Protein Surface Analysis for Functional Analysis and Prediction T. Andrew Binkowski and Andrzej Joachimiak 2009 NIGMS Workshop: Enabling Technologies for.

Protein Surface Analysis for Functional Analysis and Prediction

T. Andrew Binkowski and Andrzej Joachimiak

2009 NIGMS Workshop: Enabling Technologies for Structural Biology

March 4-6, 2009

Outline

How Can Surface Analysis Aid Your Structural Genomics Effort?

Protein Surfaces

Comparing Surfaces of Proteins

Surface Analysis in the Structural Genomics Pipeline

The Global Protein Surface Survey

3

Functional Inference in Proteins

Transfer function based on similarity to a protein with known biological activity

Sequence 30-70% Functional sites result from spatial

interactions of key residues in diverse regions of primary sequence

Structure Reveal more distant relationships 1 fold ~ many functions; vice versa

Example: generalized secondary structural element

Different SSE can bring residues in spatial proximity (Jaroszewski & Godzick, ISMB 00)

4

Functional Inference in Proteins

Functional surfaces may be the most conserved structural features of proteins Surfaces performing identical biochemical activity can be found within different

protein scaffolds or in the absence clear evolutionary relationships

Exploit ability of proteins to preserve local spatial residue patterns Presents another opportunity to infer insightful ideas about their biological function

and mechanisms

Novel heme-monooxygenase•12% sequence identity• vs. all •Experimentally verified activity

5

Surfaces of Proteins

Surface: Local grouping of solvent accessible

atoms

Pockets: Empty concavity on a protein

surfaces into which solvent can gain access

Identifying surfaces: Methods:

Solvent accessibility, Geometry, Grids, Spheres

Applications: CASTp, Surfnet, Pocket, Ligsite, Pass

Our approach: Computational geometry (alpha

shape) CASTp, PDB, Swiss-Prot,

Catalytic Site Atlas Ligand binding surfaces:

Exclusion contact surface (solvent accessibility difference)Muck & Edelsbrunner, ACM Tran Graph, 1994; Edelsbrunner, Facello, Liang, Disc Appl Math, 1996; Liang, Edelsbrunner,

Woodward, Protein Sci, 1998

6

Global Protein Surface Survey

http://gpss.mcsg.anl.gov

7

Comparing Surfaces of Proteins

SurfaceScreen Methodology for identifying similarly

shaped proteins and aligning them

Optimizes two components Global Shape

Perceived similarity Size and scale, independent of

chemistry

Local physicochemical texture Preserved atom/residue orientation Conservation of chemical

complimentarity

Global Surface Shape Filtering

Surface

Constrained SpatialSurface Refinement

Apply ScoringFunctions

Surface ShapeAlignment

8

Comparing Surfaces of Proteins:Global Shape Similarity

Surface Shape Signatures (SSS) Represent signature of a surface

as distribution sampled from a shape function (Osada et. Al., 2002)

Comparison of probability distributions

Kolmogorov-Smirnov Earth Mover’s Distance

ATP Binding sites protein kinase CK2 from Z. mays (b) phosphopantetheine

adenylyltransferase from E. coli (c) maltose/maltodextrin transport

protein from E. coli (d,cyan chain A, light blue chain B)

50 non-homologous sites (< 30% sequence identity)

9

Spatial Surface Alignment Refinement

Combinatorial comparison of residue sets in “neighborhood” Maintain “like” correspondence

of types Maximum common residues

Enumerate and evaluate alignment orientations Find optimal superposition

using SVD of correlation matrix (Umeyama 1991)

Heme binding pockets of myoglobin from different organisms.

10

Evaluating Surface Alignments

Surface Volume Overlap:

Interpretation of SVOT is not straightforward Need global and local

BABAAB VVVV ABBBAA

ABAB VVV

VSVOT

ABBBAA

ABAB VVV

VgSVOT

abbbaa

abab VVV

VlSVOT

RMSD Distance: Estimate the probability of obtaining a specific

RMSD for nres

Compute random surface alignments (108) and build lookup tables

RMSD variants: cRMSD (coordinate) oRMSD (orientation)

Benchmarking Surface Alignments

11

12

Heme Binding Site Retrieval

seq. & fold

surface analysis

Heme (iron-protoporphyrin IX) Multi-functional (i.e.oxygen binding/transport,

electron transfer and redox)

Binding on 20 different folds Between proteins <2% seq. id.

Query myoglobin (gray) against PDB structure to identify hemoproteins Retrieval rate (area under ROC

curve) Sequence: 68.7% Structure (SSM): 64.4% Surface: 95.8%

Detection of convergent heme binding site on IsdG from S. aureus Missing characteristic sequence motif 12% seq id; different scaffold Experimentally verified

monooxygenase activity

13

ATP: Retrieval of a Flexible Ligand

Adenosine 5’-triphosphate multifunctional nucleotide (i.e.cell signaling, enegry transfer)

58 unique EC classifications #.#.#.# Conformational flexibility

Retrieval rates for 4 conformations (79.1%-85.4%); method is tolerant to flexible ligands

14

Prediction and Validation of GDP Binding Surface

Structure of F420-0:gamma-glutamyl ligase from A. fulgidus

Large binding surface was searched to support functional predictions and GDP binding surface is identified Posed GDP based on superposition

of surfaces (red)

Co-crystallization experiments validates prediction

Surface Analysis in the Structural Genomics Pipeline

15

Exploiting Protein Surfaces in Structural Genomics

16

Developing surface-based tools to address specific needs of structural genomics pipeline

Ligands for co-crystallization Aid in the assignment of

electron density Functional annotation tools Drive further studies (i.e.

ligand binding, discovery)

Co-

crys

talli

zatio

nM

utat

ion

Functional Analysis

Ligand

Identification

Future StudiesDiscovery

17

Surface Identification

Partially Solvedor Low Quality

Structure

Search GPSS forBinding Sites

Co-crystallizationExperiments

Crystallization/Structure Improvement

Introduction of GDP to F420-0:gamma-glutamyl ligase from A. fulgidus improves resolution from 2.8 to 1.9 Angstroms and orders loop regions.

18

Assisted Electron Density Assignment

Unidentified ligand density

Construct surface surrounding density and search against ligand surface library Does not require entire structure

to be built

19

Assisted Electron Density Assignment

Applicable to ligands of various molecular weights and sizes Fructose (pdb id=1zx5) NADP (pdb id=2ag8)

Suggest a list in cases of ambiguity

20

Landscape Analysis: ATP

Classification based on surface similarity shows functional families have preferred (not necessarily unique) surfaces and conformation

21

Automated Protein Kinase Classification

All-against-all surface comparison of all protein

kinases in the PDB Color labeled by expert annotation (KinBase)

Surface clustering identifies: Dual substrate specificity of CK2 proteins Active/inactive states

Similarity detected between MAP p38 kinase and Abelson leukemia virus tyrosine kinase (Abl) with bound cancer drug STI-571 MAP kinase has unique DFG “out” conformation not

previously seen in ser/thr kinases

22

Function Sleuth

Conserved protein of unknown function (VCA0319) from V. cholerae apc29617

Unique arrangement of common structural motifs Problematic for secondary structure

and fold analysis

Surface analysis identifies DNA binding surface and 5 putative metal binding sites All 5 metal binding sites showed

strong preference for Mg

Putative metalloregulated repressor with Mg-regulated mechanism of DNA binding

1bdb NAD

1hoh MGD

2qwr ANP

1jbw ACQ

Function Sleuth

23

Target APC7761 (3fd3)Agrobacterium tumefaciens str. C58

Function Sleuth

Target APC61725 (3fz5)

Rhodobacter sphaeroides 2.4.1

Top 17 most similar surfaces bind B12

24

1i9c

Global Protein Surface Survey

SurfaceScreen for PSI ‘function sleuth’ targets Automated analysis of largest 5 surfaces

(per chain and unit)

Technical Note: DOE INCITE on Blue/GeneP at ANL

25

http://gpss.mcsg.anl.gov

26

Conclusion

Comparing surfaces of proteins can be a useful tool with many applications Functional characterization Assisted electron density assignment Automated classification

Global Protein Surface Survey http://gpss.mcsg.anl.gov

2727

Acknowledgements

ANL/MCSGH. An, G. Babnigg, L. Bigelow, A. Binkowski, C-s. Chang, S. Clancy, G. Cobb,M. Cuff, M. Donnelly, C. Giometti,W. Eschenfeldt, Y. Fan,C. Hatzos, R. HendricksG. Joachimiak, H. Li, L. Keigher,Y-c. Kim,N. Maltseva, E. Marland,S. Moy, R. Mulligan,B. Nocek, J. Osipiuk, M. Schiffer,

G. Montelione, Ruthgers Univ. NESGCT. Terwilliger, Los Alamos, ITCSGZ. Derewenda, Univ. of Virginia, ITCSG Z. Dauter, NCIJ. Liang, Univ. of IllinoisD. Sherman, U. Michigan

Washington Univ.D. Fremont,T. Brett, C. Nelson,

Univ. of VirginiaW. Minor, M. Chruszcz, M. Cyborowski, M. Grabowski, P. Lasota, P. Miles,M. Zimmerman, H. Zheng

Univ. of Texas SWMCZ. Otwinowski, D. Borek, A. Kudlicki, A. Q. Mei, M. Rowicka

Northwestern Univ. W. Anderson, O. KiryukhinaD. Miller, G. Minasov, L. Shuvalova, X. Yang, Y. Tang

Univ. College London @ EBI, J. Thornton, C. Orengo, M. Bashton, R. Laskowski, D. Lee, R. Marsden, D. McKenzie, A. Todd, J. Watson

Univ. of Toronto A. Edwards, C. Arrowsmith, A. Savchenko,E. Evdokimova, J. Guthrie, A. Khachatryan, M. Kudrytska, T. Skarina, X. (Linda) Xu

Univ. of ChicagoO. Schneewind, D. Missiakas, P. Gornicki, S. Koide, ITCSGW-j. Tang,B. Roux,J. L. RobertsonM.R. Rosner,T. Kossiakoff, ITCSGV. Tereshko,

Funding: NIH and DOE

ANL/MCSG A. Sather,G. Shackelford,L. Stols, K. Tan,C. Tesar,R-y. Wu, L. Volkart, R-g. Zhang, M. Zhou,ANL/SBCN. Duke, S. Ginell,F. Rotella

Thank you