungpook National University,

Daegu 702-701, Korea. E-mail: [email protected]

Received February 22, 2014, Accepted March 5, 2014

Employment of a time consuming, sophisticated calculation using the all-atom force field and generalized-Born implicit solvent model (GBIS) for refinement of NMR structures has become practical through advancesin computational methods and capacities. GBIS refinement improves the qualities of the resulting NMRstructures with reduced computational times. However, the contribution of GBIS to NMR structures has notbeen sufficiently studied in a quantitative way. In this paper, we report the effects of GBIS on the refined NMRstructures of ubiquitin (UBQ) and GB1 with subsets of distance restraints derived from experimental data.Random omission prepared a series of distance restraints 0.05, 0.1, 0.3, 0.5, and 0.7 times smaller. For eachnumber, we produced five different restraints for statistical analysis. We then recalculated the NMR structuresusing CYANA software, followed by GBIS refinements using the AMBER package. GBIS improved both theprecision and accuracy of all the structures, but to varied levels. The degrees of improvement were significantwhen the input restraints were insufficient. In particular, GBIS enabled GB1 to form an accurate structure evenwith distance restraints of 5%, revealing that the root-mean-square deviation was less than 1 Å from the X-raybackbone structure. We also showed that the efficiency of searching the conformational space was moreimportant for finding accurate structures with the calculation of UBQ with 5% distance restraints than thenumber of conformations generated. Our data will provide a meaningful guideline to judge and compare thestructural improvements by GBIS.

Key Words : NMR, Generalized-Born implicit solvent model, Ubiquitin, GB1

Introduction

Steady advances in algorithms for both NOE assignmentsand structure calculations have made automatic calculationsof protein 3D structures with raw NOE data, provided thechemical shifts are assigned for most atoms and sufficientpeaks exist.1 The algorithms automate an iterative process inwhich the assignment of NOE peaks and structure calcu-lations with the new restraints are coupled. Owing to thecurrent state of computational power, several studies havereported that fully automatic structure calculations fromprocessed NMR data were feasible without any manualinterpretation.2-5 However, the improvements and refine-ments of NMR structures are still believed to be largelydependent on the skills and experiences of the researchersinterpreting and calculating the structures.

Despite the sensitivity increments of NMR hardware andthe developments of new experiments over the last decades,the number of experimentally obtainable NMR restraints fordetermining the coordinates of a 3D structure is still muchsmaller than those obtainable using X-ray crystallography.Computational aids have contributed to improving the struc-tural qualities of NMR structures. The efforts are classifiedinto two criteria: the use of an empirical database and theapplication of sophisticated calculation methods stemming

mostly from molecular dynamics (MD) simulations. TheMD simulation consists of two main parts, an atomistic forcefield and a conformational space search. Since NMR struc-ture calculations make use of experimental restraints, forcefields simpler than those used in conventional MD simula-tions were employed,6,7 allowing for conformation searchesat higher temperatures without concern for the instability ofthe system during the computation. Through the advances ofalgorithms and computational powers, MD simulation hasgrown popular. All-atom force field, which enables a bio-molecule to behave in a realistic way during MD simulation,has matured, helping reconcile the discrepancy betweenexperimental and simulated data. Along with the force field,the implicit solvent model, a mostly generalized-Born model,has improved the qualities of resulting structures as well byapproximating the effects that explicit solvents bring abouteven with reduced computational times. This results in theoutcomes that are expected with the explicit solvents.8

Determination of the NMR structure with an all-atomforce field and a generalized-Born model (hereafter GBIS),restrained with experimental data, has successfully improvedNMR structures, particularly in the refinement stage of thecalculations. Especially, the GBIS refinement is advantage-ous for improving the local geometries that cannot be con-fined due to a lack of experimental restraints. For instance,

Refinements of NMR Structures with GBIS Bull. Korean Chem. Soc. 2014, Vol. 35, No. 7 1945

GBIS has helped in unambiguous positioning of the donorsand acceptors of hydrogen bonds, allowing further insightinto the pH dependence of binding affinity in the complexbetween UIM and ubiquitin.9 Besides improvement in thelocal geometry, GBIS refinement is effective in determiningthe global fold as well. As reported by Brooks and hiscolleagues, GBIS refinement could determine 3D folds withless than 10% of the original NOE data.10,11

Considering its potential, the applications of GBIS forrefining NMR structures are likely to increase, particularlyfor the proteins whose structures are difficult to determinewith conventional methods. The proteins include membraneproteins and complexes consisting of multiple proteins.Despite the advances in algorithms, however, the effects ofGBIS on NMR structure calculations are not straightf-orward, necessitating quantitative analyses. In this paper, wecalculated 3D structures of ubiquitin (UBQ) and GB1 usingconventional and GBIS methods with a variety of subsetsfrom experimental NMR distance restraints. The detailedquantitative interpretation facilitates the comparison of theresults from the GBIS method with those from the conv-entional methods.

Experimental

Restraints for NMR Structure Calculation. Experimentalrestraints for calculating the structures of UBQ and GB1were extracted from the PDB database (http://www.rcsb.org),where they are deposited as 1D3Z and 3GB1, respectively.Only the distance and backbone torsion angle restraints wereemployed. The numbers for the distance restraints are 1,446and 584 for ubiquitin and GB1, respectively. The number ofrestraints for UBQ(GB1) are 288(122), 294(122), 236(83),and 628(257) for intra (|i-j|=0), sequential (|i-j|=1), medium(1


with extensive experimental restraints that include residualdipolar and J-scalar couplings, resulting in very precise andaccurate backbone geometries, 0.09 and 0.18 Å for eRMSDand 0.29 and 0.58 Å for rRMSD, in UBQ and GB1, respec-tively. We quantified the geometric qualities of side-chainsusing MolProbity clash score, where the lower value indicatesthe better geometry. For example, 0.081 of the MolProbityscore corresponds to the top 100 percentile. The GBIS refin-ed structures of both UBQ and GB1 showed improvementsin the MolProbity score. In particular, the decrease in GB1from 1.785 (run-0) to 0.670 (run-1) was marked. It can beexplained by the fact that most of the additional restraintsfrom residual dipolar and J-scalar couplings were confinedto backbone atoms, whereas the effects of GBIS are spreadacross side-chains. Our data indicates that there is still a needto employ GBIS refinements in highly refined structureswith experimental restraints.

GBIS Greatly Improved both Precision and Accuracy

when the Number of Restraints Decreased. GBIS improv-ed all the structures that were calculated with subsets ofdistance restraints. The increments of omission in the distancerestraints elevated eRMSDCYA and rRMSDCYA, indicating

the difficulties in finding precise and accurate conformationsby CYANA calculations. The decreases of eRMSDGBIS andrRMSDGBIS from eRMSDCYA and rRMSDCYA, respectively,were marked. However, the side-chain packing qualitiesfrom MolProbity scores were indistinguishable in most ofthe GBIS refined structures (Tables 1 and 2). Plots ofeRMSDCYA versus (eRMSDCYA-eRMSDGBIS), and rRMSDCYA

versus (rRMSDCYA-rRMSDGBIS) clearly represent the degreesand tendencies (Fig. 1). Here please note that higher valuesin Y-axes indicate larger improvements by GBIS. All thevalues were greater than zero. The apparent linearity bet-ween RMSDCYA and (RMSDCYA-RMSDGBIS means that theimprovements were proportional. However, there were dis-similarities between the UBQ and GB1 cases. All the CYANAcalculations of both UBQ and GB1 did not generate anensemble of structures that had eRMSD and rRMSD valuesof less than 1 Å when the restraints were reduced more than10-fold (Tables 1 and 2). GBIS could not yield precise andaccurate structures of UBQ like those in GB1. Whereas therewere significant improvements in all the GBIS results ofUBQ, most of the rRMSDGBIS values were greater than 1 Åwith 10% restraints and greater than 2 Å with 5% restraints.

Table 1. GBIS-refined ubiquitin structures

Model IDDistance

Rst.eRMSD (Å)

CYANAeRMSD (Å)

GBISrRMSD (Å)

CYANArRMSD (Å)

GBISEnergy

(kcal/mol)Ramachan-dran (%)d

Mol-Probity

1D3Z 0 1,446 (628)a 0.09 (10)b n.a.c 0.29 (10) n.a. n.a. 96.7 1.308

100% 1 1,446 (628) 0.51 0.28 0.91 0.62 -3,129 88.0 1.289

70%

11 1,012 (437) 0.59 0.31 0.85 0.55 -3,171 91.1 0.901

12 1,012 (438) 0.63 0.28 0.83 0.55 -3,155 90.9 1.199

13 1,012 (441) 0.57 0.29 0.84 0.60 -3,162 90.7 1.103

14 1,012 (435) 0.58 0.38 0.87 0.60 -3,155 89.1 1.158

15 1,012 (436) 0.66 0.39 0.81 0.69 -3,156 90.8 1.011

50%

21 723 (313) 0.57 0.37 0.82 0.49 -3,217 94.5 0.628

22 723 (326) 0.69 0.46 1.02 0.60 -3,165 90.3 1.082

23 723 (296) 0.64 0.41 0.93 0.59 -3,169 90.2 1.185

24 723 (320) 0.46 0.38 0.79 0.53 -3,183 91.8 0.968

25 723 (309) 0.46 0.35 0.85 0.53 -3,193 92.3 0.999

30%

31 434 (189) 0.78 0.40 0.92 0.55 -3,207 94.2 0.873

32 434 (193) 0.60 0.47 1.03 0.60 -3,199 93.6 0.887

33 434 (193) 0.70 0.49 0.99 0.61 -3,196 94.4 0.912

34 434 (201) 0.66 0.36 0.96 0.66 -3,183 91.2 0.804

35 434 (183) 0.74 0.55 1.21 0.70 -3,172 91.9 0.917

10%

41 145 (66) 1.46 0.77 1.78 0.92 -3,173 94.3 0.787

42 145 (60) 1.28 0.82 2.05 1.32 -3,153 90.1 1.094

43 145 (59) 2.05 1.24 2.51 1.28 -3,121 90.7 1.084

44 145 (65) 1.83 1.21 2.60 1.23 -3,134 90.2 0.931

45 145 (49) 2.10 1.24 3.07 1.32 -3,117 91.7 1.024

5%

51 72 (33) 3.21 2.29 4.45 2.88 -3,047 84.8 1.407

52 72 (38) 2.69 2.23 3.99 2.58 -3,062 86.8 1.276

53 72 (31) 3.47 2.52 4.07 3.09 -3,033 83.1 1.368

54 72 (33) 3.72 3.06 3.94 3.11 -3,031 80.8 1.465

55 72 (27) 3.17 3.11 5.64 5.09 -3,026 81.1 1.291aThe numbers in parentheses of distance restraints mean those for long range NOEs. bThe number of conformations in an ensemble. If not mentioned,the number is 20. c“n.a.” means “not available”. dFor Ramachan analysis, only most favored regions are considered.


For example, the value of rRMSDGBIS at run-55 was 5.09,indicating that the structures of UBQ do not agree with theX-ray structure. On the other hand, GBIS did enable GB1 toform structures that had comparable qualities in both pre-cision (< 1 Å) and accuracy (< 1 Å) to those of the reference.In particular, GB1 could reduce eRMSD from 3.37 to 0.45and rRMSD from 10.00 to 0.56 Å at run-53. It is remarkablethat GBIS enabled a wrong fold to recover into a correctstructure. Nevertheless, it should be noted that when thestructures were refined with only 5% restraints the numberof structures that did not show significant distance violationsagainst the input data was only 10 and 9 at run-51 and -52,respectively. In all other cases, the numbers of non-violatingstructures exceeded 20. The lack of numbers at run-51 and-52 might be caused by the inefficiency of the conformationalsearch due to either insufficient restraints to guide the struc-tures to the correct fold or an initial wrong geometry thatcannot be escaped for the given sampling steps of the con-formational search.

Ensemble RMSD of GBIS-refined Structures Could be

Used as a Validation Metric. The improvements by GBISled to the decreased rRMSD, which in turn caused a better

correlation between eRMSDGBIS and rRMSDGBIS thaneRMSDCYA and rRMSDCYA (Fig. 2(a), (b)). One of the mainconcerns in calculating NMR structures of biomolecules isthe validation of the fold of the resulting structures. NOE,from which a distance restraint is prepared, originates bet-ween two atoms located proximally in space (< 6 Å). Becauseof the short-range feature, evaluating the contribution ofeach distance restraint to the global fold in a quantitativeway is not straightforward. In addition, one cannot discri-minate the accurate and inaccurate structures with just theviolations and energies information. Please note that the datafor run-55 did not show any significant violation againstinput restraints but revealed inaccurate results. Proper globalparameters that allow for judgment of the soundness ofNMR structures of biomolecules have been awaited. Wefound that two criteria of the eRMSDCYA and eRMSDGBIS

showed apparent correlation with rRMSDCYA and rRMSDGBIS,respectively (Fig. 2(a), (b)). AMBER energy also correlatedwith rRMSDGBIS (Fig. 2(c)). The Pearson’s correlationfactors (R) for these metrics were 0.91 for eRMSDCYA versusrRMSDCYA, 0.97 for eRMSDGBIS versus rRMSDGBIS, and0.73 for AMBER energy versus rRMSDGBIS. It would be

Table 2. GBIS-refined GB1 structuresa

Model IDDistance

Rst.eRMSD (Å)

CYANAeRMSD (Å)

GBISrRMSD (Å)

CYANArRMSD (Å)

GBISEnergy

(kcal/mol)Ramachan-dran (%)

Mol-Probity

3GB1 0 584 (257) 0.18 n.a. 0.58 n.a. n.a. 94.4 1.785

100% 1 584 (257) 0.46 0.29 0.68 0.59 -2,065 95.0 0.670

70%

11 409 (178) 0.48 0.44 0.82 0.57 -2,064 95.5 0.683

12 409 (184) 0.44 0.41 0.79 0.58 -2,065 96.4 0.661

13 409 (188) 0.46 0.39 0.77 0.58 -2,065 95.3 0.613

14 409 (181) 0.45 0.35 0.77 0.59 -2,067 95.9 0.644

15 409 (169) 0.45 0.34 0.78 0.59 -2,068 95.7 0.704

50%

21 292 (120) 0.52 0.49 0.87 0.55 -2,058 95.5 0.674

22 292 (132) 0.65 0.42 0.82 0.57 -2,065 95.3 0.582

23 292 (127) 0.48 0.49 0.81 0.59 -2,062 95.2 0.725

24 292 (134) 0.48 0.38 0.92 0.56 -2,066 95.9 0.650

25 292 (124) 0.48 0.42 0.88 0.54 -2,063 95.2 0.657

30%

31 175 (76) 0.81 0.37 1.46 0.60 -2,064 95.7 0.706

32 175 (69) 0.79 0.41 1.05 0.60 -2,065 95.9 0.716

33 175 (84) 0.70 0.46 1.22 0.56 -2,063 95.6 0.637

34 175 (77) 0.75 0.50 1.19 0.52 -2,063 95.7 0.678

35 175 (78) 0.69 0.44 0.87 0.57 -2,065 95.0 0.732

10%

41 58 (26) 1.89 1.64 3.08 1.98 -1,984 93.7 0.827

42 58 (29) 1.22 0.40 1.81 0.60 -2,062 96.4 0.679

43 58 (26) 1.42 0.43 2.10 0.59 -2,057 95.3 0.638

44 58 (26) 1.25 0.47 2.18 0.55 -2,060 95.6 0.679

45 58 (24) 2.39 0.40 3.47 0.55 -2,060 95.4 0.699

5%

51 29 (12) 2.78 0.29 (10)* 3.64 0.57 (10)* -2,065 95.4 0.634

52 29 (12) 4.34 0.30 (9)* 7.61 0.62 (9)* -2,064 94.9 0.929

53 29 (13) 3.37 0.45 10.00 0.56 -2,062 94.9 0.629

54 29 (15) 1.81 0.37 2.80 0.60 -2,047 95.4 0.769

55 29 (12) 2.83 0.41 4.07 0.62 -2,060 95.3 0.724aAll the values are prepared with the same rules to Table 1.


difficult to directly apply the AMBER energy for validationwithout any calibration, since the value is dependent onprotein and it is not known which value is low enough apriori. The better correlation of rRMSD with eRMSDGBIS

than eRMSDCYA suggests eRMSDGBIS as a metric for valida-tion. At least it can suggest when the accuracy of the resultswill be suspected, i.e. when the value of eRMSDGBIS islarger than 2 Å. It is noteworthy that the representative struc-tures reflect AMBER energies in parts, because the selection

criteria of the top 20 structures are based on the AMBERenergies. Further studies in this direction are necessary forgeneral use of eRMSDGBIS as a validation metric.

GBIS with Additional Psi Angle Restraints Improved

Precision and Accuracy of Ubiquitin Structures. To fairlycompare the performances of GBIS in UBQ and GB1 refine-ments and know whether the inefficiencies of the GBISprotocol in UBQ run-41–55 originated from the lack ofbackbone torsion angle restraints, we added psi angle re-straints and performed GBIS refinements with UBQ. ThePDB-deposited restraints of UBQ do not contain psi anglerestraints. The restraints by psi angle are known to beimportant in discerning secondary structures. The psi anglerestraints can be easily prepared, provided the chemicalshifts are assigned. We extracted the chemical shifts of UBQfrom the BMRB database (BMRB ID: 6457) and generated59 psi angle restraints by TALOS+ software15 using thedatabase that excluded ubiquitin. We recalculated the runswith 5 and 10% distance restraints with additional torsionangle restraints (Table 3). It is noted again that the distancerestraints were identical between 41–45 and 61–65 andbetween 51–55 and 71–75 runs. Addition of psi anglerestraints improved qualities considerably in the results fromboth CYANA and GBIS. All the rRMSDGBIS values with10% restraints reduced to a value lower than 1.0 Å. Exceptfor run-75, all the other rRMSDGBISs with 5% restraints hada value lower than 2.0 Å. Nevertheless, the accuracy andprecision were still less compared to those obtained withGB1. Our data support the idea that the contributions ofindividual distance restraints in determining NMR structuresare varied and depend on the proteins.

The Efficiency of the Conformational Space Search

was More Important than the Number of Conformations

Generated for Calculating Accurate Structures. The nextquestion was whether enhancement of the GBIS refinementcould allow a structure trapped at an inaccurate position toescape accurately with only sparse experimental restraints.Our data showed that the GBIS was typically effective withsparse experimental restraints but not in all cases. The resultsat run-75 indicated that the structures were inaccurate, eventhough there was no significant violation of the input restraints.Because our original protocols did not work properly for therestraints of run-75, we considered two ways to improveGBIS efficiency. One was to increase the number of struc-tures calculated and the other was to increase the length ofduration in the dynamics. These two methods were appliedto the recalculation of the data from run-75 with identicalrestraints. We first generated 200 and 500 structures (run-81and -82) with the same protocols as the calculations inTables 1–3. Second, we increased the duration of dynamics2- and 5-fold, leading to 40- and 100-ps restrained simulatedannealing, respectively, with 100 structures (run-83 and-84). We selected 20 structures as a final ensemble from thetwo case runs, as well. The results showed that the runs withlonger durations generated better eRMSD and rRMSDvalues (Table 4). The 100-ps GBIS refinement (run-84)generated results with comparable qualities to the other 5%

Figure 1. Plots of eRMSDCYA and (eRMSDCYA-eRMSDGBIS) (a),and rRMSDCYA and (rRMSDCYA-rRMSDGBIS) (b). Labels of X-and Y-axis are written at bottom and top of each plot, respectively.Blue squares correspond to the runs by ubiquitin, whereas redcircles represent GB1. The run-55 of ubiquitin and the run-53 ofGB1 are labeled with dashed circles in blue and red, respectively.Note that higher values in Y-axes indicate larger improvements byGBIS.


results from run-71–74. Visual inspection of structures fromrun-55, run-75, and run-81–84 represents which partsimproved according to improved methods (Fig. 3). While α-helical regions that are confined by short range restraintswere well refined, β-strands were not converged due to lackof long range distance restraints. The data clearly demon-strates that the efficiency of the conformational space searchis more important for finding the correct fold, at least underthe conditions GBIS was applied. We do not intend to argueover the optimization of GBIS protocol in this study, butthere are enough possibilities that advanced sampling, includ-

ing replica-exchange,16 could lead to better results thanrestrained simulated annealing.

Conclusion

A more accurate and precise structure is helpful in study-ing and applying the function of a protein. For example,detailed information on the protein-protein and protein-ligand complex interfaces can facilitate the discovery ofmore effective inhibitors. However, NMR signals inter-preted as structural restraints are often invisible when the

Figure 2. Plots of eRMSDCYA and rRMSDCYA (a), eRMSDGBIS and rRMSDGBIS (b), ZGBIS and rRMSDGBIS. In order to fairly compare boththe cases of ubiquitin and GB1, we calculated Z values by dividing the difference of individual AMBER Energy and mean of AMBERenergies in a protein with standard deviation of AMBER Energy in a protein. Pearson’s correlation factors (R) were described as well.

Table 3. GBIS-refined ubiquitin structures with additional ψ angle restraints

Model IDDistance

Rst.eRMSD (Å)

CYANAeRMSD (Å)

GBISrRMSD (Å)

CYANArRMSD (Å)

GBISEnergy

(kcal/mol)Ramachan-dran (%)

Mol-Probity

10%

61 145 (66) 1.02 0.77 1.47 0.87 -3,198 82.5 0.778

62 145 (60) 1.10 0.61 1.50 0.88 -3,183 83.7 0.810

63 145 (59) 1.39 0.70 1.96 0.92 -3,157 83.4 0.918

64 145 (65) 1.17 0.70 2.01 0.75 -3,178 84.2 0.820

65 145 (49) 1.39 0.73 2.10 0.75 -3,183 84.6 0.836

5%

71 72 (33) 1.45 1.20 2.90 1.52 -3,135 88.5 0.906

72 72 (38) 1.48 1.00 2.43 1.02 -3,150 83.6 0.822

73 72 (31) 1.81 1.35 2.71 1.60 -3,121 82.3 0.946

74 72 (33) 1.86 1.48 2.35 1.46 -3,135 85.1 0.850

75 72 (27) 2.30 2.87 3.52 3.72 -3,086 84.5 1.075

Table 4. GBIS-refined ubiquitin structures with more structures or longer time-steps

Model IDDistance

Rst.eRMSD (Å)

CYANAeRMSD (Å)

GBISrRMSD (Å)

CYANArRMSD (Å)

GBISEnergy

(kcal/mol)Ramachan-

dran (%)Mol-

Probity

5%

81 72 (27) 2.30 3.04 3.52 3.00 -3,120 92.7 0.923

82 72 (27) 2.30 2.31 3.52 2.39 -3,143 93.0 0.852

83 72 (27) 2.30 1.67 3.52 1.80 -3,171 93.0 0.727

84 72 (27) 2.30 0.95 3.52 1.27 -3,227 93.4 0.724


system undergoes unfavorable motion on the NMR time-scale, which in turn leads to inaccurate structures. Therefore,determining 3D structures with a limited number of re-straints has a wide range of applications. Adding to previousdata that showed the strength of GBIS for fixing localgeometries,9,17-20 the role of GBIS has been extended intoimproving global folds. Brooks and his colleagues reportedthe applications of GBIS for determining accurate globalfolds in a series of proteins with about 10% of sub-restraintsusing replica-exchange.10,11 In this study, we showed that itis even possible to obtain accurate folds with 5% of sub-restraints using restrained simulated annealing. Because ofthe differences of the datasets, a direct comparison is notstraightforward, but it is clear that our data permitted detail-ed systematic evaluations. Many successful examples thatutilize Rosetta algorithms and sparse NMR restraints indetermining 3D structures have been published.21-23 Com-parison or combined use with Rosetta remains an interestingtopic. Our data provide a meaningful starting point for thedirection.

Acknowledgments. This work was supported by theNational Research Foundation (NRF) grant funded by theKorean government (MSIP) (NRF-2012R1A1A2007246).

References

1. Güntert, P. Eur. Biophys. J. 2009, 38, 129.

2. Lopez-Mendez, B.; Güntert, P. J. Am. Chem. Soc. 2006, 128, 13112. 3. Schmidt, E.; Güntert, P. J. Am. Chem. Soc. 2012, 134, 12817.

4. Ikeya, T.; Takeda, M.; Yoshida, H.; Terauchi, T.; Jee, J. G.;

Kainosho, M.; Güntert, P. J. Biomol. NMR 2009, 44, 261. 5. Ikeya, T.; Jee, J. G.; Shigemitsu, Y.; Hamatsu, J.; Mishima, M.;

Ito, Y.; Kainosho, M.; Güntert, P. J. Biomol. NMR 2011, 50, 137.

6. Güntert, P.; Mumenthaler, C.; Wuüthrich, K. J. Mol. Biol. 1997,273, 283.

7. Brünger, A. T.; Adams, P. D.; Clore, G. M.; DeLano, W. L.; Gros,

P.; Grosse-Kunstleve, R. W.; Jiang, J. S.; Kuszewski, J.; Nilges,

M.; Pannu, N. S.; Read, R. J.; Rice, L. M.; Simonson, T.; Warren,

G. L. Acta Crystallogr. D Biol. Crystallogr. 1998, 54, 905. 8. Xia, B.; Tsui, V.; Case, D. A.; Dyson, H. J.; Wright, P. E. J.

Biomol. NMR 2002, 22, 317.

9. Jee, J. Bull. Kor. Chem. Soc. 2010, 31, 2717.10. Chen, J.; Im, W.; Brooks, C. L., III. J. Am. Chem. Soc. 2004, 126,

16038.

11. Chen, J.; Won, H. S.; Im, W.; Dyson, H. J.; Brooks, C. L., 3rd J.

Biomol. NMR 2005, 31, 59.12. Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.;

Simmerling, C. Proteins 2006, 65, 712.

13. Laskowski, R. A.; Rullmannn, J. A.; MacArthur, M. W.; Kaptein,R.; Thornton, J. M. J. Biomol. NMR 1996, 8, 477.

14. Davis, I. W.; Leaver-Fay, A.; Chen, V. B.; Block, J. N.; Kapral, G.

J.; Wang, X.; Murray, L. W.; Arendall, W. B., 3rd.; Snoeyink, J.;Richardson, J. S.; Richardson, D. C. Nucleic. Acids Res. 2007, 35,

W375.

15. Shen, Y.; Delaglio, F.; Cornilescu, G.; Bax, A. J. Biomol. NMR2009, 44, 213.

16. Sugita, Y.; Okamoto, Y. Chem. Phys. Lett. 1999, 141.

17. Fujiwara, K.; Tenno, T.; Sugasawa, K.; Jee, J. G.; Ohki, I.;Kojima, C.; Tochio, H.; Hiroaki, H.; Hanaoka, F.; Shirakawa, M.

J. Biol. Chem. 2004, 279, 4760.

18. Ohno, A.; Jee, J.; Fujiwara, K.; Tenno, T.; Goda, N.; Tochio, H.;Kobayashi, H.; Hiroaki, H.; Shirakawa, M. Structure 2005, 13,

521.

19. Jee, J.; Ahn, H. C. Bull. Korean Chem. Soc. 2009, 30, 1139.20. Jee, J.; Mizuno, T.; Kamada, K.; Tochio, H.; Chiba, Y.; Yanagi,

K.; Yasuda, G.; Hiroaki, H.; Hanaoka, F.; Shirakawa, M. J. Biol.

Chem. 2010, 285, 15931.21. Raman, S.; Lange, O. F.; Rossi, P.; Tyka, M.; Wang, X.; Aramini,

J.; Liu, G.; Ramelot, T. A.; Eletsky, A.; Szyperski, T.; Kennedy, M.

A.; Prestegard, J.; Montelione, G. T.; Baker, D. Science 2010, 327,1014.

22. Lange, O. F.; Rossi, P.; Sgourakis, N. G.; Song, Y.; Lee, H. W.;

Aramini, J. M.; Ertekin, A.; Xiao, R.; Acton, T. B.; Montelione, G.T.; Baker, D. Proc. Natl. Acad. Sci. USA 2012, 109, 10873.

23. Thompson, J. M.; Sgourakis, N. G.; Liu, G.; Rossi, P.; Tang, Y.;

Mills, J. L.; Szyperski, T.; Montelione, G. T.; Baker, D. Proc. Natl.Acad. Sci. USA 2012, 109, 9875.

Figure 3. Overlaid 20 structures of ubiquitin. Top 20 structures were overlaid with backbone atoms of residues 1-70. Each ensemble waslabeled with corresponding run number. Ribbon diagram shows reference X-ray structure (PDB ID: 1UBQ).

systematic assessment of the effects of an all-atom force field...

Documents