Charles Ballard (original GáborBunkóczi) CCP4 Workshop 7 December 2011

Transcription

1 Experimental phasing Charles Ballard (original GáborBunkóczi) CCP4 Workshop 7 December 2011

2 Anomalous diffraction F P protein F A anomalous substructure ano F A " -FA" A F F -* F A F P

3 Phasing Substructure location Anomalous substructure Substructure refinement Electron density map Phase calculation

4 Phase calculation in SAD F o, F o - measurements F o - F o -* calculated from substructure (-2F A ") α 1,α 2,α best phase angles α 2 F o - F o -* F o o - o Approximate map given by i mf best e α With =, best α best F F F best m = cos ( α ) α F o - α 1

5 SAD likelihood function Probability of experimental observations. Observations correlated, multivariate distribution. P(F O, F, H, H ) * P(FO, FO ;H P( F O, α, F,H, * * * O O α Factor joint probability into two parts P(F O, F O ;H, H ) ;H * ) = P(FO ;FO,H, H ) P(FO, H * *,H ) * ) Marginalise phases α and α -

6 SAD function normal scattering ( ) exp ;, 2 * Σ Σ = i o o o D e F F F P H H α π α scattering. Primarily normal and H between for errors accounts -* Σ o F H -* F o -

7 SAD function -anomalous ( ) ( ) accounts for errors between 2 ei exp 2,, ; 0 2 * Σ Σ Σ Σ = F F F F F F P c o c o o O o H H F F o F -F -* ) ( where, and F F accounts for errors between * Φ = Σ Φ H e F e D H F i i C C O α α Primarily anomalous scattering.

8 SAD function Total likelihood is integral product of the distributions under the black circle Normal scattering Anomalous scattering Phase probability H -* F o - F o <H - H -* >

9 Phaserphasing scenarios Known anomalous substructure missing atoms model errors (xyz, occ, B) weak normal component Known partial structure missing (anomalous) atoms (partial) model errors weak anomalous component

10 SAD phasing Parameters: atom type position occupancy B-factor f Input Substructure atoms H, H - estimates SAD-function

11 MR-SAD phasing Input: partial structure possible atom types Heavy atom parameters H, H - estimates SAD-function

12 SAD function H -* F o H -* F o <F o - F o -* > <F o - F o -* > F o - F o - Scenario: anomalous substructure Scenario: partial MR model

13 Log-likelihood gradient maps Peak height dependent on f /f ratio distinguish atom type (fooled if normal scattering is present at site) Very sensitive to detail e.g. anisotropic temperature motion Iterative improvements in substructure description increases sensitivity

14 Log-likelihood gradient maps Δ ano LLG1 LLG2 LLG3 Br Co Na Na Na Element f f Na P K Ca

15 F1-ATPase testcase 3500 residue protein complex 81 S atoms, 3.50 Å resolution Wavelength: 1.77 Å (f S =0.72 e - ) Anomalous signal extends to 21.5 Å, or to 12.7 Å very optimistically. 1bmf - Abrahams J.P., Leslie A.G.W., LutterR., Walker J.E. Nature (1994) 370:621 2ck3 - high resolution (1.95 Å ) equivalent Bowler M.D., Montgomery M.G. Leslie, A.G Walker J.E. Proc. Natl. Acad. Sci. USA (2006) 103:8646

16 F1 ATPasetestcase pdb 2CK3 Present sites Met Cys PO 4 Ion ? Found sites Model Met Cys PO 4 Ion F1-ATPase F1 ATPase 3α chains Yeast ATPase

17 F1 ATPasetestcase ADP in active site (not included in model) P Mg 2 P Located sites

18 Phaser Macrocycle Outlier rejection Substructure refinement convergence? Substructure editing Find new atoms in LLG map Reject atoms with low occupancy Phase calculation Automated mode

19 Data flow in substructure solution Measured data Anomalous differences Solution Electron density map Anomalous substructure Phasing

20 Rescore trials with phaser LLG Insulin full search against 1.7 Å data Solved 400 Not solved CC Clearly separated solution and non-solution groups Impossible to identify solutions

21 CC for phaser-refined substructures LL LG Insulin full search against 1.7 Å data Solved 400 Not solved Lower CC for non-solutions Refinement: reduced overfitting CC Increased CC for solutions Completion: improved substructure

22 Fibronectintestcase Full search with CC rescoring Solution buried in noise LLG CC Rudiño-PiñeraE, RavelliRB, SheldrickGM, NanaoMH, KorostelevVV, Werner JM, Schwarz-Linek U, Potts JR, Garman EF J Mol Biol. 368,

23 Fibronectintestcase Full search with phaser rescoring Clear solutions LLG CC Rudiño-PiñeraE, RavelliRB, SheldrickGM, NanaoMH, KorostelevVV, Werner JM, Schwarz-Linek U, Potts JR, Garman EF J Mol Biol. 368,

24 Patterson seeding Dual-space recycling Substructure seeding Substructure is available at this point Compare: atoms compatible with the difference Patterson atoms compatible with the SAD LLG map Refinement and scoring

25 Comparison Transaminase, 60 Se sites Patterson seeding LLG-map seeding CC scoring Phaser scoring KSM42, 160 Se sites Patterson seeding LLG-map seeding CC scoring Phaser scoring 28 37

26 When to use Phaser scoring default run default give no solution LLG-map seeding default run large substructures with phaser scoring?

27 Phaser Mode SAD (start from heavy atoms) MRSAD (start with partial MR model) Reflection data F /F - possible space groups Scatterer information atom type and wavelength measured f /f values Absolute scale macromolecule weight

28 Results Phase information FWT, PHWT HLA, HLB, HLC, HLD PHIB FLLG, PHILLG PDB file anomalous substructure Best map or initial map for density modification Phase probability for density modification or refinement Centroidphase for density modification/refinement (use HL-coefficients!) Log-likelihood gradient map (flat after automated mode!) Logfile

29

30 Acknowledgements Python-based Hierarchical Environment for Integrated Xtallography Computational Crystallography Initiative / LBNL Paul Adams, Ralf Grosse-Kunstleve, PavelAfonine, Nathaniel Echols, Jeffrey Headd, Nigel Moriarty, Nicholas Sauter, Peter Zwart Los Alamos National Lab Tom Terwilliger, Li-Wei Hung UNIVERSITY OF CAMBRIDGE University of Cambridge Randy Read, Airlie McCoy, Robert Oeffner, Gabor Bunkoczi DUKE University Duke University Jane and Dave Richardson, Vincent Chen