Refinement with REFMAC

1
Refinement with REFMAC

• Garib N Murshudov
• York Structural Laboratory
• Chemistry Department
• University of York

2
(No Transcript)
3
(No Transcript)
4
Contents

1. Refinement program Refmac
2. Simple refinement Selection of weights
3. Automatic twin refinement Rfactor warnings
4. Low resolution refinement tools

5
What can REFMAC do?

• Simple maximum likelihood restrained refinement
• Twin refinement
• Phased refinement (with Hendrickson-Lattmann
  coefficients)
• SAD/SIRAS refinement
• Structure idealisation
• Library for more than 10000 ligands (from the
  next version)
• Covalent links between ligands and ligand-protein
• Rigid body refinement
• Low res NCS local, restraints to external
  structures, jelly body
• TLS refinement
• Map sharpening
• Occupancy refinement
• etc

6
Simple refinement
7
Simple refinement
8
Optimisation of weights
9
Optimisation of weights

• After refinement final statistics are
• Initial Final
• R factor 0.2783 0.1831
• R free 0.2668 0.2030
• Rms BondLength 0.0284 0.0327
• Rms BondAngle 4.5704 2.3083
• Rms ChirVolume 0.1696 0.1645
• RMSD of bond lengths is too large.

10
Optimisation of weights

• If rmsd of bond lengths is too large (gt0.022) or
  too tight (lt0.01) then you may want to change
  weights. It can be done using weight matrix on
  the interface.
• Look at the log file. Refmac prints out current
  weights it is using.
• Weight matrix 4.4438701
• Actual weight 10.000000 is applied to the
  X-ray term
• If rmsd is large then you can use half of
  currently used weight matrix (around 2.2).

11
Optimisation of weights
Change weight matrix
12
Optimisation of weights

• With new weights RMSD is reasonable.
• Initial Final
• R factor 0.2783 0.1876
• R free 0.2668 0.2052
• Rms BondLength 0.0284 0.0201
• Rms BondAngle 4.5704 1.6554
• Rms ChirVolume 0.1696 0.1063

13
Twin refinement
14
merohedral and pseudo-merohedral twinning
Crystal symmetry P3 P2
P2 Constrain - ß 90º - Lattice
symmetry P622 P222 P2 (rotations
only) Possible twinning merohedral
pseudo-merohedral -
Crystal lattice is invariant with respect to
twinning operator. The crystal is NOT invariant
with respect to twinning operator.
15
Twin refinement (it works with older version also
16
Twin refinement

• Twin refinement in REFMAC is carried out in
  several stages
• Stage 1 Identify potential twin operators. It is
  done by analysis of lattice and crystal symmetry.
• In this case space group is P31 and there are
  four potential twin operators
• Potential twin domain 1 with operator H, K,
  L, metric score 0.000
• Potential twin domain 2 with operator -K, -H,
  -L, metric score 0.000
• Potential twin domain 3 with operator -H, -K,
  L, metric score 0.000
• Potential twin domain 4 with operator K, H,
  -L, metric score 0.000

17
Twin refinement Group/subgroup
18
Twin refinement

• 2) Stage 2 Filter using agreement between twin
  related reflections (using Rmerge)
• Filtering out small twin domains, step 1
  
• Twin ops with Rm gt 0.44 will be removed
• SymOp -K,-H,-LR_m0.248twin is probable
• SymOp -H,-K, LR_m0.237twin is probable
• SymOp K, H,-LR_m0.027twin or higher symm
• At this stage REFMAC may suggest that space group
  could be higher

19
Twin refinement Effect of twin on Rmerge

• R merges without experimental error
• No twinning
  50
• Along non twinned axes with another axis than
  twin 37.5

Non twin
Twin
20
Twin refinement

• 3) Stage 3 Estimate twin fractions and remove
  small twin domains
• Filtering out small twin domains, step 2
  
• Twin domains with fraction lt 7.00000003E-02
  are removed
• Twin operators with estimated twin fractions
  
• Twin op H, K, L Fr 0.391 Eq ops K,
  -H-K, L -H-K, H, L
• Twin op -K, -H, -LFr 0.112 Eq ops -H, HK,
  -L HK, -K, -L
• Twin op -H, -K, LFr 0.108 Eq ops -K,
  HK, L HK, -H, L
• Twin op K, H, -LFr 0.390 Eq ops H,
  -H-K, -L -H-K, K, -L

21
Twin refinement

• 3) Stage 4 Perform twin refinement with all
  survived twin operators (in this example all four
  operators survive)
• Twin fractions 0.3773 0.1246
  0.1206 0.3775
• Rfactors look very good
• Initial Final
• R factor 0.1912 0.1566
• R free 0.1796 0.2047
• Rms BondLength 0.0088 0.0235
• Rms BondAngle 1.4825 2.1812
• Rms ChirVolume 0.1077 0.1336

22
Rfactors from non-twinned refinement
Initial Final R factor
0.3103 0.2779 R free 0.3184
0.3496 Rms BondLength 0.0088 0.0129
Rms BondAngle 1.4825 1.5648 Rms
ChirVolume 0.1077 0.1034
23
Twin refinement Rfactors be careful
Cyan perfect twin and twin modelled Black no
twin and not modelled Red perfect twin and
not modelled Blue no twin and perfect twin
modelled
Rfactor drop can be as large as 15 without
atomic model improvement
24
Twin refinement Alternative indexing

• If crystal can be twinned then there may be more
  than one indexing of hkl. Different indexing are
  related with the symmetry operator of lattice but
  not the crystal.
• Best way of dealing with indexing problem is to
  use the program pointless by Phil Evans. You can
  either give a reference mtz file or a reference
  structure. Then all subsequent data will be
  indexed in consistent manner.

25
Low resolution refinement
26
Low resolution refinement tools

• Jelly body (implicit normal modes) refinement
• NCS local and global restraints
• NCS constraints
• Restraints to reference structures
• Regularised map sharpening
• Long range B value restraints based on
  Kullback-Liebler distances
• Murshudov GN, Skubak P, Lebedev AA, Pannu NS,
  Steiner RA, Nicholls RA, Winn MD, Long F, Vagin
  AA REFMAC5 for the Refinement of Macromolecular
  Crystal Structures Acta Cryst , D67, 355-367

27
Restraints to external structures are generated
by the program ProSmart1) Aligns structure in
the presence of conformational changes. Sequence
is not used2) Gernates restraints for aligned
atoms3) Identifies secondary structures (at the
moment helix and strand, but the approach is
general and can be extended to any motif).4)
Generates restraints for secondary
structuresNote 1 ProSmart has been written by
Rob Nocholls and available from him (now). It
will be distributed by ccp4 (hopefully from the
next release)Note 2 Robust estimator functions
are used for restraints. I.e. if differences
between target and model is very large then their
contributions are downweighted
External (reference structure restraints)
28
Restraints to current distances
The term is added to the target
function Summation is over all pairs in the
same chain and within given distance (default
4.2A). dcurrent is recalculated at every cycle.
This function does not contribute to gradients.
It only contributes to the second derivative
matrix. It is equivalent to adding springs
between atom pairs. During refinement
inter-atomic distances are not changed very much.
If all pairs would be used and weights would be
very large then it would be equivalent to rigid
body refinement. It could be called implicit
normal modes, soft body or jelly body
refinement.
29
The program will be available from ccp4.
Currently if you want to try it you should ask
Rob Nicholls at ran105_at_york.ac.ukOnce you have
downloaded you can run using this
commandprosmart p1 refined_structure.pdb p2
reference_structure.pdbIt will generate many
useful info including restraints to the reference
structure.
External (reference structure restraints)
30
Auto NCS local and global

• Align all chains with all chains using
  Needleman-Wunsh method
• If alignment score is higher than predefined
  (e.g.80) value then consider them as similar
• Find local RMS and if average local RMS is less
  than predefined value then consider them aligned
• Find correspondence between atoms
• If global restraints (i.e. restraints based on
  RMS between atoms of aligned chains) then
  identify domains
• For local NCS make the list of corresponding
  interatomic distances (remove bond and angle
  related atom pairs)
• Design weights
• The list of interatomic distance pairs is
  calculated at every cycle

31
Add external keywords file in refmac interface
Browse files
32
Add external keywords file in refmac interface
Select keywords file
33
Add external keywords file in refmac interface
Keywords file
34
(No Transcript)
35
Jelly bodyRidge dist sigma 0.01ridge dist
dmax 4.2 ncs ncsr local to control
restraints to reference structures. Restraints
are generated by prosmart external dmax
4.2external weight scale 4external cut 10
Instructions you may want to play with
36
Low resolution refinement Some results
Initial Simple Jelly NCS local Jelly/NCS Reference structure
R 0.3605 0.2218 0.2533 0.2232 0.2535 0.2557
Rfree 0.3563 0.3116 2961 0.3124 0.2955 0.2907
If you want to use current version then you may
need to run several time to get parameters right.
In this case maximum radius for reference
structure restraint was 4.0, maximum radius for
NCS local was 4.2, if deviation between reference
distance and current distance was more than 10
sigma then it was excluded, sigmas for reference
structures were 0.07. At lower resolution (5-7Å)
radius may need to be 5.5 and sigma 0.02
37
Conclusions

• Auto weight works fine for large class of cases,
  however you may need to change weights
• Twin is automatic but Rfactors are poor
  indicators
• Use of available information may improve low
  resolution refinement

38
Acknowledgment

• York Leiden
• Alexei Vagin Pavol Skubak
• Andrey Lebedev Raj Pannu
• Rob Nocholls
• Fei Long
• CCP4, YSBL people
• __________________________________________________
  ____________________
• REFMAC is available from CCP4 or from Yorks ftp
  site
• www.ysbl.york.ac.uk/refmac/latest_refmac.html
• Balbes and other programs
• www.ysbl.york.ac.uk/refmac/YSBLPrograms/index.jsp
• This and other presentations can be found on
• www.ysbl.york.ac.uk/refmac/Presentations/