This document provides an overview of protein structure analysis tools and techniques: 1) It describes exploring the Protein Data Bank (PDB) to view and analyze X-ray crystallography and NMR protein structures, comparing similar structures, and using tools like FoldX for in silico mutagenesis and homology modeling. 2) Key concepts covered include PDB file formats, atomic coordinates, B-factors, resolution, RMSD, and the principles of X-ray crystallography, NMR structure determination, and homology modeling. 3) Visualization software like YASARA, SwissPDBViewer and PyMOL are introduced for viewing protein structures from the PDB.

1. BITS Training Protein Structure Joost Van Durme VIB Switch Laboratory Vrije Universiteit Brussel http://www.bits.vib.be/training

2. Topics for today Exploring the protein structure databank (PDB) Viewing and analyzing protein structures with YASARA Comparing similar protein structures In silico mutagenesis with FoldX Homology modeling with FoldX

3. PDB contains 65000 structures EMBL-Bank contains 114,475,051 sequences or 215,540,553,360 nucleotides! Sequences and structures

4. The sequence-structure gap

5. X-ray crystallography (crystals) Nuclear Magnetic Resonance (NMR) (in solution) Electron microscopy (in native tissue) Structures can be solved

6. Solving structures is lots of work (6 months to years) Need lots of material/reagents. Solubility is a problem. Some protein structures are really difficult to solve: membrane proteins, extremely large proteins, protein complexes The field evolves fast. Techniques improve, more user friendly software, more automatisation (x-ray infrastructure, crystal growth) Despite this progress it is not expected that the sequence-structure gap will ever be closed. But ...

7. What can we learn from models/structures? Active site structure: structure-based drug design Protein-protein interactions Function Antigenic behavior / vaccine development Stabilising proteins using structural knowledge

8. Human vs parasite Parasite Active site

9. 1918 Influenza Epidemic Influenza Virus

10. NEURAMIDASE POCKET SIALIC ACID

11. RELENZA SIALIC ACID

12. RELENZA TAMIFLU 5.000.000+ doses in NL

13. Trouw – 3 maart 2009

14. RELENZA TAMIFLU WT K i = 1.0 H274Y K i = 1.9 WT K i = 1.0 H274Y K i =265 H274Y H274Y

15. PDB structures come from ... X-Ray crystallography experiments NMR structure determination The PDB no longer contains: EM structures (too low resolution) Models (too unreliable)

16. Principle of X-Ray crystallography initial model electron densities

17. X-Ray structure

18. X-Ray models components x, y, z coordinates: define the mean atom position disorder about this mean: B-factor and occupancy variations in time and space B-factor: model the ‘smearing out’ of disorder around the mean atom position (ellipsoids) higher B-factor means more uncertainty about position Occupancy: consider alternative conformations of the same sidechain how often do we find this sidechain in one conformation and how often in the other conformation

19. Occupancy ATOM 625 C ILE A 77 -11.322 28.374 -1.179 1.00 28.77 C ATOM 626 O ILE A 77 -11.946 29.453 -1.112 1.00 28.84 O ATOM 627 CA AILE A 77 -11.432 27.329 -0.087 0.70 28.15 C ATOM 628 CB AILE A 77 -12.918 26.874 0.087 0.70 28.64 C ATOM 629 CG1AILE A 77 -13.042 25.758 1.141 0.70 26.75 C ATOM 630 CG2AILE A 77 -13.516 26.421 -1.241 0.70 28.13 C ATOM 631 CD1AILE A 77 -13.378 26.302 2.501 0.70 26.47 C ATOM 632 CA BILE A 77 -11.423 27.327 -0.082 0.30 28.50 C ATOM 633 CB BILE A 77 -12.874 26.775 0.117 0.30 28.79 C ATOM 634 CG1BILE A 77 -13.519 26.423 -1.227 0.30 28.62 C ATOM 635 CG2BILE A 77 -13.748 27.739 0.916 0.30 28.40 C ATOM 636 CD1BILE A 77 -14.720 25.518 -1.100 0.30 28.69 C ATOM 637 N ARG A 78 -10.521 28.048 -2.183 1.00 28.70 N ATOM 638 CA ARG A 78 -10.258 28.952 -3.268 1.00 28.47 C ATOM 639 C ARG A 78 -10.857 28.469 -4.584 1.00 28.22 C 2VWC

20. Atomic B-factors Value which determines the precision of an atom’s given position Atoms with the largest B-factors will have the largest positional uncertainty Indication of mobility of an atom 0 < B < 20 : Atom is most likely OK 20 < B < 40 : Atom is probably OK, but positional errors up to 0.5 Ångstrom are normal 40 < B < 60 : Atom is probably reasonably OK, but be careful, because positional errors up to 1.0 Ångstrom can be observed B > 60 : Atom is not likely to be within 1.0 Ångstrom from where you see it B around 100 : Atom is guaranteed not within 1.0 Ångstrom from where you see it

21. B-factor www.YASARA.org Low High

22. Resolution (Angstrom) Level of detail that can be observed in the electron density map The greater the disorder in the crystal, the lower the resolution (proportional to the protein size) 3.0A 2.0A 1.2A

23. R-factor The difference between the observed and computed diffraction pattern A measure of how well the refined structure predicts the observed data Higher values mean less agreement 0.40-0.60: very unreliable 0.20 seems to be the standard threshold electron density map

24. NMR Structure determination

25. NMR models components In solution study protein dynamics solve protein structures that are difficult to crystallize Nuclear Overhauser Effect or NOE intensities of signal peaks correspond to short inter-atomic distances between spatially close protons (NOE distances) NOE constraints are known with low precision. E.g. NOEs are binned 2.5-4.0, 4.0-5.5, and 5.5-7.0 Angstrom Multiple models are generated that are consistent with the distance and angle constraints using e.g. molecular dynamics: the NMR ensemble Take average or best model from ensemble for PDB deposition, or just deposit a selected ensemble of superposed structures

26. Structure superposition (root mean square distance) n = number of atoms di = distance between 2 corresponding atoms i in 2 structures The more atoms superpose on each other, the lower the RMSD Unit of RMSD => Ångstroms identical structures => RMSD = “0” similar structures => RMSD is small (1 – 3 Å) distant structures => RMSD > 3 Å However, care has to be taken as RMSD is length dependent and dominated by outliers: comparison of two short peptide structures can result in a small RMSD even if their structure is visibly different. very similar structures can have a bad RMSD due to a short part of the structures that is very different (loops) Insertions and deletions are not implemented in the RMSD calculation, since we only look at equivalent atoms/residues (see figure)

27. NMR ensemble RMSD (root mean square distance) Superpose the NMR models Calculate RMSD of local regions and also whole models Regions with high RMSD are less well defined by the data

28. Structural data is stored in the Protein Data Bank (PDB) http://www.pdb.org Protein Data Bank (PDB)

29. ©CMBI 2009 ©CMBI 2009 Protein Data Bank (PDB) Databank for 3-dimensional structures of biomolecules: Protein DNA RNA Ligands Obligatory deposit of coordinates in the PDB before publication ~ 65 000 entries (April 2010) ( ~27000 “unique” structures) PDB file is a keyword-organised flat-file ( 80 column) 1) human readable 2) every line starts with a keyword (3-6 letters) 3) platform independent

30. ©CMBI 2009 PDB important records (1) PDB nomenclature Filename= accession number= PDB Code Filename is 4 positions (often 1 digit & 3 letters, e.g. 1CRN.pdb) HEADER describes molecule & gives deposition date HEADER PLANT SEED PROTEIN 30-APR-81 1CRN CMPND name of molecule COMPND CRAMBIN SOURCE organism SOURCE ABYSSINIAN CABBAGE (CRAMBE ABYSSINICA) SEED

31. ©CMBI 2009 PDB important records (2) SEQRES Sequence of protein; be aware: Not always all 3d-coordinates are present for all the amino acids in SEQRES!! SEQRES 1 46 THR THR CYS CYS PRO SER ILE VAL ALA ARG SER ASN PHE 1CRN 51 SEQRES 2 46 ASN VAL CYS ARG LEU PRO GLY THR PRO GLU ALA ILE CYS 1CRN 52 SEQRES 3 46 ALA THR TYR THR GLY CYS ILE ILE ILE PRO GLY ALA THR 1CRN 53 SEQRES 4 46 CYS PRO GLY ASP TYR ALA ASN 1CRN 54 SSBOND disulfide bridges SSBOND 1 CYS 3 CYS 40 SSBOND 2 CYS 4 CYS 32

32. ©CMBI 2009 PDB important records (3) and at the end of the PDB file the “real” data: ATOM one line for each atom with its unique name and its x,y,z coordinates ATOM 1 N THR 1 17.047 14.099 3.625 1.00 13.79 1CRN 70 ATOM 2 CA THR 1 16.967 12.784 4.338 1.00 10.80 1CRN 71 ATOM 3 C THR 1 15.685 12.755 5.133 1.00 9.19 1CRN 72 ATOM 4 O THR 1 15.268 13.825 5.594 1.00 9.85 1CRN 73 ATOM 5 CB THR 1 18.170 12.703 5.337 1.00 13.02 1CRN 74 ATOM 6 OG1 THR 1 19.334 12.829 4.463 1.00 15.06 1CRN 75 ATOM 7 CG2 THR 1 18.150 11.546 6.304 1.00 14.23 1CRN 76 ATOM 8 N THR 2 15.115 11.555 5.265 1.00 7.81 1CRN 77 ATOM 9 CA THR 2 13.856 11.469 6.066 1.00 8.31 1CRN 78 ATOM 10 C THR 2 14.164 10.785 7.379 1.00 5.80 1CRN 79 ATOM 11 O THR 2 14.993 9.862 7.443 1.00 6.94 1CRN 80

33. PDB entry

34. PDB entry

35. PDB entry

36. ©CMBI 2009 Structure Visualization Structures from PDB can be visualized with: YASARA (http://www.yasara.org) SwissPDBViewer (http://spdbv.vital-it.ch/) PyMOL (http://www.pymol/org) Chimera (http:// www.cgl.ucsf.edu/chimera )

37. YASARA View nomenclature Atom Residue = any continuous stretch of atoms sharing the same residue name, residue number and molecule name Molecule = any continuous stretch of residues sharing the same molecule name (PDB calls this a CHAIN) Object = a collection of molecules and additional items

38. Standard atom colors C = cyan O = red N = blue H = white S = green

39. Atom nomenclature N-term C-term C α C β C γ O γ N N O C α C β C C C γ C δ 1 C δ 2 OT1 OT2

40. FoldX: a molecular design toolkit Predict the effect of point mutation on the protein stability Predict the 3D structure of a sequence: homology modeling

41. Predict effect of point mutation FoldX is an empirical force field It is validated with calorimetric experiments E.g. If such an experiment concludes that breaking a hydrogen bond costs 1.5 kcal/mol, FoldX uses this knowledge rather than using theoretical physics equations FoldX compares WT and mutant for: Hydrogen bonds, electrostatics, Van der Waals clashes and contacts, entropy, desolvation, ... FoldX energies Energy of a single molecule is meaningless The difference in energy of two molecules (such as WT and a point mutant) approaches realistic values

42. Predict effect of point mutation FoldX calculates the stability of WT and MT and makes the difference (net effect of mutation): Δ G MT - Δ G WT = ΔΔ G m u t at i o n If ΔΔ G m u t at i o n > 0 : mutation is bad for stability < 0 : mutation is good for stability FoldX error margin is 0.5 kcal/mol, so changes within this margin are meaningless

43. Introduction to homology modeling Goal: predict a structure from its sequence with an accuracy that is comparable to the best results achieved experimentally (X-Ray) Protein modeling is the only way to obtain structural information when experimental techniques (x-ray, NMR, EM) fail

44. Homology Modeling

45. Principles of Homology Modeling Search for a sequence with a known structure that is very similar to the sequence with the unknown structure. Build model using known structure as template The structure of a protein is uniquely determined by its amino acid sequence Structure is more conserved than sequence Similar sequences adopt nearly exact same structure Distantly related sequences can still fold into a similar structure

46. Sequence similarity rule Rost (1999) modeled lots of structures and compared them to the real ones in the PDB Derived precise limits for homology modeling This rule tells you whether a model will be reliable or unreliable

47. FoldX plugin for YASARA

48. Acknowledgements Gert Vriend, Radboud Universiteit Nijmegen, NL ( www.cmbi.ru.nl ) Sander Nabuurs, Lead Pharma, Nijmegen, NL Greet De Baets, VIB Switch Laboratory