Outline of this Lecture

1. Electrostatic Potentials (10 minutes)
2. Other Molecular Interaction Fields (5 minutes)
3. 2D QSAR (10 minutes)
4. 3D QSAR (15 minutes)
   1. Active Analog Approach
   2. CoMFA
5. 10-minute break
6. Docking (20 minutes)
7. Automated Ligand Design (20 minutes)

Optional material is shown in gray boxes.

Drug Discovery

• up to 10 compounds were clinically tested in humans
• up to 100 compounds were tested for safety
• up to 10,000 compounds were synthesized & tested for activity

Electrostatic Potentials

• association rates
• binding constants
• functional group properties:
  change in pKa values (where pKa is strongly dependent on ionization energy)
  change in redox midpoints (electron transfer), change in acid-base equilibria (proton transfer)
  change in stabilization energy
  changes due to solvation effects; free energy of solvation

Response of system to electrostatic forces:

• dipole reorientation (dielectric = 2.5 to 4)
• charge rearrangement, especially due to mobile ions (dielectric > 30)
• electronic polarizability (dielectric = 2 to 2.5)
  
  • polarity: density of charged & dipolar groups
    polarizability: potential for reorganizing charges, orienting dipoles, inducing dipoles.
    polarizability: polarity plus the freedom to reorganize.
  • A folding protein has high polarity, high polarizability.
    A folded protein has high polarity, low polarizability.
  • Small protein dipole reorientations: polarization can be represented by a dielectric model.
    Large protein dipole reorientations: polarization can't be represented by a dielectric model---must use explicit charge distribution.

Calculating the Electrostatic Field

• by atom type (does NOT account for 3D conformation)
• by Quantum Mechanical calculations.
  Milliken population anaysis example.
  ElectroStatic Potential Fit (ESP Fit) method (adjusts Milliken population analysis to approximate charge distribution of molecule).

Other Molecular Interaction Fields

Hydrophobicity: Measure (log P) of solvent partition coefficient of a molecule between water and an organic solvent (usually octanol, which has a dielectric constant and hydrogen bond/molecular weight ratio that is similar to a protein "core"). Used as a qualitative measure of entropy of solvent.

2D QSAR Analysis

• chemical information systems
• synthesis planning, reaction databases
• property prediction
• pharmacophore generation based on structure

3D QSAR

1. Evaluate minimum energy conformations via conformational analysis techniques.
2. Use Principle Component Analysis (PCA) to optimize conformational descriptors:
   Distance between two atoms
   Orientation of ring system
   Vector of hydrogen bond donor/acceptor, or charged functional group
   In general, a few Principle Components account for most of the system.
3. Perform Cluster Analysis classification methods to select binding conformation The "active cluster" should have a high "concentration" of active compounds and a low "concentration" of inactive compounds.
   The lowest energy conformation of each structure in the active cluster is the structure's "binding conformation".
4. Determine properties of the pharmacophore:
   

   dipole moment	HOMO & LUMO
   polarizability	shape
   point charges	volume
   molecular electrostatic potential

5. Select mot important pharmacophore properties
   

Active Analog Approach

• Based on hypothesis that all ligands interact with biological target in a similar, specific conformation---the "Binding Conformation".
• All active ligands must have all necessary functional groups and all must be bioisostereic (all necessary functional groups must behave int eh same way during interaction with the binding site).
• Generates a geometric pharmacophore.

CoMFA:
Comparitive Molecular Field Analysis

• Depends on ligand conformation
  
• Depends on molecular alignments
  • atom-by-atom pairwise RMSD fit
  • fit of molecular interaction fields and associated properties
  • fit of molecular surfaces
• Depends on parameters describing Interaction Fields
• Depends on scoring schemes.
  Scoring Schemes are not reliable---not well correlated with relative binding affinities. This is likely related to the problem of rigid ligand conformations.
• Doesn't explicitly account for hydrogen bonds.

Docking

1. manual docking with a grid
2. "Traction Beam" docking with energy minimization/molecular dynamics
3. Monte Carlo docking

1. docked molecules are usually rigid
2. Scoring Schemes are not reliable---not well correlated with relative binding affinities. This is likely related to the rigid docking weakness.

Ligand Design

• USING JUST ONE PROGRAM MAY NOT COVER ALL POSSIBILITIES.

• It is VERY difficult to account for flexibility of receptor
• Must consider water molecules in active site (Lam et. al at Dupont Merck did successful lead discovery by binding water to LIGAND during CAMD).

Atom-by-Atom ligand design: Better for ligands with conformational & structural variations slower than fragment-addition method can generate ligands that are not chemically or synthetically plausible Examples:

• LEGEND (authors: Nishibata & Itai)
• CONCEPTS (authors: Pearlman & Murcko)
• GROWMOL (author: Bohacek)

Fragment-addition ligand design:

• Better for ligands with less variability
• more likely to create chemically & synthetically plausible structures
• Examples:
  1. GROW (authors: Moon & Howe; natural peptide residue fragments only)
  2. LUDI (author: Bohm, MSI)
  3. GroupBuild (authors: Rotstein & Murcko)
  4. SPROUT (authors: Gillet & co-workers)
  5. MCSS (authors: Miranker & Karplus) and HOOK (Holland & co-workers)

ligand design by pharmacophore pattern:

• Ususally used when no receptor structure is available.
• works best if you have ring structures
• Examples:
  1. CAVEAT (author: Bartlett & co-workers)
  2. NEWLEAD (Tschiniki & Cohen)
  3. LINKOR (authors: Itai & co-workers)
  4. RECEPS (Kato and co-workers)

Bohm's Scoring Function for Free Energy Upon Binding
Interaction/Motion	Free Energy
loss of translational & rotational motion of ligand	+5.4 kJ/mol
loss of freedom for each rotatable bond in ligand	+1.4 kJ/mol
generation of ideal hydrogen bond	-4.7 kJ/mol
generation of ideal ionic interaction	-8.3 kJ/mol
lipophilic contact	-0.17 kJ/ang2

SCORING BASED ON GEOMETRY:

SCORING BASED ON ENERGY:

Correlation between Score and Dissociation Constant

Score {100 log(Ki}	Ki
100	100 mM
200	10 mM
300	1 mM
600	1 uM
900	1 nM

LINK SCORING: