Systems Biology: Applications in pharma research 20 September 2010, - - PowerPoint PPT Presentation

Systems Biology: Applications in pharma research

20 September 2010, TU München Andrea Schafferhans

Andrea Schafferhans @ TU München

Similar proteins have similar interaction partners (?)

20 January 2011 Introduction 2

Andrea Schafferhans @ TU München

Applications

• Function prediction
• Drug development

– “Target Class” approach – Side effects – “Polypharmacology” / “Network pharmacology”

20 January 2011 Introduction 3 Hopkins,A.L. (2008) Network pharmacology: the next paradigm in drug

• discovery. Nat Chem Biol, 4, 682-690.

Andrea Schafferhans @ TU München

Contents

• 1. Introduction
• 2. Protein comparison

– Computational binding site identification – Binding site comparison

• 3. Application examples

20 January 2011 Introduction 4

Andrea Schafferhans @ TU München

Types of protein similarity

• Function
• Sequence

– Paralogs – within species – Orthologs – across species

• Binding sites / interaction patterns

20 January 2011 Protein similarity 5

Andrea Schafferhans @ TU München

What is a binding site?

• Function

– Binding other proteins (e.g. signal transduction) – Binding substrates (enzymes) – Binding Co-Factors (e.g. Heme) – …

• Form

– Cavity in the protein – CAVE: induced fit / conformational selection more realistic

• Pragmatic

– Around all HETATM records in PDB (CAVE: e.g. metals…)

20 January 2011 Protein similarity 6

Andrea Schafferhans @ TU München

Binding site characteristics

• Usually a pocket or cleft in the protein
• Less hydrophobic than the interior of a protein
• Specific through complementarity of

– Form – Electrostatic interactions – Hydrogen bonds – Hydrophobic interactions

Henrich S, Salo-Ahen OM, Huang B, et al.: Computational approaches to identifying and characterizing protein binding sites for ligand design. Journal of Molecular Recognition 2010, 23:209-219 20 January 2011 Protein similarity 7

Andrea Schafferhans @ TU München

Binding site analysis – Applications

• Automated drug target annotation

– E.g. estimation of druggability (binding site size, hydrophobicity, etc.)

• Virtual screening

– Restrict the search space for docking experiments

• Function prediction
• Prediction of drug side effects

20 January 2011 Protein similarity 8

Andrea Schafferhans @ TU München

Finding binding sites – geometrically

Observation: Binding sites usually are the largest pockets e.g. 83% of enzyme active sites found in the largest pocket

(Laskowski RA, et al. Protein clefts in molecular recognition and function. Protein

• Sci. 1996; 5:2438-2452.)

20 January 2011 Protein similarity 9

Andrea Schafferhans @ TU München

POCKET

• Fill the protein with a grid (3 Å spacing)
• Mark grid points as “protein“

(within 3 Å of an atom ) or “solvent“

• Go along grid and mark “solvent” points

that lie between “protein” points for potential pocket

• Find largest “clusters” of “pocket” points

Levitt D, Banaszak L. POCKET: a computer graphics method for identifying and displaying protein cavities and their surrounding amino acids. J. Mol. Graph 1992, 10:229-234.

20 January 2011 Protein similarity 10

Andrea Schafferhans @ TU München

LIGSITE

Differences to POCKET

• More efficient searching for

neighbour atoms

• Cubic diagonals also used for

finding pockets  less dependent on orientation

• Grid points scored by the number of times they are

found (between 0 and 7)  adjustable “buriedness“

• Smaller and adjustable grid spacing (best: 0.5 to 0.75 Å)

Hendlich M, et al.: LIGSITE: automatic and efficient detection of potential small molecule-binding sites in proteins. J. Mol. Graph. Mod. 1997, 15:359-363

20 January 2011 Protein similarity 11

Andrea Schafferhans @ TU München

Finding binding sites – energetically

Binding sites interact with the bound molecules  Find location of favourable interaction energies

20 January 2011 Protein similarity 12

Andrea Schafferhans @ TU München

GRID

• Calculates interaction energies of probe molecules
• Uses three terms:

– Lennard-Jones (attraction + repulsion) – electrostatic – directional hydrogen bond

Goodford, P.J. A computational procedure for determining energetically favorable binding sites on biologically important macromolecules. J. Med. Chem. 1985 28:849-857

20 January 2011 Protein similarity 13

Andrea Schafferhans @ TU München

GRID application

• Cluster energy minima  binding site
• BUT:

– Hard to cluster – Computationally intensive

• Good for binding site characterisation

Picture from: Henrich S, Salo-Ahen OM, Huang B, et al. JMR 2010, 23:209-19.

20 January 2011 Protein similarity 14

Andrea Schafferhans @ TU München

Q-SiteFinder

• GRID methyl probe (0.9 Å grid)
• Cluster:

adjacent grid points that meet energy criterion → Success: > 70% first predicted binding site > 90% first three → 68% average precision (precision: overlap between ligand and predicted binding site)

Laurie AT, Jackson RM: Q-SiteFinder: an energy-based method for the prediction of protein-ligand binding sites. Bioinformatics 2005, 21:1908-16

20 January 2011 Protein similarity 15

Andrea Schafferhans @ TU München

i-Site

20 January 2011 Protein similarity 16

Variation of Q-Site:

• Better probe distribution

(more dense grid)

• Two energy limits

– low value for cluster seeds – higher value for extension  filtering out meaningful clusters

• AMBER force field

Morita M, Nakamura S, Shimizu K: Highly accurate method for ligand-binding site prediction in unbound state (apo) protein structures. Proteins 2008, 73:468-479

Andrea Schafferhans @ TU München

i-Site

20 January 2011 Protein similarity 17

Variation of Q-Site:

• Better probe distribution

(more dense grid)

• Two energy limits

– low value for cluster seeds – higher value for extension  filtering out meaningful clusters

• AMBER force field

Morita M, Nakamura S, Shimizu K: Highly accurate method for ligand-binding site prediction in unbound state (apo) protein structures. Proteins 2008, 73:468-479

Andrea Schafferhans @ TU München

Challenges in binding site identification

• Protein flexibility can “hide” binding sites

→ Use multiple experimental conformations → Use molecular dynamics to generate conformations

• Dimerisation has to be considered

→ Carefully look at PDB unit cell → Carefully look at information about the protein

20 January 2011 Protein similarity 18

Andrea Schafferhans @ TU München

Characterising binding sites

Properties to characterise:

• Geometry
• Amino acid composition
• Solvation
• Hydrophobicity
• Electrostatics
• Interactions with functional groups

20 January 2011 Protein similarity 19

Andrea Schafferhans @ TU München

Hydrophobicity

Measured by logP (partitioning between water and octanol)

• Map atom / residue based

contributions

• Calculate interaction

energies of hydrophobic probes (e.g. GRID)

20 January 2011 Protein similarity 20

Andrea Schafferhans @ TU München

Electrostatics

• Map electrostatic

potential onto surface

(e.g. using DelPhi, see http://structure.usc.edu/ howto/delphi-surface- pymol.html)

• CAVE:

dependence on protonation!

20 January 2011 Protein similarity 21

Andrea Schafferhans @ TU München

Functional groups

• Superstar

– Analyse the spatial distribution of functional groups in CSD  density maps – Break the protein into fragments found in CSD – Map the observed distribution of interaction partners onto the protein

Verdonk ML, Cole JC, Taylor R: SuperStar: a knowledge-based approach for identifying interaction sites in proteins. Journal of molecular biology 1999, 289:1093-108.

20 January 2011 Protein similarity 22

Andrea Schafferhans @ TU München

Binding site comparison

• Align structures in 3D
• Analyse differences and similarities of

– Amino acid composition – Local conformation – Pocket size – Presence of interaction partners

• Straightforward in case of

– Sequence similarity or – Structural similarity

20 January 2011 Protein similarity 23

Andrea Schafferhans @ TU München

RELIBASE

20 January 2011 Protein similarity 24

Andrea Schafferhans @ TU München

RELIBASE

• Stores binding sites from PDB structures
• Allows superposition of related binding sites
• Computes differences between binding sites

Hendlich M, Bergner A, Günther J, Klebe G: Relibase: Design and Development of a Database for Comprehensive Analysis of Protein-Ligand Interactions. Journal of Molecular Biology 2003, 326:607-620. http://relibase.ccdc.cam.ac

20 January 2011 Protein similarity 25

Andrea Schafferhans @ TU München

• cAMP-dependent protein kinase (1cdk)

with adenyl-imido-triphosphate

• trypanothione reductase (1aog)

with flavine-adenine-dinucleotide

20 January 2011 Protein similarity 26

Similar but not homologous binding sites

Andrea Schafferhans @ TU München

20 January 2011 Protein similarity 27

Similar but not homologous binding sites

Graphics from www.ebi.ac.uk/pdbsum/

Andrea Schafferhans @ TU München

20 January 2011 Protein similarity 28

Similar but not homologous binding sites

Graphics from Schmitt S, Kuhn D, Klebe G. Journal of molecular biology 2002, 323:387-406

Andrea Schafferhans @ TU München

Problems in binding site comparison

• Automatically locate binding site
• Capture important features in efficient representation
• Search efficiently across all structures

– Find best superimposition – Score the alignment

20 January 2011 Protein similarity 29

Andrea Schafferhans @ TU München

Binding site comparison methods

• Representation by

– Coordinate set with physico-chemical or evolutionary properties

• Atoms
• Chemical groups
• Surface points

– 3D shape descriptors

• Superimposition by

– Geometric hashing – Graph theory, clique search

• Similarity measurement by

– RMSD – Residue conservation – Physico-chemical property similarity 20 January 2011 Protein similarity 30

Andrea Schafferhans @ TU München

CavBase – Structure representation

• Cavity detection with LIGSITE (stored in Relibase)
• Cavity-flanking residues represented as pseudo-centers:

– Donor – Acceptor – Donor-Acceptor – Aliphatic – PI – several per residue if necessary

• Create Graph:

– Nodes: pseudo-centers – Edges: distances between the pseudo-centres

Graphics from Schmitt S, Kuhn D, Klebe G. Journal of molecular biology 2002, 323:387-406

20 January 2011 Protein similarity 31

Andrea Schafferhans @ TU München

CavBase – Alignment

Create associated graph:

Node: node from protein A and node from protein B 
 with similar interaction properties Edge: member nodes in protein A and B are connected
 member node distance <12Å distance difference <2Å

Find maximal common subgraph (Bron-Kerbosh)  similar arrangement of pseudo-centers in original graphs

20 January 2011 Protein similarity 32

Andrea Schafferhans @ TU München

CavBase – Scoring

• Scoring based on
• verlap of similarly

typed surface patches

Kuhn D, Weskamp N, Schmitt S, Hüllermeier E, Klebe G: From the Similarity Analysis of Protein Cavities to the Functional Classification of Protein Families Using Cavbase. Journal of Molecular Biology 2006, 359:1023-1044

20 January 2011 Protein similarity 33

Andrea Schafferhans @ TU München

SOIPPA – Structure representation

• Delaunay tesselation of Cα atoms
• > 1 tetrahedron/Cα
• Environmental boundary (red) and

protein boundary (blue)

Bourne PE, Xie L: A robust and efficient algorithm for the shape description of protein structures and its application in predicting ligand binding sites. BMC Bioinformatics 2007, 8:S9. Bourne PE, Xie L: A unified statistical model to support local sequence order independent similarity searching for ligand-binding sites and its application to genome-based drug discovery. Bioinformatics 2009, 25:i305-312.

20 January 2011 Protein similarity 34

Andrea Schafferhans @ TU München

SOIPPA – Structure representation (2)

• Each Cα characterized by

– Vector with distance and direction

• f boundaries

– Substitution matrix

• Graph:

Node: Cα Edge: connection of tetrahedra

Xie L., Bourne PE. Bioinformatics 2009, 25:i305-312.

20 January 2011 Protein similarity 35

Andrea Schafferhans @ TU München

SOIPPA - Alignment

Create associated graph:

Node: node(A) + node(B) with similar geometric potential weight: amino acid frequency profile similarity Edge: member nodes in protein A and B are connected distance difference <2Å surface normal difference <30°

Find maximum-weight common subgraph (MWCS)

Xie L., Bourne PE. Bioinformatics 2009, 25:i305-312.

20 January 2011 Protein similarity 36

Andrea Schafferhans @ TU München

SOIPPA – Scoring

• Sum over aligned residue pairs:

Residue similarity weighted by distance and normal vector angle

• Statistical significance of score

Background score distribution: – compare unrelated structures with random sequences – fit resulting score distribution to extreme value distribution  function giving probability of randomness dependent on score

Sij = (Mij × paij × pdij )

i, j

∑

Xie L., Bourne PE. Bioinformatics 2009, 25:i305-312.

20 January 2011 Protein similarity 37

Andrea Schafferhans @ TU München

Isocleft

• Structure representation: Cα / atoms within 5 Å of ligand
• Alignment: Bron-Kerbosh of associated graph
• Scoring:

Najmanovich R, Kurbatova N, Thornton J: Detection of 3D atomic similarities and their use in the discrimination of small molecule protein-binding sites. Bioinformatics 2008, 24:i105 http://www.ebi.ac.uk/thornton-srv/databases/cgi-bin/icfdb/StartPage.pl

S = NC NA + NB − NC

20 January 2011 Protein similarity 38

Andrea Schafferhans @ TU München

Isocleft - innovations

• Two iterations of alignment:
• 1. Nodes: Cα atoms,

Edges: distance difference <3.5 Å, minimal residue similarity  Superimpose based on found graph

• 2. Nodes: all heavy atoms,

Edges: distance <4 Å, similar atom type (hydrophilic, acceptor, donor, hydrophobic, aromatic, neutral, neutral-donor and neutral- acceptor)

• Use first result of Bron-Kerbosch, then terminate

Najmanovich R, Kurbatova N, Thornton J: Detection of 3D atomic similarities and their use in the discrimination of small molecule protein-binding sites. Bioinformatics 2008, 24:i105

20 January 2011 Protein similarity 39

Andrea Schafferhans @ TU München

Example 1: Explaining side effects

Problem: side effects of ERα modulators (SERMs) Finding “off target” effects:

• Map sequences to structures (BLAST)
• Limit to “druggable” proteins (?)
• Search with SOIPPA

=> SERCA (SarcoplasmicReticulum Ca2+ channel ATPase)

20 January 2011 Application examples 40

Xie L, Wang J, Bourne PE (2007) In silico elucidation of the molecular mechanism defining the adverse effect of selective estrogen receptor

• modulators. PLoS Comput Biol 3(11)

Andrea Schafferhans @ TU München

Example 1: Validating results

• Inverse search
• Docking

– SERM – similar compounds, correlate (?)

20 January 2011 Application examples 41

Graphics from Xie L, Wang J, Bourne PE (2007) PLoS Comput Biol 3(11)

Andrea Schafferhans @ TU München

Example 2: Repositioning known drug

Problem: new tuberculosis drugs needed, but many parameters to optimise Finding compound to reuse against InhA:

• Search other structures binding Adenine

(ATP, ADP, NAD, FAD, ...)

• Compare binding sites with SOIPPA

=> SAM-dependent methyltransferases

20 January 2011 Application examples 42

Kinnings SL, Liu N, Buchmeier N, Tonge PJ, Xie L, et al. (2009) Drug Discovery Using Chemical Systems Biology: Repositioning the Safe Medicine Comtan to Treat Multi-Drug and Extensively Drug Resistant Tuberculosis. PLoS Comput Biol 5(7)

Andrea Schafferhans @ TU München

Example 2: Structure match

20 January 2011 Application examples 43

Graphics from Kinnings SL et al. (2009) PLoS Comput Biol 5(7)

Andrea Schafferhans @ TU München

Example 3: Analysing target relationships

Nodes: proteins Edges: similar binding (within factor 103)

20 January 2011 Application examples 44

Paolini,G.V. et al. (2006) Global mapping of pharmacological space. Nature biotechnology, 24, 805-15.

Andrea Schafferhans @ TU München

Example 3: Analysing target relationships

20 January 2011 Application examples 45

Paolini,G.V. et al. (2006) Global mapping of pharmacological space. Nature biotechnology, 24, 805-15.

Andrea Schafferhans @ TU München

Summary

Pharma research focus moving from

• nly individual interactions to

system oriented research Challenges:

• How to compare?
• Computational overhead

20 January 2011 Summary 46