Quality
assessment
of
protein
NMR
structures
§Antonio
Rosato
1,
Roberto
Tejero
2and
Gaetano
T
Montelione
3,4BiomolecularNMRstructuresarenowroutinelyusedin biology,chemistry,andbioinformatics.Methodsandmetrics forassessingtheaccuracyandprecisionofproteinNMR structuresarebeginningtobestandardizedacrossthe biologicalNMRcommunity.Theseincludeboth knowledge-basedassessmentmetrics,parameterizedfromthedatabase ofproteinstructures,andmodelversusdataassessment metrics.Onlineserversareavailablethatprovide
comprehensiveproteinstructurequalityassessmentreports, andeffortsareinprogressbytheworld-wideProteinDataBank (wwPDB)todevelopabiomolecularNMRstructurequality assessmentpipelineaspartofthestructuredeposition process.Thesequalityassessmentmetricsandstandardswill aidNMRspectroscopistsindeterminingmoreaccurate structures,andincreasethevalueandutilityofthesestructures forthebroadscientificcommunity.
Addresses
1MagneticResonanceCenterandDepartmentofChemistry,University
ofFlorence,50019SestoFiorentino,Italy
2
DepartamentodeQui´micaFi´sica,UniversidaddeValencia,AvenidaDr. Moliner50,46100Burjassot,Valencia,Spain
3CenterforAdvancedBiotechnologyandMedicine,Departmentof
MolecularBiologyandBiochemistry,andNortheastStructuralGenomic Consortium,Rutgers,TheStateUniversityofNewJersey,Piscataway, NJ08854,USA
4DepartmentofBiochemistryandMolecularBiology,RobertWood
JohnsonMedicalSchool,Rutgers,TheStateUniversityofNewJersey, Piscataway,NJ08854,USA
Correspondingauthor:Montelione,GaetanoT(guy@cabm.rutgers.edu)
CurrentOpinioninStructuralBiology2013,23:715–724 ThisreviewcomesfromathemedissueonBiophysicalmethods EditedbyWahChiuandGerhardWagner
ForacompleteoverviewseetheIssueandtheEditorial
Availableonline21stSeptember2013
0959-440X/$–seefrontmatter,#2013TheAuthors.ElsevierLtd.All rightsreserved.
http://dx.doi.org/10.1016/j.sbi.2013.08.005
Introduction
Nuclear Magnetic Resonance (NMR) spectroscopy, along with X-ray diffraction and cryo-electron micro-scopy,isoneofthethreemajorexperimentaltechniques providingthree-dimensional(3D)structuresofbiological
macromolecules.Inadditiontoitsuniquerolein charac-terizingbiomoleculardynamics,NMRisroutinelyusedfor structure determinations of small (<20kDa) proteins [1,2,3] and is beginning to be used more routinely for determining structures of larger (20–50kDa)soluble and membrane proteins (e.g. Refs. [4,5,6,7]). NMR-derivedstructuremodelscanbeusedinterchangeablywith models generatedbyX-raycrystallography inmany bio-logicalapplications.Itisthereforenaturalthatmany qual-ity assessment metrics are common between the two techniques.Inaddition,thereisaportfolioofmetricsthat arespecifictoNMR,whichtakeintoaccountthe distinc-tive features of the NMR data used in the structure determinationprocess.
Inthisreview,weoutlinesomeofthemetricsincommon useforproteinNMRstructurequalityassessment.Alarge partofourreviewreflectsrecentlypublished recommen-dations of theworld-wide ProteinData Base(wwPDB) taskforcesonvalidationofbiomolecularstructures deter-minedbyX-raycrystallography[8]andNMRmethods [9].
Knowledge-basedmeasures
Knowledge-based (KB) metrics describe how well the structuremodelconformstoexpectationswithrespectto selectedfeaturesthatcanbeassessedbycomparisonwith theextensivedatabaseofexperimentalstructures.These include bondlengthandbondangledistributions, dihe-dralangle distributions,atomic packing,hydrogenbond geometries,andothergeometricfeatures.Idealvaluesor valuedistributionsarederivedfromstatisticalanalysesof high-resolution X-ray structures, and are generally con-sistent withbasicprinciplesofbiophysicalchemistry. There hasbeen some debate regarding the use of KB informationderivedfromX-raycrystalstructuresof bio-molecules in assessingsolutionNMR structures. There are often differences in the sample conditions used in determining NMR and X-ray structures, and particular conformations from the distributionpresent in solution maybeselectedbytherequirementsofthecrystallattice. Nevertheless,thereisnocogentreasontoadoptdifferent KBparameterdistributionsforassessingsolutionor solid-stateproteinNMRstructureswithrespecttothoseused toassessstructuresdeterminedbyX-raycrystallography. Problems indicated byKBassessments can bemapped onto the 3D structure to identify local hot spots of structuralinaccuracy[10,11].
Themostgeneralproteinmodelassessmenttoolslookat residuepair-distributionfunctions(e.g.PROSA2[12])or
§
Thisisanopen-access articledistributedunderthetermsofthe CreativeCommonsAttribution-NonCommercial-NoDerivativeWorks License, whichpermits non-commercialuse,distribution,and repro-duction inanymedium,providedtheoriginalauthorandsource are credited.
distributions of hydrophobic and hydrophilic residues (e.g. Verify3D [13]) which are characteristic of native protein structures. These analyses are important first steps in structure validation. However, their value is primarily in identifying severely incorrect folds [14], whichrarelyresultfromNMR structure determinations done with high restraint densities. These scores may, however,beimportant whenassessingstructures deter-mined from sparse restraint networks. Non-globular proteinfolds(e.g.coil–coilstructures) mayexhibitpoor PROSA2or Verify3D scoresevenwhenthemodelsare accurate.
DihedralangledistributionsarethemostprominentKB statisticusedinassessingproteinNMRstructures.They aregenerallyreportedasZ-scoresrelativetodistributions observedinhigh-resolutioncrystalstructures,oracrossall structuresthathavebeendepositedinthePDB.Dihedral angle distributionsare generallyreported separately for amino-acidresiduebackboneandsidechains.Backbone f, cdistributionsaregenerally assessedbased on com-pliance with the Ramachandran plot. Historically, the program ProCheck[15] hasbeenused forthis analysis, but recent work using a larger set of X-ray structures determinedathighresolutionsuggeststhatmoreaccurate assessmentscanbemadeusingimprovedbackbonef,c distribution statistics [8,16]. The assessment of side chain dihedral angle distributions (also referred to as rotamer normality) can be more subtle. Protein side chains have been observed to largely adopt standard rotamerstates(g,t,g+)evenwhenburiedinthecores of protein crystal structures. While NMR data can in principle determine accurate side-chain conformations, surfacesidechainsareoftendynamicallyaveraged com-plicating the interpretation of the corresponding NMR data.
Severaltoolshavebeenusedtoassesscoreatompacking, includingtheMolprobity[11]program for assessment of overpacking, and both the Molprobity and Rosetta-Holes [17] programs for assessment of underpacking. Severe atomic overlaps are rarely an issue in NMR structures, unless there are errors in the restraints, becauseoftheuseoflower-distancebounds.High-energy contactscanoccurduetosimplifiedtreatmentsofvander Waalsandhydrogen-bondedinteractions,butare gener-allyrelievedbyenergyrefinement. Underpacking, how-ever,maybeamoregeneralproblemfor NMR-derived structuresthan generallyappreciated,and methodslike RosettaHoles[17]shouldbeakeycomponentofprotein NMRstructure assessment.
TheroleofKB-drivenenergyrefinementindetermining proteinNMR structures, suchas KBpotentials or frag-mentlibraries[18,19],isalsosomewhatcontroversial.In our opinion and experience, appropriate protocols for energy refinement, including the use of KB potentials
and fragment libraries, can significantly improve the accuracy of NMR structures (see e.g. [1,3,20,21]), particularly for larger proteins determined with sparse restraintnetworks[4,5].Approachesbasedonmolecular dynamics can also often provide appreciable improve-mentsinstructurequality [22].
Modelversusdatameasures
Modelversusdata(MvD)metricsdescribehowwellthe NMRstructuremodelmatchesexperimentaldata.MvD qualityassessmentsincludedatathathavebeenusedin the structure generation process and, where feasible, cross-validation using data that have not been used in structuregenerationcalculations.
NMRrestraintanalysis
ThemostgeneralformofMvDvalidationinvolves com-parisonofdistances anddihedral anglesin modelswith thecorrespondingexperimentalrestraints. Table1 pro-vides a standard format used to report such restraint violations[9,20],althoughotherformatsarealso com-mon intheNMRliterature. Thesemetricsprovide an overview ofglobal (or average) restraintviolation stat-istics, as well as information on the most significant outliers. Clusters of restraint violations in regions of the3Dstructuremayindicateerrors inthelocal struc-ture.Methodshavebeendescribedtoconvertbetween restraint formats [20,23,24], allowing initial restraint listsgenerated foronestructuregenerationprogramto beusedwithalternativestructuregenerationprograms. TheNOEcompletenessscore[20,25]isausefulmetric of structural accuracy, assessing the fraction of short distancesinthemodelstructurethatareconsistentwith restraintdataset.
NOESYdata
Restraint analysis has the significant shortcoming that therestraintsarethemselvesinterpretationsofNOESY andotherNMRdata.Accordingly,NMRstructure qual-ityassessmentshouldalsoincludesomemetrics validat-ing models against uninterpretedspectral data. In the caseofNOESYdata,severalmethodshavebeen devel-oped for back-calculating NOESY spectral data (e.g. Refs.[26–28]),althoughtodatenoneofthesehascome into general use. A more rapid, though approximate, approachistocomparemodelswithunassignedNOESY peaklists.TheRPFprogram[29]quantifiesthe agree-ment between the NOESY peak list, chemical shift assignments, and NMR models by calculating recall, precision, and F measures, aswell as a normalized F-measure(calledaDiscriminatingPowerDPscore).The normalized DP score is highly correlated with the accuracyoftheNMRmodel[30,31].
RDC,scalarcoupling,paramagnetic,andSAXSdata
MvD metrics could also include assessment of scalar coupling,residualdipolarcoupling(RDC),chemicalshift
Table1
SummaryofKBand MvSstructuralstatisticsforbacterialprotein Alr2454.Thisanalysis wasgeneratedbythe PSVSproteinNMR structurequalityassessmentserver[10,14].TheseresultsareadoptedfromRef.[66]
Alr2454a
Completenessofresonanceassignmentsb
Backbone(%) 99.4
Sidechain(%) 98.3
Aromatic(%) 96.6
Stereospecificmethyl(%) 100
Conformationallyrestrictingrestraintsc Distancerestraints
Total 2478
Intra-residue(i=j) 688
Sequential(jijj=1) 619
Mediumrange(1<jijj<5) 462
Longrange(jijj5) 709
Dihedralanglerestraints 162
Hydrogenbondrestraints 0
Disulfidebondrestraints 0
No.ofrestraintsperresidue 25.5
No.oflongrangerestraintsperresidue 6.8
Residualrestraintviolationsc
Averageno.ofdistanceviolationsperstructure:
0.1–0.2A˚ 8.75
0.2–0.5A˚ 1.85
>0.5A˚ 0
Largestdistanceviolation(A˚) 0.35
Averageno.ofdihedralangleviolationsperstructure:
1–108 8.75
>108 0
Largestdihedralangleviolation(8) 3.8
NOEcompletenessscore 0.692
Modelqualityc
RMSDbackboneatoms(A˚)d 0.6
RMSDheavyatoms(A˚)d 0.9
RMSDbondlengths(A˚) 0.018
RMSDbondangles(8) 1.1
MolProbityRamachandranstatisticsc,d
Mostfavoredregions(%) 96.8
Allowedregions(%) 3.1
Disallowedregions(%) 0.1
Globalqualityscores(raw/Z-score)c
Verify3D 0.40 0.96
ProsaII 0.66 0.04
ProCheck(phi-psi)d 0.15 0.28
ProCheck(all)d 0.03 0.18
MolProbityclashscore 12.51 0.62
RPFscorese
Recall/precision 0.976 0.934
F-measure/DP-score 0.955 0.817
Modelcontents
Orderedresidueranged 1–100
Totalno.ofresidues 108
BMRBaccessionnumber: 17965
PDBID 2LJWa
aStructuralstatisticscomputedforanensembleof20structures.
b ComputedusingAVSsoftware[67]fromtheexpectednumberofassignableresonances,excluding:highlyexchangeableprotons(N-terminaland
Lysaminogroups,Argguanidogroups,hydroxylsofSer,Thr,Tyr),carboxylcarbonsofAspandGlu,non-protonatedaromaticcarbons,andthe C-terminalHis6tag.Methylprotonsarecountedasasingleassignableresonance.
c
CalculatedusingPSVS1.4[10].RamachandranstatisticswerecalculatedbyMolprobity[11].Averagedistanceviolationswerecalculatedusing thesumoverr6fordegenerateprotonsandstereochemicallydistinctprotonslackingstereospecificassignments.
anisotropy(CSA),paramagneticresonanceenhancement (PRE), paramagnetic pseudo-contact shift (PCS), solid-statedipolarcoupling,and smallangleX-rayorneutron scattering(SAXSorSANS)data.Severaltoolsfor validat-ingstructuresagainstthesedataare available,including methodsforvalidationofproteinstructuresagainstRDC
data[32–34],CSAdata[35]andSAXSdata[36].Residual
dipolar coupling data provide information about the orientations of internuclear bond vectors with respect to the molecular orientation tensor, and hence provide long-range information including, for example, the relative orientations of secondary structure elements, suchashelixtiltangles,whicharesometimesinaccurate in protein NMR structures. The most commonly used MvD metric for these data is the RDC Q-factor [37]. Aromatic RDCs also have been found to provide an importantapproachforvalidatingaccuracyincore struc-turesof proteins[38].
FreeRfactors
Inspite ofextensivediscussionsand ofvarious formu-lationsthathavebeenproposed[25,27–29,37,39],a cross-validationmetricanalogoustothefreeR-factorofX-ray crystallographyhasnotyetbeenbroadlyadoptedbythe biologicalNMRcommunity.Thisisattributable,atleast inpart, tothe sparseness ofNMR data, andthe need tomanually evaluate each individual peak in NOESY spectrawhenapplyingtrulyquantitativemethods.The RDCfreeQ-factor[37]ispotentiallymoreaccessible,as quantifyingRDCdataisgenerallymorestraightforward than the quantification of large numbers of NOESY peaks.
Chemicalshiftdata
ChemicalshiftdatahavegreatpotentialforNMR struc-ture validation. The most straightforward validation involvescomparingexperimentalchemicalshiftdatawith values predicted from the 3D model structure [40–
42,43,44,45,46,47]. Extensive chemical shift data
must be obtained atthe onset of any protein structure determination, and chemical shift data often used indirectlyasrestraintsondihedralangles[48].Several recentpapers havedescribed important progress in cal-culating chemical shifts from molecular models using bothempirical[49,50,51,52]andquantumchemical
[44,47,53,54] approaches. However, systematic, large
scaletestsusingchemicalshiftsasastandardmetricfor proteinNMR structurevalidationarenotyetavailable.
Whichpartsofthestructureshouldbevalidated?
Protein NMR structures are generally represented by ensemblesofconformershavingthesamelevelof agree-mentwiththeexperimentaldata.Differentregionsofthe structureareoftenconvergedtodifferentdegreesamong conformers. In common practice, a distinction is made between well-defined and ill-defined (i.e. not-well-defined) regions [9,20]. In single-domain proteins,
thebackbone or all-heavy-atomrootmean square devi-ationof coordinates(RMSD)computedafter superposi-tionofallthewell-definedregionsistakenasameasured ofstructureprecision.Similarly,theper-residueRMSDis taken as a measure of local precision. Alternative measures of local or global precision include dihedral angle order parameters (DAOP) [55,56,57] and dis-tance variance matrix methods [57,58,59] (Snyder DA, Grullon J, Huang YJ, Tejero R, Montelione GT: Theexpandedfindcoremethodforidentificationofacore atomset forassessment ofprotein structureprediction. Proteins 2013 (submitted for publication). This is an extension of the variance matrix method of Ref. [58] for annotating well-defined versus ill-defined atoms in an NMR ensemble. FindCore also provides criteria for identifyingregionsoftheproteinstructurethatare intern-allywell-defined,butnotwell-definedwithrespecttoone another, guiding the independent superimposition of theseregionsforRMSDcalculations.)(Figure1).Often, there is a good, albeit qualitative, correlation between localrestraint density andlocal precision,implyingthat the ill-defined regions of the structure result from the experimental data providing insufficient information. This variability can be due to local protein dynamics (sothat locallyasingle conformation in solutionin fact does not exist) or to experimental factors limiting the information that can be extracted from spectra (e.g. extensiveresonancedegeneracy).
Theprecision ofNMR structuresindicatedbythe con-vergence across the ensemble of NMR conformers is operational, and does not provide atrue representation of theBoltzmanndistribution ofconformations actually present in the NMR sample. Indeed, NMR structure ensemblesdonotevenprovideastatisticallymeaningful descriptionofthetrueprecisionofcoordinatesgiventhe experimentaluncertaintiesinderivingdistancerestraints. Forexample,fastexchangemaygiverisetoinconsistent restraints, which when simultaneouslysatisfied canpin the local conformational distribution into an unrealisti-callynarrowrangeof conformations[60].
BecauseNMRexperimentsdonotprovideenoughdata tocharacterize them,theconformationsobserved in ill-definedregionslargelyresultfromofthecombinationof random initial conformations with the potential energy functions.Hence,theseregionsshouldnotbeincludedin global structure validation. However, even such ill-defined regions may be restrained to some degree by sparse experimental data, which may be biologically relevant. Ill-defined regions may also contribute to the back-calculationofNMRobservables(e.g.NOESYpeak lists).Hence, in ouropinion allatoms forwhich exper-imentaldataareavailable(includingonlychemicalshift data)shouldbeincludedin theatomiccoordinates that are deposited in the PDB, consistent with recommen-dationsof thewwPDB NMRVTF[9].
Proteinstructurequalityassessmentservers
Recently, several on-line servers havebecome available whichintegratebothKBandMvDassessmentmetricsto providea comprehensiveNMR structurequality assess-mentreport.Web-basedtoolsdescribedinTable2include the CiNG[61,62],Molprobity [11],PSVS [10,14], Vivaldi[63],andResProx[43]servers.These,aswellas other programs for NMR structure quality assessment, have been critically reviewed in a recent publication [57].The wwPDBNMRVTFhasalsorecommended
metrics and standardsfor biomolecular structurequality assessment[8,9],andtheseguidelinesarebeingusedin developingsoftwarepipelinesthatwillgenerate standar-dized structure quality reports for all NMR and X-ray crystalstructuressubmittedtothewwPDB.
Qualitymeasuresversusstructureaccuracy
SatisfactionofKBmetricsisanecessarybutnotsufficient criteria for validating the accuracy of protein NMR structures.Thisconclusionwasillustratedbytherecent Figure2 0.2 0.4 0.6 0.8 1.0 0 0.5 1 1.5 2 2.5 3 3.5 4 DP-Score RMSD (Å) 0.2 0.4 0.6 0.8 1.0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 DP-Score RMSD (Å)
Current Opinion in Structural Biology
CorrelationbetweenbackboneRMSDtothereferencestructure,whichistakenasameasureofstructuralaccuracy,andtheDP-score.Theseresults demonstratethediscriminatingpoweroftheDPscoreindistinguishingaccuratefromless-accurateproteinNMRmodels.Thehorizontalandvertical linesindicateempiricalthresholdsforaccuratestructuremodels[30].Dataarefromthe2010and2013roundsoftheCASD-NMRproject[1,2],which havebeenmerged.Leftpanel:fullrange;rightpanel:expansionofthe0–4.0A˚ RMSDrange.
Figure1
Phe 47
(a) (b)
Phe 20
Phe 36
Current Opinion in Structural Biology
ComparisonofDOAPandvariancedistancematrixresultsforidentifyingwell-definedatomsets.(a)NMRstructureensemblesuperimposition showingresiduesdefinedas‘welldefined’bythePDBStatDAOPanalysis[20]andthoseidentifiedas‘welldefined’bytheFindCore[20,58]method, forproteinSgR42(PDB_id2jz2).Residuesidentifiedas‘welldefined’bybothmethodsareshownindarkblue,thoseidentifiedas‘welldefined’by variancematrixbutnotbyDAOPinlightblue,andthoseidentifiedas‘welldefined’byDAOPbutnotbyvariancematrixingreen.Residuesidentifiedas ‘illdefined’bybothmethodsareshowninyellow.ThebackboneatomsofthecorrespondingX-raycrystalstructure(PDBid3c4s)areshowninred.(b) Expansionshowingatom-specific‘welldefined’(darkblue)and‘illdefined’(yellow)designationsforthesidechainsofresiduesPhe20,Phe36,and Phe47inproteinSgR42.Thisimagedemonstratesthevalueofatom-specificdesignationsofwell-definedregionsofaproteinNMRstructure. AdoptedfromRef.[20].
CriticalAssessmentofProteinStructureDetermination byNMR(CASD-NMR)[1,2],inwhichNMRdatawas publicly released, and the results of automated NOE assignmentandstructuregenerationmethodswere com-pared inablindfashion with manually refined protein NMRstructures. Inmanycases,modelswithexcellent energeticswereinfactquitedifferentfromthemanually refined ‘gold standard’ structures. However, energetic considerationscanidentifyinaccuraterestraintdata,and thus guide the experimentalist to more accurate interpretationsoftherawNMRdataintermsof exper-imentalrestraints[20,21].
The CASD-NMR experiment also compared corre-lations between several KB and MvD metrics with structural accuracy. In thisanalysis, the DP score was observed to provide the highest correlation between submittedmodelsandthemanuallyrefinedtarget
struc-ture (Figure 2). Although the DPscore has significant
shortcomingsto the extent that it uses an interpreted
NOESYpeaklistratherthanrawNOESYspectraldata, theseresultsdemonstratetheimportanceofcombining KBandMvDvalidationscoresinstructurequality assess-ment. For example, the DP score has recently been combined with several KB scores by linear regression analysis to provide an ‘equivalent resolution’ metric [31,64].
Areasofcurrentresearch
Thetrendtowardsdetermininglargerbiomolecular struc-turesusingsparseNMRdataandhybridmethods incor-porating small angle X-ray scattering, chemical cross links, and other data, requires increasing use of KB informationandpotential energyfunctionsin the struc-turedeterminationprocess.Somesolid-stateNMR struc-ture determinations may also rely on relatively sparse networksof experimentaldata. Thesemethodsprovide lessindependentinformationfor useincross-validation, creating important challenges to the field in terms of robuststructure qualityassessment.
Table2
Webaccessibleserversprovidingknowledge-based(KB)and/ormodelversusdata(MvD)proteinNMRstructurequalityassessment reports
Server url Validationstatisticsprovided
CING[62]v.1.0:CommonInterface forNMRStructureGeneration
https://nmr.le.ac.uk/icing KB:Bondlengthsandbondangles,backboneRamachandran distributions,peptideomegas,packing,sidechainrotamer normalility,disulfides,salt-bridges,chemicalshiftvalidations. Providesresidue-specificROCscores[68].UsesDSSP[69,70], WHATIF[71],PROCHECK[15,72],andBMRBchemicalshift validation
MvD:DOAPanalysis.Restraintviolationanalysis,redundancy andduplicatedrestraints,backcalculationofchemicalshiftfrom structure.UsesSHIFTX[49]andVASCO[73]
Molprobity[11] http://molprobity.biochem.duke.edu/ KB:BackboneRamachandrandistributions,peptideomegas, packing,H-bondsatisfaction,sidechainrotamernormalility MvD:RDCanalysiswithRDCvis[74]
PSVS[10,14]v.1.5.Protein StructureValidationSoftwaresuite.
http://psvs.nesg.org/ KB:Bondlengthsandbondangles,backboneRamachandran distributions,peptideomegas,packing,sidechainrotamer normality,chemicalshiftvalidationandcompletenessusingAVS [67].ProvidesZscoresrelativetohigh-resolutioncrystal structuresforProsaII[12],Verify3D[13],Procheck_bb[15,72], Procheck_all[15,72],andMolProbity[11].AlsousesDSSP [69,70],PDBStat[20],LACS[75],andPDBClashscore (deposit.rcsb.org/validate)
MvD:DAOPorVarianceMatrixanalysis.Restraintviolation analysis,redundancyandduplicatedrestraints,NOE
completeness,NOEDPscores,RDCQscores,andGLM-RMSD [31]‘equivalentresolution’score.Providesmappingof RamachandranoutliersandRPFviolationsonto3Dstructure. AlsousesPDBStat[20],FindCore[20,58]andRPF[29,30] software
ResProx[43]ResolutionbyProxy http://www.resprox.ca/ KB:Assesses‘equivalentresolution’basedon25proteinfeatures includingbackboneRamachandrandistributions,peptide omegas,H-bondgeometry,overandunderpacking.UsesVader [40],PROSESS[42],Molprobity[11],RosettaHoles[17],and GeNMR[41]software
Vivaldi[63]v.1:Visualizationand validationofbiomacromolecular NMRstructuresfromthePDB
http://www.ebi.ac.uk/pdbe/vivaldi/ KB:Bondlengthsandbondangles,peptideomegas,chemical shiftvalidations,backboneRamachandrandistributions,CING [61]ROCscores.UsesWHATIF[71],CING[61],andVASCO[73] MvD:DAOPandVarianceMatrixanalysis.Restraintviolationand RDCanalysis.UsesNMRCore[76]
QualityassessmentconsiderationsforNMRstructuresof nucleicacidsandcarbohydrates,whilegenerallysimilarto those forproteins,havenotyetbeenextensively inves-tigated.Methodsalsoneedtobedevelopedfor generat-ingBoltzmannensemblesofconformersthatbestsatisfy theexperimentaldata(e.g.[37,65]),ratherthanfittingall the datato asingle conformer multipletimes,as isthe common practice. Ensemble-averaging interpretations are particularlyimportant in highly dynamic regionsof abiomolecularstructure,andnewmethodsareneededfor qualityassessmentoftheensemblesofmodelsproposed for suchdisorderedregions.
Conclusions
AhighqualityNMR-derivedstructureshouldmeet mini-mal standards based on awide range of KB and MvD validationassessment metrics.Nosingle metricscore is sufficientto validateaproteinNMRstructure.Usersof biomolecular NMR structuresalsoneed to beinformed aboutwhichpartsofthestructurearewell-definedbythe data, and which are less-reliably defined in terms of a unique structure. This information can be critical in interpretingstructure–functionrelationships.
Structure determinations byNMR are often marginally underdetermined,andrelyineithersubtleorsubstantial ways on KB information, including at the very least standard values of bond lengths and bond angles. Although they are useful for biological studies, and in some cases quite accurate,the relativelylow density of data constituting a typical protein NMR structure pre-sentschallengestocrossvalidation,usingpartofthedata to generatethestructureandanotherpartof thedatato validatethestructure.Inthisregard,chemicalshiftdata, which are available for most atoms reported in modern biomolecularNMRstructures,holdpromiseforproviding a robust and general approach for data-based protein structure qualityassessments.
Acknowledgements
WethankallofthescientistsparticipatingintheCASD-NMRProject [1,2]forgeneratingthedatasummarized(Figure2).SpecialthankstoJ. Aramini,L.Ferella,Y.J.Huang,B.Mao,D.Snyder,andthemembersofthe wwPDBNMRValidationTaskForceforhelpfulscientificdiscussions.This workwassupportedbygrantsfromtheProteinStructureInitiative— BiologyoftheNationalInstitutesofHealth,GrantU54-GM094597(to G.T.M),bytheCONSOLIDERINGENIOCSD2010-00065and GeneralitatValencianaPROMETEO2011/008(toR.T),andbythe EuropeanCommunityFP7e-Infrastructure‘WeNMR’project,Grant 261572(toA.R.).
References
and
recommended
reading
Papersofparticularinterest,publishedwithintheperiodofreview, havebeenhighlightedas:ofspecialinterest ofoutstandinginterest 1.
RosatoA,AraminiJM,ArrowsmithC,BagariaA,BakerD,CavalliA, DoreleijersJF,EletskyA,GiachettiA,GuerryP,GutmanasAetal.: Blindtestingofroutine,fullyautomateddeterminationof proteinstructuresfromNMRdata.Structure2012,20:227-236.
ThispaperreportsthefirstresultsoftheCASD-NMRinitiative.Inthis collaborativeeffort,rawdatawereprovidedtoeightteamsdeveloping softwareforautomatedNMRstructuredetermination,whilethe corre-sponding3Dcoordinateswerenotreleased.Anindependentteamthen compared theautomatically generated structures with thereference structure, manually determined from the same data. All automated NOESY-basedmethodsperformedwellonalltargets.
2. RosatoA,BagariaA,BakerD,BardiauxB,CavalliA,DoreleijersJF, GiachettiA,GuerryP,GuntertP,HerrmannT,HuangYJetal.: CASD-NMR:criticalassessmentofautomatedstructure determinationbyNMR.NatMethods2009,6:625-626.
3.
MaoB,GuanR,MontelioneGT:Improvedtechnologiesnow routinelyprovideproteinNMRstructuresusefulformolecular replacement.Structure2011,19:757-766.
ThispaperdemonstratesthatNMRstructuresdeterminedandvalidated usingmodernmethodsprovidecoordinateswhicharesufficientlyaccurate to allow their routine use in determining phases of corresponding or homologousX-raycrystalstructuresbymolecularreplacementmethods. Insomecases, however, theaccuracyoftheseNMRstructures,asassessed bytheirphasingpower,isimprovedbyKBrefinementwithRosetta.
4.
LangeOF,RossiP,SgourakisNG,SongY,LeeHW,AraminiJM, ErtekinA,XiaoR,ActonTB,MontelioneGT,BakerD:
Determinationofsolutionstructuresofproteinsupto40kDa usingcs-rosettawithsparseNMRdatafromdeuterated samples.ProcNatlAcadSciUSA2012,109:10873-10878.
ThispaperdemonstratestheuseofKBmethods, includingfragment libraries, in determining accurate protein structures by NMR, using perdeuterated,methylprotonatedproteins,inthe20–40kDasizerange. RelatedresultsarepresentedinRef.[5].
5.
RamanS,LangeOF,RossiP,TykaM,WangX,AraminiJ,LiuG, RamelotTA,EletskyA,SzyperskiT,KennedyMAetal.:NMR structuredeterminationforlargerproteinsusing backbone-onlydata.Science2010,327:1014-1018.
ThispaperdemonstratestheuseofKBmethods, includingfragment libraries,indeterminingaccurateproteinstructuresbyNMRusing per-deuterated proteinswithoutmethyl protonation inthe20–40kDasize range.RelatedresultsarepresentedinRef.[4].
6. HillerS,WagnerG:TheroleofsolutionNMRinthestructure determinationsofvdac-1andothermembraneproteins.Curr OpinStructBiol2009,19:396-401.
7. HillerS,GarcesRG,MaliaTJ,OrekhovVY,ColombiniM, WagnerG:Solutionstructureoftheintegralhumanmembrane proteinvdac-1indetergentmicelles.Science2008,
321:1206-1210.
8.
ReadRJ,AdamsPD,Arendall WB3rd,BrungerAT,EmsleyP, JoostenRP,KleywegtGJ,KrissinelEB,LuttekeT,OtwinowskiZ, PerrakisAetal.:Anewgenerationofcrystallographic validationtoolsfortheproteindatabank.Structure2011, 19:1395-1412.
AnoverviewoftherecommendationsofwwPDBX-rayCrystallography ValidationTaskForce,providingguidelinesforbothKBandMvD valida-tion ofbiomolecular structures determined by X-ray crystallography. Recommendationsarealsooutlinedforthepresentationoftheseresults instructurevalidationreportstobepreparedbythewwPDB.
9.
MontelioneGT,NilgesM,BaxA,Gu¨ntertP,HerrmannT, RichardsonJS,SchwietersCD,VrankenWF,VuisterGW, WishartDS,BermanHMetal.:Recommendationsofthewwpdb NMRvalidationtaskforce.Structure2013,21:1563-1570.
AnoverviewofconsensusconclusionsbythewwPDBNMRValidation TaskForceprovidingrecommendationsforvalidationofproteinNMR structures inthreephases:(i) validationmethodsreadytoimplement immediatelyaspartofaconsensusNMRstructurevalidationpipeline,(ii) validation methods which work well, but havenot yet been broadly adoptedby thebiomolecularNMRcommunity,and(iii)keyissuesof biomolecularstructurevalidationwhichrequireadditionalresearch.
10.
BhattacharyaA,TejeroR,MontelioneGT:Evaluatingprotein structuresdeterminedbystructuralgenomicsconsortia. Proteins2007,66:778-795.
ThispaperdescribesthePSVSserver,apopulartoolforNMRstructure validation.PSVSprovidesan extensiveanalysisofbothKBandMvD validation metrics for protein NMRstructures, and a comprehensive structurevalidationreport.Althoughtheoriginalpaperwas published in 2007, the serveris updatedregularly by the introduction of new validation metrics. See alsoRef. [14] exploring theuse ofPSVSfor structurequalityassessmentofhomologymodels.
11.
ChenVB,ArendallWB3rd,HeaddJJ,KeedyDA,ImmorminoRM, KapralGJ,MurrayLW,RichardsonJS,RichardsonDC: Molprobity:all-atomstructurevalidationformacromolecular crystallography.ActaCrystallogr.Sect.D,Biol.Crystallogr.2010, 66(Pt1):12-21.
MolprobitysoftwareisakeycomponentofmostproteinNMRstructure validationservers.ItprovidescomprehensiveKBvalidationofprotein structures,andincludestheRDCvismodule(Ref.[75])forvisualvalidation ofproteinNMRstructuresusingRDCdata.
12. SipplMJ:Recognitionoferrorsinthree-dimensional structuresofproteins.Proteins1993,17:355-362.
13. LuthyR,BowieJU,EisenbergD:Assessmentofproteinmodels withthree-dimensionalprofiles.Nature1992,356:83-85.
14. BhattacharyaA,WunderlichZ,MonleonD,TejeroR,
MontelioneGT:Assessingmodelaccuracyusingthehomology modelingautomaticallysoftware.Proteins2008,70:105-118.
15. LaskowskiRA,RullmannnJA,MacArthurMW,KapteinR, ThorntonJM:Aquaandprocheck-NMR:programsforchecking thequalityofproteinstructuressolvedbyNMR.JBiomolNMR 1996,8:477-486.
16. Arendall WB3rd,TempelW,RichardsonJS,ZhouW,WangS, DavisIW,LiuZJ,RoseJP,CarsonWM,LuoM,RichardsonDC:A testofenhancingmodelaccuracyinhigh-throughput crystallography.JStructFunctGenomics2005,6:1-11. 17. ShefflerW,BakerD:Rosettaholes:rapidassessmentofprotein
corepackingforstructureprediction,refinement,design,and validation.ProteinSciPublProteinSoc2009,18:229-239.
18.
vanderSchotG,ZhangZ,VernonR,ShenY,VrankenWF, BakerD,BonvinAM,LangeOF:Improving3Dstructure predictionfromchemicalshiftdata.JBiomolNMR2013, 57:27-35.
Thisworkdemonstratesanewandimprovedfragmentpickerandthe iterative sampling algorithm RASREC toprovide significant improve-mentsinconvergenceandaccuracyoftheCS-Rosettamethod(Ref. [19]).Theyalsointroduceimprovedcriteriaforassessingtheaccuracyof themodelsproducedbyCS-Rosetta.
19. ShenY,LangeO,DelaglioF,RossiP,AraminiJM,LiuG,EletskyA, WuY,SingarapuKK,LemakA,IgnatchenkoAetal.:Consistent blindproteinstructuregenerationfromNMRchemicalshift data.ProcNatlAcadSciUSA2008,105:4685-4690.
20.
TejeroR,SnyderD,MaoB,AraminiJM,MontelioneGT:PDBStat: auniversalrestraintconverterandrestraintanalysissoftware packageforproteinNMR.JBiomolNMR2013,56:337-351.
PDBStatsoftwareprovidesarobusttoolforMvDvalidations,including extensiverestraintanalysisandconversionutilities.Thesoftwarecanalso beusedtoconvertan initialrestraintlistintorestraintlistsforother software(e.g.CYANArestraintstoRosettarestraints),allowingvalidation by consensus structuregeneration with multiple software packages. PDBStatalsoprovidesMvDvalidationagainstRDCdata.
21. RamelotTA,RamanS,KuzinAP,XiaoR,MaLC,ActonTB, HuntJF,MontelioneGT,BakerD,KennedyMA:ImprovingNMR proteinstructurequalitybyrosettarefinement:amolecular replacementstudy.Proteins2009,75:147-167.
22. BertiniI,CaseDA,FerellaL,GiachettiA,RosatoA:Agrid-enabled webportalforNMRstructurerefinementwithamber. Bioinformatics2011,27:2384-2390.
23. NederveenAJ,DoreleijersJF,VrankenW,MillerZ,SpronkCA, NabuursSB,GuntertP,LivnyM,MarkleyJL,NilgesM,UlrichEL etal.:Recoord:arecalculatedcoordinatedatabaseof500+ proteinsfromthepdbusingrestraintsfromthe
biomagresbank.Proteins2005,59:662-672.
24. VrankenWF,BoucherW,StevensTJ,FoghRH,PajonA,LlinasM, UlrichEL,MarkleyJL,IonidesJ,LaueED:Theccpndatamodel forNMRspectroscopy:developmentofasoftwarepipeline. Proteins2005,59:687-696.
25. DoreleijersJF,RavesML,RullmannT,KapteinR:Completeness ofnodesinproteinstructure:astatisticalanalysisofNMR.J BiomolNMR1999,14:123-132.
26. BorgiasBA,JamesTL:Two-dimensionalnuclearoverhauser effect:completerelaxationmatrixanalysis.MethodsEnzymol 1989,176:169-183.
27. GronwaldW,KirchhoferR,GorlerA,KremerW,GanslmeierB, NeidigKP,KalbitzerHR:Rfac,aprogramforautomatedNMR r-factorestimation.JBiomolNMR2000,17:137-151.
28. ZhuL,DysonHJ,WrightPE:Anoesy-hsqcsimulationprogram, spirit.JBiomolNMR1998,11:17-29.
29. HuangYJ,PowersR,MontelioneGT:ProteinNMRrecall, precision,andFmeasurescores(RPFscores):structure qualityassessmentmeasuresbasedoninformationretrieval statistics.JAmChemSoc2005,127:1665-1674.
30.
HuangYJ,RosatoA,SinghG,MontelioneGT:Rpf:aquality assessmenttoolforproteinNMRstructures.NucleicAcidsRes 2012,40(WebServerissue):W542-W546.
ThisarticledescribesawebserverforthecalculationofRPFscoresfor NMRstructures.ThisMvDmeasureexploitsunassignedNOESYdata. Oneoftheoutputmeasures,theDP-score,isausefulpredictorofthe accuracyofthequerystructure(seealsoRef.[29]).
31.
BagariaA,JaravineV,HuangYJ,MontelioneGT,GuntertP: Proteinstructurevalidationbygeneralizedlinearmodel root-mean-squaredeviationprediction.ProteinSciPublProteinSoc 2012,21:229-238.
Thisarticledescribesapredictorofthequalityofproteinstructuresthat integratesavarietyofdifferentmeasuresintoasingle‘equivalent resolu-tion’score.Theauthorsdemonstratethatthisscoreprovidesan indica-tionofstructuralaccuracy,providedastheestimatedRMSDfromthetrue structure.
32. CloreGM,RobienMA,GronenbornAM:Exploringthelimitsof precisionandaccuracyofproteinstructuresdeterminedby nuclearmagneticresonancespectroscopy.JMolBiol1993, 231:82-102.
33. ValafarH,PrestegardJH:Redcat:aresidualdipolarcoupling analysistool.JMagnReson2004,167:228-241.
34. BrysonM,TianF,PrestegardJH,ValafarH:Redcraft:atoolfor simultaneouscharacterizationofproteinbackbonestructure andmotionfromrdcdata.JMagnReson2008,191:322-334.
35. CornilescuG,MarquardtJL,OttigerM,BaxA:Validationof proteinstructurefromanisotropiccarbonylchemicalshiftsin adiluteliquidcrystallinephase.JAmChemSoc1998, 120:6836-6837.
36. GrishaevA,WuJ,TrewhellaJ,BaxA:Refinementof multidomainproteinstructuresbycombinationofsolution small-angleX-rayscatteringandNMRdata.JAmChemSoc 2005,127:16621-16628.
37. CloreGM,SchwietersCD:Howmuchbackbonemotionin ubiquitinisrequiredtoaccountfordipolarcouplingdata measuredinmultiplealignmentmediaasassessedby independentcross-validation? JAmChemSoc2004, 126:2923-2938.
38.
SathyamoorthyB,SingarapuKK,GarciaAE,SzyperskiT:Protein conformationalspacepopulatedinsolutionprobedwith aromaticresidualdipolar13C–1Hcouplings.Chembiochem 2013,14:684-688.
AninnovativeapproachdemonstratingthecorrelationofaromaticTyrand Phe13C–1HRDCmeasurementswithproteinstructureaccuracy.These dataarerelativelyeasytomeasureandprovideimportantinformation usefulforeitherstructuredetermination,refinement,orvalidation.
39. BrungerAT,CloreGM,GronenbornAM,SaffrichR,NilgesM: Assessingthequalityofsolutionnuclearmagneticresonance structuresbycompletecross-validation.Science1993, 261:328-331.
40. WillardL,RanjanA,ZhangH,MonzaviH,BoykoRF,SykesBD, WishartDS:Vadar:awebserverforquantitativeevaluationof proteinstructurequality.NucleicAcidsRes2003,31:3316-3319.
41. BerjanskiiM,TangP,LiangJ,CruzJA,ZhouJ,ZhouY,BassettE, MacDonellC,LuP,LinG,WishartDS:GeNMR:awebserverfor rapidNMR-basedproteinstructuredetermination.Nucleic AcidsRes2009,37(WebServerissue):W670-W677.
42. BerjanskiiM,LiangY,ZhouJ,TangP,StothardP,ZhouY,CruzJ, MacDonellC,LinG,LuP,WishartDS:Process:aprotein structureevaluationsuiteandserver.NucleicAcidsRes2010, 38(WebServerissue):W633-W640.
43.
BerjanskiiM,ZhouJ,LiangY,LinG,WishartDS: Resolution-by-proxy:asimplemeasureforassessingandcomparingthe overallqualityofNMRproteinstructures.JBiomolNMR2012, 53:167-180.
ResProx calculates equivalent X-ray crystal structureresolution of a proteinNMRstructurefrom25measurableproteinfeatures.Itcanbe used to identify under-restrained, poorly refinedor inaccurate NMR structures.
44.
VilaJA,AraminiJM,RossiP,KuzinA,SuM,SeetharamanJ, XiaoR,TongL,MontelioneGT,ScheragaHA:Quantumchemical 13c(alpha)chemicalshiftcalculationsforproteinNMR structuredetermination,refinement,andvalidation.ProcNatl AcadSciUSA2008,105:14389-14394.
QuantumchemistrycalculationsbytheprogramCheShiftareusedto predict 13Ca chemical shifts for various residue types, flanking sequences,andbackbone f,Cvalues.Theresultingweb-accessible database provides a rapid approach for protein backbone structure qualityassessment.
45. MartinOA,VillegasME,VilaJA,ScheragaHA:Analysisof 13calphaand13cbetachemicalshiftsofcysteineandcystine residuesinproteins:aquantumchemicalapproach.JBiomol NMR2010,46:217-225.
46. VilaJA,SerranoP,WuthrichK,ScheragaHA:Sequential nearest-neighboreffectsoncomputed13calphachemicalshifts.J BiomolNMR2010,48:23-30.
47.
MartinOA,VilaJA,ScheragaHA:Cheshift-2:graphicvalidation ofproteinstructures.Bioinformatics2012,28:1538-1539.
CheShift-2providesagraphicaluserinterfacetoidentifypossibleflawsin anNMRstructurebasedonquantummechanicalcomputationof13Ca shifts.Theshiftsarecalculatedasaveragesoverallmembersofthe ensemble,andtheiraveragecomparedtotheexperimentaldata. 48.
ShenY,BaxA:Sparta+:amodestimprovementinempirical NMRchemicalshiftpredictionbymeansofanartificialneural network.JBiomolNMR2010,48:13-22.
TogetherwithShiftX2,thisisoneofthemostcommonlyusedandreliable methodsforNMRchemicalshiftpredictions.
49.
HanB,LiuY,GinzingerSW,WishartDS:Shiftx2:significantly improvedproteinchemicalshiftprediction.JBiomolNMR 2011,50:43-57.
Together withSparta+, thisis oneofthemost commonly usedand reliablemethodsforNMRchemicalshiftpredictions.SeealsoRef.[50].
50. WangB,WangY,WishartDS:Aprobabilisticapproachfor validatingproteinNMRchemicalshiftassignments.JBiomol NMR2010,47:85-99.
51.
SahakyanAB,VrankenWF,CavalliA,VendruscoloM:Using side-chainaromaticprotonchemicalshiftsforaquantitative analysisofproteinstructures.AngewChemIntEdEngl2011, 50:9620-9623.
The authorsdemonstrate the useofaromatic chemical shiftdata in distinguishingcorrectstructuralmodelsfromincorrectones.Theyalso exploreitsusetodetectdifferencescausedbycofactororligandbinding, orbysequencealterationsbetweenstructures.
52.
SahakyanAB,VrankenWF,CavalliA,VendruscoloM: Structure-basedpredictionofmethylchemicalshiftsinproteins.J BiomolNMR2011,50:331-346.
ThispaperdescribestheCH3Shiftmethodofperformingstructure-based predictionsofmethylchemicalshifts.Themethodachieveshigh accu-racyinthepredictionsof1HchemicalshiftsforAla,Thr,Val,LeuandIle
methylgroups.Thisapproachisusefulbothforassignmentofmethyl groups from 3D protein structure, and for validation of 3D protein structures, particularly thosedeterminedusing perdeuterated,methyl protonatedsamples.
53. VilaJA,SueSC,FraserJS,ScheragaHA,DysonHJ:Cheshift-2 resolvesalocalinconsistencybetweentwoX-raycrystal structures.JBiomolNMR2012,54:193-198.
54. VilaJA,ArnautovaYA,VorobjevY,ScheragaHA:Assessingthe fractionsoftautomericformsoftheimidazoleringofhistidine inproteinsasafunctionofph.ProcNatlAcadSciUSA2011, 108:5602-5607.
55. HybertsSG,GoldbergMS,HavelTF,WagnerG:Thesolution structureofeglincbasedonmeasurementsofmanynodes andcouplingconstantsanditscomparisonwithX-ray structures.ProteinSci1992,1:736-751.
56.
KirchnerDK,GuntertP:Objectiveidentificationofresidue rangesforthesuperpositionofproteinstructures.BMC Bioinformatics2011,12:170.
TheCYRANGEmethodfortheidentificationofwell-orderedregionsin NMRstructurescombinesdifferentaspectsofapproachesbasedonboth dihedralangleorderparametersanddistancematrices,andintegrates themwithnewideas.Overall,theaimistodefineresiduerangesthat containfewgapsandcovermostofthesequence.Thismethodhasbeen adoptedbythewwPDBNMRasastandardfor definingwell-defined residuerangesforqualityassessment.
57.
VuisterGW,FoghRH,HendrickxPM,DoreleijersJF,GutmanasA: AnoverviewoftoolsforthevalidationofproteinNMR structures.JBiomolNMR2013.
Thisreview reportsa comparisonofthe applicationofninedifferent validationtoolsontwoexampleNMRstructures.Interestingly,thesame property as analyzed by different tools is not always consistently assessedasbeinggoodorbad.
58. SnyderDA,MontelioneGT:Clusteringalgorithmsforidentifying coreatomsetsandforassessingtheprecisionofprotein structureensembles.Proteins2005,59:673-686.
59. AndrecM,SnyderDA,ZhouZ,YoungJ,MontelioneGT,LevyRM: Alargedatasetcomparisonofproteinstructuresdetermined bycrystallographyandNMR:statisticaltestforstructural differencesandtheeffectofcrystalpacking.Proteins2007, 69:449-465.
60. TejeroR,Bassolino-KlimasD,BruccoleriRE,MontelioneGT: Simulatedannealingwithrestrainedmoleculardynamics usingcongen:energyrefinementoftheNMRsolution structuresofepidermalandtype-alphatransforminggrowth factors.ProteinSci1996,5:578-592.
61.
DoreleijersJF,SousadaSilvaAW,KriegerE,NabuursSB, SpronkCA,StevensTJ,VrankenWF,VriendG,VuisterGW:CiNG: anintegratedresidue-basedstructurevalidationprogram suite.JBiomolNMR2012,54:267-283.
CiNGisapowerfulweb-basedtoolforNMRstructurevalidation.CiNG providesanextensive analysisofbothKBandMvD parameters,and provideresidue-specificqualityscores,ROCscores,whichareusefulfor proteinstructurevalidationandrefinement.
62.
DoreleijersJF,VrankenWF,SchulteC,MarkleyJL,UlrichEL, VriendG,VuisterGW:NRG-CiNG:integratedvalidationreports ofremediatedexperimentalbiomolecularNMRdataand coordinatesinwwpdb.NucleicAcidsRes2012,40(Database issue):519-524.
NRG-CiNGcontainsvalidationreportsbyCiNGforallNMRstructuresin thePDB,basedonremediated,thatis,correctedforavarietyofpossible errorsorinconsistencies,NMRrestraints.
63.
HendrickxPM,GutmanasA,KleywegtGJ:Vivaldi:visualization andvalidationofbiomacromolecularNMRstructuresfromthe pdb.Proteins2013,81:583-591.
Vivaldiprovidesaweb-basedgraphicalinterfacetointeractivelyvisualize andanalyzeNMRstructures andthecorrespondingrestraints that is directlyembeddedinthePDBe.VivaldialsopermitstheanalysisofRDC data.
64. BagariaA,JaravineV,GuntertP:Estimatingstructurequality trendsintheproteindatabankbyequivalentresolution. ComputBiolChem2013,46C:8-15.
65. Lindorff-LarsenK,BestRB,DepristoMA,DobsonCM, VendruscoloM:Simultaneousdeterminationofprotein structureanddynamics.Nature2005,433:128-132.
66. AraminiJM,PetreyD,LeeDY,JanjuaH,XiaoR,ActonTB, EverettJK,MontelioneGT:SolutionNMRstructureofalr2454 fromnostocspPcc7120,thefirststructuralrepresentativeof pfamdomainfamilypf11267.JStructFunctGenomics2012, 13:171-176.
67. MoseleyHN,SahotaG,MontelioneGT:Assignmentvalidation softwaresuitefortheevaluationandpresentationofprotein resonanceassignmentdata.JBiomolNMR2004,
28:341-355.
68. DoreleijersJF,SousadaSilvaAW,KriegerE,NabuursSB, SpronkCA,StevensTJ,VrankenWF,VriendG,VuisterGW::Cing: Anintegratedresidue-basedstructurevalidationprogram suite.JBiomolNMR2012.
69. KabschW,SanderC:Dictionaryofproteinsecondarystructure: patternrecognitionofhydrogen-bondedandgeometrical features.Biopolymers1983,22:2577-2637.
70. JoostenRP,teBeekTA,KriegerE,HekkelmanML,HooftRW, SchneiderR,SanderC,VriendG:Aseriesofpdbrelated databasesforeverydayneeds.NucleicAcidsRes2011, 39(Databaseissue):D411-D419.
71. VriendG:Whatif:amolecularmodelinganddrugdesign program.JMolGraph1990,8:52-5629.
72. LaskowskiRA,MacArthurMW,MossDS,ThorntonJM: Procheck:aprogramtocheckthestereochemicalqualityof proteinstructures.JApplCrystallogr1993,26:283-291.
73. RiepingW,VrankenWF:Validationofarchivedchemicalshifts throughatomiccoordinates.Proteins2010,78:2482-2489.
74. BlockJN,ZielinskiDJ,ChenVB,DavisIW,VinsonEC,BradyR, RichardsoJS,RichardsonDC:Kinimmerse:macromolecularvr forNMRensembles.SourceCodeBiolMed2009,4:3.
75. WangL,EghbalniaHR,BahramiA,MarkleyJL:Linearanalysisof carbon-13chemicalshiftdifferencesanditsapplicationtothe detectionandcorrectionoferrorsinreferencingandspin systemidentifications.JBiomolNMR2005,32:13-22.
76. KelleyLA,GardnerSP,SutcliffeMJ:Anautomatedapproachfor definingcoreatomsanddomainsinanensembleof NMR-derivedproteinstructures.ProteinEng1997,10:737-741.