Jean-Charles BOISSON. New models for the docking problem. Image from the ArgusLab software. Outline. The docking problem: Protein Structure Prediction (PSP). Protein Folding Problem (PFP). Existing software and visualization tools. Existing Models: Coding. Evaluation function. Methods.
How can a ligand be « docked » in a binding site ?
Only one conformation for the ligand and the site:
The ligand can have several (or a lot of) conformations for the site:
The ligand and the site can have several (or a lot of ) conformations:
Designing a new drug:
Find the ligand that can interact with a given site.
The number of existing ligands is enormous (million).
All the ligands can not be tested in vitro.
How do we choose the right ones to test ?
Thanks to expert opinion.
Allows to screen the right ligands for a site.
The remain ligands can be tested with a high confidence.
Drug designing is boosted !!!
In the flexible DP:
The ligand and the site have to modify their shape to find the best complementarity
How do we find the new conformations ?
The ligand and the site have to form a stable complex
How do we find the best complex conformation ?
From a protein in its linear form (amino acid chain), find the final 2D or 3D structure.
A problem with a large search space.
No unique solution there are several equivalent solutions (due to the environment).
Very consuming process for « real » protein.
How do a protein exactly fold itself to gain its 3D structure ?
Finding the moves to reach a stable state.
A problem with a large search space.
Very time consuming process…
A paradox in the real life, fastest known proteins can fold themselves in only 1 μs.
The size of the search space due to all the conformations of the ligand and the site.
The most stable states for the ligand and the site do not necessary form the right complex.
The best ligand/site complex ( with the minimal energy) is not necessary the right one.
Several sites can exist…
Need to find ALL the complexes of minimal energy.
Population based algorithm:
B.D. Bursulaya, M. Totrov, R. Abagyan, and C.L. Brooks. Comparative study of several algorithms for flexible ligand docking. Journal of Computer-Aided Molecular Design, 17(11) :755-763, November 2003.
E.A. Wiley and G. Deslongchamps. PostDock : A novel visualization tool for the analysis of molecular docking. Computing and Visualization in Science, 2005. Published online.
Simplified coding the lattice models:
HP-model (most used).
Very generic modelling.
The shape of the lattice allows to limit the number of potential conformation.
The linked evaluation methods Allow to compute « real » proteins.
A solution coordinates of each amino acid.
K.A. Dill. Theory for the Folding and Stability of Globular Proteins. Biochemistry, 24 :1501-1509, 1985.
No real energetic model.
Based on two global evaluation functions:
Maximisation of the H-H contact number.
Protein surface minimisation.
M.T. Hoque, M. Chetty, and L.S. Dooley. A Hybrid Genetic Algorithm for 2D FCC Hydrophobic-Hydrophilic Lattice Model to Predict Protein Folding. Proceedings of Australian Joint Conference on Artificial Intelligence (AI 2006), LNAI 4303 :867-876, 2006.
Complex models :
Energetic model force field utilisation:
CHARMM, AMBER, …
Combination of energetic terms:
Van der Waals,
And force terms:
For the same terms, using different force fields give different results.
Bi-objective model (most used):
Bonded atom energy:
Non bonded atom energy:
Van der Walls energy.
V. Cutello, G. Narzi, and G. Nocisia. A Class of Pareto Archived Evolution Strategy Algorithms Using Immune Inspired Operators for Ab-Initio Protein Structure Prediction. Proceedings of Evo Workshops 2005, LNCS 3449 :54-63, 2005
V. Cutello, G. Narzi, and G. Nocisia. A multi-objective evolutionary approach to the protein structure prediction problem. Journal of the Royal Society Interface, 2005. Published online.
Another bi-objective model:
Energetic interaction between the ligand and the site.
Energetic interaction between the atoms of the ligand.
Using the Autodock evaluation function.
S. Janson and D. Merkle. A New Multi-objective Particle Swarm Optimization Algorithm Using Clustering Applied to Automated Docking. Proceedings of Second International Workshop, HM 2005, LNCS 3636 :128-141, 2005.
Bonded atom energy.
Non bonded atom energy
Ordinary the shape completion is only a conformational filter before the docking process.
A. Oduguwa, A. Tiwari, S. Fiorentino, and R. Roy. Multi-Objective Optimisation of the Protein-Ligand Docking Problem in Drug Discovery. In GECCO’06, July 8-12 2006. Seattle, Washington, USA.
M.E.B. Yamagishi, N.F. Martins, G. Neshich, W. Cai, X. Shao, A. Beautrait, and B. Maigret. A fast surface-matching procedure for protein-ligand docking. Journal of Molecular Modelling, 12 :965-972, 2006.
First algorithms Metropolis algorithms.
Genetic algorithms (Autodock, GOLD, …):
Autodock hybridizing GA/local search
Lamarckian genetic algorithm
Constructive algorithms (Flex, DOCK, …).
Others proposed methods:
A Bi-objective model for PSP.
Based on the bonded and non bonded energy model.
A Lamarckian genetic algorithm for the resolution.
Using grid5000 power to compute complete protein folding.
A-A. Tantar, N. Melab, E-G. Talbi, and B. Toursel. Solving the Protein Folding Problem with a Bicriterion Genetic Algorithm on the Grid. In Proceedings of the Sixth IEEE International Symposium on Cluster Computing and the Grid (CCGRID’06), page 43, 2006.
Respectful of the repulsive/attractive system linked to the docking.
Repulsive Van der Waals energy.
Attractive Van der Walls energy.
Finding new criteria:
Hydrogen bridge number.
Van Der Waals surface.
J. Ryu, R. Park, and DS. Kim. Connolly Surface on an Atomic Structure via Voronoi Diagram of Atoms. Journal of Computer Science and Technology, 21(2) : 255-260, 2006.