Kein Folientitel

1
How to address synthetic accessibility in
automated De Novo Design?
2
Example 1
Combinatorial Computational Method gives new
Picomolar Ligands for a Known Enzyme Grzybowski
et al., PNAS 2002, 99(3), 1270-1273 (Group of E.
Shakhnovich, Department of Chemistry, Harvard)
3
The CombiSMoG approach
4
CombiSMoG Design of HCA inhibitors
Kd 30 pM
5
CombiSMoG Design of HCA inhibitors
6
Example 2
Synergy between Combinatorial Chemistryand De
Novo Design Leach et al., J. Mol. Graphics Mod.
2000, 18, 358-367 (GlaxoSmithKline)
7
Combinatorial De Novo Design
8
Combinatorial De Novo Design
9
Combinatorial De Novo Design
10
Combinatorial De Novo Design
11
Combinatorial De Novo Design
12
Combinatorial De Novo Design
13
Combinatorial De Novo Design
14
Combinatorial De Novo Design
Decompose
Generalise query
Search Reagent Database
15
Combinatorial De Novo Design
16
Combinatorial De Novo Design
17
Example 3
Structure-Based Generation of a New Class of
Potent CDK4 inhibitors New De Novo Design
Strategy and Library Design Honma et al., J.
Med. Chem. 2001, 44, 4615-4627 (Banyu Tsukuba
Research Institute)
18
De Novo Design of CDK4 inhibitors
19
De Novo Design of CDK4 inhibitors
20
De Novo Design of CDK4 inhibitors
Soaking into CDK2 binding site confirmed proposed
binding mode
21
Docking Scoring
22
Docking

• Instead of stepwise and interactive addition of
  fragments, whole ligands from a database are
  docked into a protein binding pocket
• Quality of ligand-receptor match is measured by
  scoring functions
• Several standard programs
• DOCK (Kuntz et al., UCSF, USA)
• FlexX (Rarey et al., GMD, Germany)
• GOLD (Jones et al., Univ. Sheffield, UK)
• AutoDock (Olson et al., Scripps, USA)

23
The DOCK approach

• Score ligand-receptor match

24
Scoring Functions in DOCK

• Contact Scoring
• Only number of ligand atom - receptor atom
  contacts are counted, regardless of atom types
• Force field Scoring
• Non-bonded terms (Van-der-Waals and Coulomb) are
  calculated by a force field

25
The FlexX Approach
Step 4 Scoring
26
The FlexX Approach
27
The GOLD Approach

• GOLD (Genetic Optimization for Ligand Design)
  uses a genetic algorithm (GA) for ligand-protein
  docking
• GAs imitate nature's method for adapting to a
  changing environment
• GAs are general optimization algorithms which can
  be applied to numerous optimization problems
• GOLD applies a force field scoring method
  (similar to DOCK)

28
The GOLD Approach - Genetic Algorithms
29
The GOLD Approach - Genetic Algorithms

• GAs in general require
• a fitness function
• a way to describe the individuals ("chromosomes")
  
• operations to alter the chromosomes (typically
  mutation and crossover)
• For Docking
• Fitness function ligand-receptor scoring
  function(i.e. those individuals that yield
  "better" binding energy are "fitter" and have a
  higher chance of reproduction)
• Chromosomes bit string of torsion angles and
  rotations/translations of molecule in binding
  pocket

30
The GOLD Approach - Genetic Algorithms
31
The GOLD Approach - Genetic Algorithms
32
The GOLD Approach - Genetic Algorithms
Conformational change upon mutation
33
The GOLD Approach - Genetic Algorithms
Crossover Operator
34
The GOLD Approach - Genetic Algorithms