An Integrated Approach to Protein-Protein Docking

1
An Integrated Approach to Protein-Protein Docking

• Zhiping Weng
• Department of Biomedical Engineering
• Bioinformatics Program
• Boston University

2
What is Protein Docking?
Protein docking is the computational
determination of protein complex structure from
individual protein structures.
3
Motivation

• Biological activity depends on the specific
  recognition of proteins.
• Understand protein interaction networks in a cell
  
• Yield insight to thermodynamics of molecular
  recognition
• The experimental determination of protein-protein
  complex structures remains difficult.

4
Ubiquitination
5
Experimental Tools for Studying Protein-Protein
Interactions

• 3-D structures of protein-protein complexes
  X-ray crystallography NMR
• Binding affinity between two proteins SPR,
  titration assays
• Mutagenesis and its affect on binding
• Yeast 2-hybrid system
• Protein Chips?

6
Computational Tools for Studying Protein-Protein
Interactions

• Protein docking
• Binding affinity calculation
• Analysis of site-specific mutation experiments
• Protein design
• The kinetics of protein-protein interactions

7
Protein-Protein Interaction Thermodynamics
8
The Lowest Binding Free Energy DG
water
9
General Derivations
10
Two kinds of docking problems

• Bound docking
• The complex structure is known. The receptor and
  the ligand in the complex are pulled apart and
  reassembled.
• Unbound docking
• Individually determined protein structures are
  used.

11
Challenges

• Large search space
• Imperfect understanding of thermodynamics
• Protein flexibility
• Heterogeneities in protein interactions

12
Divide and Conquer

• Initial stage of unbound docking
• Assume minimum binding site information
• Try to predict as many near-native structures
  (hits) as possible in the top 1000, for as many
  complexes as possible
• Post-processing
• Re-rank the 1000 structures in order to pick out
  near-native structures

13
Energy Components
14
An Effective Binding Free Energy Function
Number of atoms of type i Desolvation energy for
an atom of type i
15
Fast Fourier Transform
Increase the speed by 107
16
DOCK by Kuntz et al.
17
Evaluate Performance

• Gold Standard Crystal structure of the complex
• A near-native structure (hit) RMSD of Ca after
  superposition lt 2.5 Å
• Success rate Given some number of predictions,
  percentage of complexes with at least one hit

18
Docking Benchmark
55 non-redundant complexes http//zlab.bu.edu/ron
g/dock/
19
(No Transcript)
20
Post-Processing Using RDOCK
21
CAPRI Results
Target Total Contacts Top Predictions Our Prediction
1 52 17 (1st) 5
2 52 27 (2nd) 50 (1st)
3 62 45 (1st), 43 (2nd) 37 (3rd)
4 58 1 (1st) 0
5 64 10 (1st) 4 (2nd)
6 65 60(1st) 18
7 37 30(2nd,3rd), 29(4th,5th) 31(1st)
22
Target 2 Antibody/VP6 Red Crystal
Structure Blue Prediction 50/52 1st
23
Target 7 T Cell Receptor / Toxin Red Crystal
Structure Blue Prediction 31/37, 1st
24
Target 3 Antibody/Hemagglutinin Red Crystal
Structure Blue Prediction 37/62, 3rd
25
Target 6 Camelide Antibody/a amylase Red
Crystal Structure Blue Prediction 18/65
26
Target 1Hpr/HPrK Red Crystal Structure Blue
Prediction 5/52
27
Summary

• Conformational change tolerant target functions
  are needed for unbound docking
• We need to balance shape complementarity,
  desolvation, electrostatics components
• If we submit 10 predictions, we have a 60
  success rate.

28
Future Work

• An automatic protein-protein docking server
• Large scale comparison of all docking algorithms
  on the benchmark
• Post processing with binding site information,
  conformation space search, clustering and
  detailed free energy calculation
• Make predictions!