2. Rational Drug Discovery PC session Protein sequence analysis Biocomputing Primary etc structure X-ray crystallography Structural genomics Homology modelling Protein Structure QSAR History Objectives Limitations Statistics Steric Electrostatics Hydrophobic PC sessions Molecular modelling Theory Drug structure Drug conformation Docking De Novo ligand design PC sessions FBDD Lead compound Physiological Biochemical Chemical (prodrugs) Targeting and delivery

3. 3D considerations Protein structure X-ray crystallography NMR Homology modelling Forming quality crystals The resolution Solving the structure New for 2010: FBDD replaces 3D QSAR (Comfa)

4. atoms X-rays 6. Protein structure 6.1. Introduction QSAR most useful when no structure for the target when structure of target protein is known, molecular modelling more useful (PC activities are divided between structure known and structure not known) FBDD is heavily dependent on structure for designing ways to grow the fragment. Resolution for light to view an object is about /2. Wavelength, (visible) is about 4000 - 7000 Å Compare to C-C bond length of 1.5 Å  (X-rays) 20

5. 6.3. X-ray crystallography Requires a single crystal so scattered radiation from each atom is in phase A major difficulty in X-ray crystallography is obtaining crystals which diffract For myoglobin (1st protein structure, 1957), add solution of ammonium sulphate to concentrated solution of proteins  crystals Some proteins crystallise easily, some don’t trial and error, vary solvent, source of protein, temperature 10 years for thymidylate synthase (TS) Membrane proteins (e.g. GPCRs, ion channels, transporters) are the target for the majority of drugs but very few crystallised Lysozyme crystals http://crystal.uah.edu/ ~carter/protein/protein.htm A diffraction pattern

6. 6.4. Diffraction apparatus Illustrations: (above left) liquid nitrogen is used to freeze the crystal which allows for increased reliability of information gathered from testing. The area detector, which collects the diffracted x-rays once they pass through the crystal, is the black plate located behind the nitrogen stream Not examinable

7. 6.5. Recombinant DNA plasmids can be Used to produce a specific protein (right); robots produce crystals (below) Not examinable

8. 6.7. Detection Film: collects several reflections (spots) at once - inefficient Scinillation counters collect one reflection at a time Note X-ray damage to crystal limits time available to collect data as image gets foggy Area detectors collect several reflections at once A good structure may require over 100 000 reflections A low resolution structure may be misleading - see below More powerful X-ray beam (e.g. from a synchrotron as at Daresbury, Cheshire) can be used to study very large proteins e.g. virus 8500 kdaltons, 240 protein subunits

9. 3 Å 2 Å 1.2 Å 125 K 293 K 6.8. Resolution and temperature The quality and reliability of the structure depends on the resolution. Generally, below 2 Å is good; 1.2 Å is excellent. From: http://www-structure.llnl.gov/xray/101index.html at 4 Å can see the chain at 2 Å can see the sidechains at 1 Å all atoms and solvent visible

10. 6.9. Resolution depends on crystal quality Good quality crystals but they only diffract to low resolution The further back the X-rays scatter (i.e. the bigger  ), the better the resolution. 40

11. 6.9. Not a serious slide - but it does show the difficulty in producing good crystals Perfect crystals disrupt their formation. The generation of perfect crystals can sometimes be the limiting factor in determining a protein's structure. are difficult to achieve on Earth. Gravity and turbulence in the atmosphere Scientists are now looking at conducting protein crystallization experiments on the International Space Station. By eliminating variables such as gravity, crystals are able to form slower and more precise in space. NASA, in association with the University of Alabama in Huntsville and the University of California in Irvine, have been working on the development of the equipment and the procedure needed to produce protein crystals in space. Not examinable

12. Crystallization tricks β 2 -adrenergic receptor co-crystallized with lysozyme (red) – as lysozyme forms crystals easily but membrane proteins do not β 1 -adrenergic receptor – thermostabilized (binds 50% of ligand at a higher temp) by 6 mutations so that it crystallizes more easily.

13. The active GPCR structure The  2 -adrenergic receptor (yellow) in complex with its heterotrimeric G- protein (Gs) (alpha blue, beta cyan and gamma green). Lysozyme (red) is attached to the N-terminus and a nanobody (a small camel antobody) is shown in red binding to the G-protein). Lysozyme and the antibody aid crystallization.

14. 6.10. Solving the structure Calculate the electron density (using computers) from the diffraction pattern (we don’t need to know how) Fit the amino acids one by one into the electron density contours (e.g. using computer graphics) to do this need to know the amino acid sequence (automatic degradation of protein or from cDNA) In the illustration, the 5-atom imidazole ring from the amino acid histidine is adjusted from an incorrect location (panel A) to a correct location (panel B).

15. 6.10. Solving the structure (continued) Structures can be refined using molecular mechanics and molecular dynamics. Scattering of X-rays is proportional to electron density so hydrogen atoms are not visible. Cannot distinguish -CO 2 - from -COOH, -CO 2 - from CONH 2 water molecules from Na + ions Why is this a problem? 6.10.1. Neutron diffraction No free radicals generated - no crystal damage, neutrons are scattered by nuclei, not electrons so H-atoms can be observed. Disadvantage? Need nuclear reactor! It is a different exercise to design a drug to bind to –CO 2 - rather than –COOH or – CONH 2

16. 6.11. Modelling protein structure Some out of date statistics: protein sequences available protein structures available Probably there is no structure for the protein of interest Need to predict the structure from sequence 6.11.1. Modelling protein structure by homology Proteins exist in families with similar 3D structures (e.g. TIM barrel): i.e. the same pattern of  -helices,  -sheets, loops many identical residues at corresponding positions Main differences in loops which connect important parts 2 proteins show homology if there is similarity in their sequences 45 5 000 000 30 000

17. 6.11.2. Modelling proteins by homology Align sequences, preferably in a multiple sequence alignment Display co-ordinates of known structure Alter sequence of known to that of unknown make insertions or deletions in the loops building the loops may be guided by loops of similar structure in the protein structure databank Adjust orientation of side chains to remove steric clashes Use molecular mechanics/molecular dynamics to “refine” structure Does the structure we have modelled explain the known experimental facts? If not, adjust accordingly This method works well if the similarity is high. See PC activity. If there is no related structure, use structure prediction methods?

18. 6.12. Protein structure by NMR 6.12.1. Nuclear Overhauser effect (NOE) Change in intensity of one peak when another is irradiated is called an NOE There is a single absorption frequency for transition E -  E + since most of the molecules lie in lower energy state. Apply intense radiation corresponding to the transition frequency for one type of proton - saturate the NMR peak - change spin population so equal number in two levels The nuclei interact with neighbouring nuclei and change their spin populations so intensity changes Effect is proportional to 1/r 6. Advantages: Gives distance information on protons close together “Distance Geometry” can be used to turn distance information into 3D structure Works for proteins that don’t crystalize Disadvantages Need powerful NMR > 500 MHz Difficulty in assigning peaks Currently works only for small proteins