1
NOESY spectrum structure
NOEs and ambiguity
• 15N- or 13C edited 1H-1H NOESY
3D
1
H
13
C
1H1
H
15
N
1H
1
H
15
N
1H
13C
1
H
13
C
1H
13C
1
H
15
N
1H
15N
4D
1H 1H2D• 1H-1H NOESY
• 15N- or 13C edited 1H-1H NOESY
Working with 3D-NOESY
assign only intra-residue cross-peaks to
generate accurate chemical shift lists
NB: do not assign long-range NOEs !!!
Let CYANA do the job
User considerations
•Completeness of chemical shift assignments
should be higher than 90%
•Lack of aromatic chemical shifts is harmful to
the outcome of a structure calculation because
they give rise to a higher-than-average number
of NOEs
4. 1H-1H Distances from NOEs
A B C D Z• • • •Intraresidue
Sequential
Medium-range
(helices)
Long-range
(tertiary structure)
Challenge is to assign all peaks in NOESY spectra
- semi-automated processes for NOE assignment using
NOESY data and table of chemical shifts yet still
significant amount of human analysis
Traditionally NOE Assignment is done
manually
Distance restraints from not uniquely assigned NOEs:
Ambiguous distance restraints
Robustness against erroneous assignments:
Constraint combination / violation confinement
Reduction of assignment ambiguity prior to the structure calculation:
Probabilistic network-anchored assignment
CANDID/CYANA
Automated NOE Assignment and de novo
Structure Calculation
• User is biased against the data (erroneous assignments - rejected peaks)
• Time consuming (several months)
NOE assignment and ambiguous distance
restraints
In general, several different 1
H
chemical shifts ωA, ωB match the
position of a NOESY peak within
the experimental uncertainty Δω.
Assignment ambiguity
Manual assignment is very
cumbersome!
For peak lists obtained from 13Cor
15N-resolved 3D NOESY
spectra, the ambiguity in one of
the proton dimensions can usually
be resolved by reference to the
heteroatom
2
Ambiguous distance constrains
Constraint with multiple assignments
Allows delay of assignment choice until structures are better
defined
If one assignment possibility leads to a sufficiently short
distance, then the ambiguous distance restraint will be
fulfilled.
The presence of wrong assignment possibilities has no (or
little) influence on the structure, as long as the correct
assignment possibility is present.
Nilges et al., J. Mol. Biol. 269, 408-422 (1997)
A
B
C
9.52 ppm
4.34 ppm
4.34 ppm
Due to resonance overlap between
atoms B and C, an NOE crosspeak
between 9.52 ppm and 4.34 ppm
could be A to C or A to B - this
restraint is ambiguous.
But if an ensemble generated with this
ambiguous restraint shows that A is never
close to B, then the restraint must be A to C.
Resolving ambiguity
during structure calculation
9-11 Å
3-4 Årange of inter-atomic
distances observed in
trial ensemble
80
Constrain combination
Problem: Peaks with wrong
(long-range) assignments
may severely distort the
structure, especially in the
first cycles, and may lead to
convergence to a wrong
structure.
Idea: From two long-range peaks each, combine the assignments into a
single distance constraint for the first two cycles.
Result: occurrence of erroneous
assignments is reduced at the expense of
temporary loss of information
Effect of constrain combination
Example: 1000 long-range peaks, 10% of which
would lead to erroneous constraints
Individual constraints:
1000 constraints, ≈1000 x 0.1 = 100 wrong (10%)
2->1 constraint combination:
500 constraints, ≈500 x 0.12 = 5 wrong (1%)
4->4 constraints combination:
1000 constraints, ≈1000 x 0.12 = 10 wrong (1%)
The number of long-range constraints is halved by the 2->1
combination but stays constant on 4->4 pair-wise combination!!!
Network-anchoring
The generalized relative contribution is determined from chemical shift tolerance, crosspeak
symmetry, covalent structure compatibility, and the convergence of network anchoring
and three-dimensional structure compatibility of multidimensional experiments
Herrmann et al., J. Mol. Biol. 319, 209-227 (2002)
Conditions for valid assignment
of a NOESY cross-peak
chemical shift
agreement
network anchoring spatial proximity in
the structure
3
CYANA overview
• input data
protein sequence, chemical shift lists, NOESY peaks,
other constarints (RDC s, angles, hydrogen or disulphide
bonds)
• initial assignments
one or several assignments are defined based on
chemical shift lists
• rank of initial assignments
filtering criteria include presence of symmetry-related
cross-peaks, agreement of chemical shifts and peak
position, self consistency with the entire NOE network
• calibrate distance constraints
from the NOESY peak volumes or intensities upper
distance bounds are derived for the corresponding,
ambiguous or unambiguous distance restraints
• eliminate spurious NOESY cross-peaks
• constraints combination
unrelated long-range distance constraints are combined
into new virtual distance constraints
• structure calculation
NMR structure calculation
molecular dynamics
Etotal = EvdW + Ebs + Eab + Etorsion + Eelctrostatics + …
We use a force field, or equations that
describe the energy of the system as a function of <xyz>
coordinates.
In general, it is a sum of different energy terms:
NOE data
ENOE = KNOE * ( rcalc - rmax )2 if rcalc > rmax
ENOE = 0 if rmax > rcalc > rmin
ENOE = KNOE * ( rmin - rcalc )2 if rcalc < rmin
Strong NOE 1.8 - 2.7 Å
Medium NOE 1.8 - 3.3 Å
Weak NOE 1.8 - 5.0 Å
The potential energy function related to these ranges looks like this:
It is a flat-bottomed quadratic function. The further away the distance
calculated by the computer (rcalc) is from the range, the higher the penalty.
As long NOEs relate our experimental data with the <xyz> coordinates,
we include them at the end of the energy function.
Similarly, we include torsions as a range constraint:
Torsion angles
EJ = KJ * ( fcalc - fmax )2 if fcalc > fmax
EJ = 0 if fmax > fcalc > fmin
EJ = KJ * ( fmin - fcalc )2 if fcalc < fmin
Or any other type of contraints (RDC, PRE, PCS, chemical shifts, etc)
Penalty function
Rcalc or fcalc
E
0
rmin
fmin
rmax
fmax
4
Energy minimization
f
y
E
(Kcal/mol)
functions in the potential energy expression for the molecule,
represent bonded interactions (bonds, angles, and torsions), and
non-bonded interactions (vdW, electrostatic, NMR constraints).
to get the structural model we must be able to minimize the energy
of the system, which means to find a low energy (or the lowest
energy) conformer or group of conformers.
Nearly impossible, because we are looking at a n-variable surface
We have energy peaks (maxima) and valleys (minima).
Simulated annealing
Provide energy to the system (rise the ‘temperature’) and see how it evolves with
time. Temperature usually translates into kinetic energy, which allows the peptide to
surmount energy barriers.
Restrained Molecular Dynamics structure
determination of protein 1GB1 from NMR
Judge your structure
CANDID criteria
• Average CYANA target function value of cycle 1 below 250 Å2
• Average final CYANA target function value below 10 Å2
• Less than 20% unassigned NOEs
good data sets can reach 95% of input peaks assigned
always check the unassigned peaks !!!
• Less than 20% discarded long-range NOEs
not straightforward to assess due to chemical shift ambiguity
• RMSD value in cycle 1 below 3 Å
• RMSD between the mean structures of the first and last cycle below 3 Å
Water refinement
Improving the Quality of NMR Structures
• Water Refinement
Ø protein structures generally calculated in vacuum.
Ø water has a significant effect on protein structures
t explicit solvent model
–MD simulation in box of water
– box > 10 Å, keep solvent from edge
– 1000 to 10,000s water molecule
– Computationally expensive
Water refinement
compare structures in vacuum to water
– no visible difference
5
Water refinement
subtle, but significant improvements
t compare structures in vacuum to water
– improves NH to CO hydrogen bonds
– improves f and y angle distributions
to keep or not to keep manual assignments
[do not keep]
Phe102 HZ has been assigned based on unique NOE cross-peaks
CYANA consistently rejected these NOEs
Xray structure confirmed our suspicions
We fixed only 3 NOEs from Phe102 HZ
Introducing Val/Leu stereospecific assignments resolved the problem
TALOS predictions
and their effect on NMR structures
TALOS TALOS+
TALOS vs Xray (1)
TALOS TALOS+
TALOS in structure calculations
X-ray
NMR wrong PSI 34 (RMSD 0.65 Å)
NMR without PSI 34 (RMSD 0.61 Å)
NMR without wrong angles (RMSD 0.57 Å)
what is a good NMR structure
IPSE (106aa)
sequential: 499
intra-residual: 651
medium-range: 286
long-range: 1214
total: 2650
ramachandran
core: 91.1%
allowed: 8.9%
generous: 0.0%
disallowed: 0.0%
Wattos Surplus Analysis Summary
Found number of to do constraints: 2650
Found number of exceptional constraints: 0
Found number of constraints to be double with others: 17
Found number of impossible constraints: 0
Found number of fixed constraints: 2
Found number of redundant constraints: 1
Found number of non-redundant constraints: 2630
Found number of constraints to be surplus (E+C+D+I+F+R): 20
Overall NOE completeness is 68.10 percent
Input spectra
hNH
hCH
hCH2
cNH_
CcH_
hH_noN_
hH
selected peaks: 9345
assigned: 8865 (94.8%)
unassigned: 480
6
f, y, c1, c2 distribution
Comparison of main chain and side-chain parameters to
standard values
PROCHECK analysis Wattos analysis
examples
Qua1 symmetric dimer
• two 13C edited noesy spectra as input
• no filter NOESY experiments
Full NOE set + RDCs
crystals coming to your rescue
bb rmsd 1.4 Å
dimer vs dimer
examples
MYND the sinful structure
45 aa !!!
• Structure calculation without the zinc atoms
• Identification of the zinc coordination residues from the fold
• Repeat calculation with the zinc atoms fixed
CYANA uses only distances
S - ZN: 2.3 S - S: 3.65
N - ZN: 2.0 S - N: 3.35
CNS uses both distances and angles definitions with possibility of using
different weights
S – ZN: 2.3 S – ZN – S : 109.5
N – ZN: 2.0 N – ZN – S : 120
In both cases one
needs to give the
Zn chelating
residues
Defining tetrahedra
7
TYR552
examples
aromats troubleshooting
4 Tryptophans
3 Tyrosines
4 Histidines
5 Phenylalanines
Y552-CQD-QD Y552-CQE-QE
Number of NOE derived distance restraints
total 2687
short-range, |i-j|<=1 1378
medium-range, 1<|i-j|<5 445
long-range, |i-j|>=5 864
RMSD (residues 519-622)
Average backbone RMSD to mean 1.04 +/- 0.28A
Average heavy atom RMSD to mean 1.62 +/- 0.35A
Ramachandran plot
Residues in most favoured regions 93.2%
Residues in additionally allowed regions 6.5%
Filtered/Edited NOE: based on selection of NOEs from two
molecules with unique labeling patterns
1
H 1H
13C
Unlabeled
peptide
Labeled
protein
Intermolecular NOEs
Protein-ligand structures
Summary
• CYANA will determine the correct fold
• you should take care for the input data
• you should take care for the local geometry
• understand how CNS works to refine your structure
In general to determine an NMR structure is (not) straightforward