NMR Structure Calculation: Insights and Practical Tips Trent Bjorndahl Faculty of Pharmacy Universit

1
NMR Structure CalculationInsights and
Practical TipsTrent BjorndahlFaculty of
PharmacyUniversity of AlbertaJune 17, 2005
2
Structure Calculation

• Cornerstone of Structural Biology
• Necessary to formulate valid hypotheses
• Protein Protein Interaction
• Protein Ligand Interactions
• Cofactors, substrates and metabolites
• Drug Design
• Dynamics

3
NMR Structures

• Importance
• 24,747 Protein structures in PDB (June 16, 2005)
• 3,814 derived from NMR (15)
• Account for gt 25 of novel folds
  (Nederveen et al., 2005 Proteins Struc.
  Func and BioInform)
• Genuine or erroneous?

4
Structural Error

• Inexperienced or Under-supervised
• Computers as Black Boxes
• New Methods not adopted
• Intermediate models not subjected to systematic
  quality analysis
• Use of quality indicators that are strongly
  correlated with restrained parameters

5
Structural Error

• NMR methods are heterogeneous
• RECOORD (April 2005)
• New methods of refinement still not adopted
• NMR structures still not choice for drug design

6
Tips Data Collection

• Homogeneous Sample conditions
• Protein concentration, buffer, salt, DTT, DSS,
  EDTA, NaN3
• Samples Need three
• Water, D2O, perdeuterdated
• Collect Additional Experiments and Data
• D2O exchange for H - Bonds
• Prolines, aromatics, histidines, N and Q
  side-chains, methionine methyls

7
Additional Data

• HRV14-3C
• 4 Histidines (20 protons)
• 4 Tyrosines (16 protons)
• 6 Phenylalanines (30 protons)
• 3 Methionines (3 protons)
• 14 N and Q residues (28 protons)
• 7 Prolines (35 protons)

132 protons of 1052 12.5
8
Histidines

• Tautomeric State
• 15N HSQC
• 4 to 20ppm in proton
• Up to 280ppm for nitrogen

9
Histidines and Aromatics

• 13C HSQC (aromatic region)

13C NOESY-HSQC
13C HSQC
13C NOESY-HSQC
10
Prolines
g_hc_co_nnh
g_c_co_nnh
Cis Peptide Bonds ? - ? Carbon Chemical Shift
Difference
Cis Trans
13C HSQC
Delta Protons
11
Methionine Methyls
CCH-TOCSY
13C NOESY- HSQC
HCCH-TOCSY

• No Connectivity
• Strong signal in 13C-NOESY 13C-HSQC

12
Glutamine and Asparagine Side Chains
Amide group on side chain
15N HSQC
HNCO
HNCACB
13
Hydrogen Bonds
D2O Exchange 15N HSQC IDs Residues involved in
H-bonds, but not to which residues! Assign only
after NOE only structure is calculated Do
sequential series of experiments to identify
cut-off time for positive identification
Post 4hrs
Post 2hrs
Nitrogen
N114 - In AP Beta Sheet Solvent exposed - cannot
form H-Bond
14
Hydrogen Bond Restraints

• Do not apply till end of structure calculation
• i.e. Don't force structure
• Use appropriate distances
• Parallel vs. Anti-Parallel beta sheets

15
3J HNHA Coupling Contsants

• Used to derive Phi angles

16
3J HNHA Coupling Constants Early Attempt
17
3J HNHA Coupling Constants

• Derived from HNHA experiment
• Ix/IN -tan(3JPidel1del2)2
• based on intensity ratio

Overlapped Peaks Erroneous Intensities Error
in Result
18
3J HNHA Coupling Constants

• HRV14-3C results
• 98 Phi Angles
• Cor. Coef 0.98
• Ave. Dev. 9
• Max. Dev. 24

3J HNHA gt 8 Beta Sheet 3J HNHA lt 4 Alpha Helix
19
3J HNHA Coupling Constants Lesson

• Use only unambiguous data
• Assign and implement at end of calculation
• i.e. Don't force structure
• Don't over restraint
• Calculate error use appropriate tolerances
• NMRView - getval command

20
NOEs

• Cornerstone of structure calculation
• Use highest field strengths possible!
• Resolution, resolution, resolution.......
• Methyl to Methyl contacts
• Extremely important
• Can drive folding
• Calibration issues with NMRView

21
Resolution
500 Mhz vs.
800 Mhz
22
Resolution
500 Mhz vs. 800
Mhz
Resolved into two peaks
23
NOEs

• Calibration issues Vol. vs. Int.
• Backward linear prediction artificially inflates
  volume of fourier transformed data
• Necessary in 13C NOESY-HSQC to correct baseline
  distortion

24
NOEs

• Calibration method
• May need to adjust bounds or cut-offs
• Keep within reason (5.5 Angstroms)
• Scaling
• Only available for first peak list loaded
• Redundancy
• Remove stronger peak
• Pascals Scripts

25
Spin Diffusion

• Bi-product of experiment
• Keep mixing times as short as possible
• Do not assign weak methyls!
• i.e. Divide intensity by 3
• Keep upper bounds within reason!(5.5 Angstroms)

Assignment of weak methyl can drive structure to
wrong fold
26
Ambiguous Restraints

• Geminal ambiguity
• sum. vs. center averaging in Xplor/CNS
• Pseudo atom restraints
• Overlapped ambiguity
• Leave till the end!
• Assign as structure comes together
• Confirm all assignments with cross correlation

27
Cross Correlation
13C NOESY HSQC
15N NOESY HSQC
28
Diagonal NOEs

• Do not use!
• Will mess up calibrations
• Remove completely
• Peaks close to diagonal
• Proximity to diagonal may artificially inflate
  intensity manually assign upper bound limit

Assign only peaks that can be cross
correlated! i.e. make sure it is not a truncation
artefact
29
Noes Auto assignment

• Referencing issues and Tolerance settings
• Auto Assignment Programs
• Cyana and Aria
• Protein 180 residues
• Proton Tolerance 0.05 gt 150,000 Possible
  assignments
• Time consuming and error prone

30
NOEs Tolerance Issues
WOW!
OUCH!
31
Programs for Calculation

• Xplor/CNS with Aria, Cyana
• Cartesian vs. Torsion angle space
• Pseudo atom correction
• Use Appropriate weight averaging
• Sum. vs. center averaging

32
Water Refinement

• RECOORD and DRESS databases
• Water refinement not implemented in CYANA
• Dramatic improvement of Packing and Bump score
  www.ebi.ac.uk/msd/recoord
• RECOORD scripts
• Default uses sum averaging!

33
RECOORD Water Refinement
Uses full Lennard-Jones and OPLS Energy Potentials
60 Phi/Psi Core 69 Phi/Psi Core
81 Phi/Psi Core
Jan 2005 May 2005
June 2005
34
Structure Analysis

• Validation NMRQ, PROCHECK, Whatif
• Back calculation Queen (use cluster)
• Consistent NOE viol. gt 0.5 Angstroms
• Ramachandran Plot (omega angles)
• Chi angles / X1-X2 normality / Packing
• Bumps / 100 Residues
• What-check Z scores (low is good)
• RMSD

35
NOE Violations PROCHECK
36
Ramachandran Plots PROCHECK

• Individual Plots
• Should cluster

37
Chi Angles PROCHECK

• Core Packing BCAAs
• Want Low RMSD
• X1/X2 normality Plots
• NOE derived assignments
• Perdeuteration

38
Chi-1 Angles
Most Favored gt Second gt
Least Favored
Seen for serines and threonines in turns and helix
39
VADAR, NMRQ and Shiftcor

• Omega Angles
• DRESS Early water refinement - C. Spronks
  scripts
• Angles varied from 180 to 155 degree!
• RECOORD now uses omega restraints 180 /- 2?
• Packing and residual volume plots
• Shiftcor - use to back calculate assignments
• NOE assignments Must be confident with chemical
  shift assignments first!
• RMSD
• Should be low over secondary structure

40
Lessons

• Collect all the data!
• Must be confident with proton assignments
• Use water refinement for structure generation
• Implement 3J-HNHA, Phi angles, and H-bonds at end
  of calculation
• Watch log files carefully

41
Error Logs
NOES NOEgtassign (segid HRV and resid 160 and
name HE1) (segid HRV and resid 160 and name HE)
5.500 3.7 0.0 ! aronoesy.5 restraint
successfully read 1628 reading restraint
1629 SELRPN 1 atoms have been selected out
of 2952 SELRPN 2 atoms have been
selected out of 2952 NOESET-ERR i and j set
have to be disjoint NOESET-ERR error in
selection - no atoms spec. NOEgtassign (segid HRV
and resid 160 and name HN) (segid HRV and resid
160 and name HA) 5.500 3.7 0.0 ! hinnoesy.153
NOE-ERR problem at 1629 -999.000 -999.000
-999.000 -999.000 reading restraint 1630
Typos
atom nomenclature DIHEDRALgt !! T111
DIHEDRALgt assign (segid HRV and resid 111 and
name n ) (segid HRV and resid 111 and name ca
) SELRPN 1 atoms have been selected out
of 2952 SELRPN 1 atoms have been
selected out of 2952 SELRPNgt (segid HRV
and resid 111 and name cb) (segid HRV and
resid 111 and name cg ) 1.0 -60.0 30.0 2
SELRPN 1 atoms have been selected out of
2952 SELRPN 0 atoms have been selected out
of 2952 CSTRAN-ERR selection has to contain
exactly one atom.
Nomenclature change
42
Lessons

• Get to know the programs you use
• Use all the tools for Analysis
• Back calculate NOEs and chemical shifts
• Models Use Only for Global Fold
• eg. HRV14-3C SWISS-MODEL
• Backbone and H-bond discrepancies
• 20 difference in X1 X2 angles (packing)
• Trust your NOEs!

43
Ligands

• Enter the Chop Shop!
• Hic-Up Server
• http//alpha2.bmc.uu.se/hicup/
• PRODRG Server
• http//davapc1.bioch.dundee.ac.uk/programs/prodrg/
  
• Strips hydrogens
• Both produce Improper and Dihedral restraints
• CNS nomenclature incompatibilities
• No backward compatibility with .par and .top
  files
• Check log files closely

44
Useful links

• http//www.bionmr.com/forum/index.php?acthome