E N D
1. An Introduction to Rational Drug Discovery: QSAR & CADD
2. Goal: How to judge whether or not a molecule might be active, based on only structural info? Basics of Protein-Ligand Binding
QSAR & Pharmacophore Modelling
Building a model for ACE Inhibitors
3. Drug Discovery
4. Drug Discovery High attrition rate of compounds
Far less than 1 percent of screened compounds enter preclinical testing
For every 250 compounds in preclinical testing, only 5 survive to enter clinical testing
With only 1 approved by the FDA after an average of 15 years of total R&D!!!!
Costs run >$500,000,000
5. Integrated Drug Discovery Target Discovery & Validation
Genomics
Proteomics
Lead Identification
High Throughput Screening
Diverse Compound Repository
Lead Optimization
Medicinal Chemistry/Targeted Synthesis
Combinatorial Chemistry
Computational Chemistry
Lead/Pre-clinical Development
Formulation & Process Chemistry
In Vivo Optimization
Pharmacology, PK, Toxicology
6. “Da Basics” :Ligand-Protein Binding where Ki is the kinetically derived inhibition constant and Kd is the dissociation constant (see MM kinetics)
7. “Da Basics” :DGb, the Free Energy of Binding
8. “Da Rules”:Protein-Ligand Binding Binding affinity:
Predominantly determined by hydrophobic interactions
Depends on ‘buried’ non-polar surface area
Entropy driven
Depends on states (conformation, protonation, tautomeric, etc.) of unbound species
Binding specificity:
Structural complementarity (steric & electronic factors)
Hydrogen bonding (enthalpy driven)
9. Hydrogen Bonding Hydrogen bonds contribute to specificity but not significantly to binding affinity.
A desolvation penalty must be paid when water is ‘removed’ from the ligand and the protein binding sites
This results from the loss of H-bonds between the ligand or protein and solvent waters
This penalty is approximately balanced by the H-bonding energy gained through ligand-protein interactions.
A significant free-energy price is paid for unsatisfied H-bonds, which reduces the overall binding free energy.
The desolvation penalty results from the loss of H-bonds between the ligand or protein and solvent waters
10. Binding Affinity & DGb A factor of 10 in binding affinity,
usually measured as Kd ,
is equivalent to
~1.3 kcal mol-1 of binding free energy at room temperature
11. Rational Drug Discovery:Lead Optimization Goals
? Binding Affinity
? Specificity
? Novelty – Intellectual Property (IP)
? Synthetic Accessibility
? Cell Penetration
? ADME/Tox Prop’s. Bioavailability, Plasma Stability, pKa, pKD, etc.
MEET PROJECT TEAM OBJECTIVES!
12. Rational Drug Discovery:Lead Optimization Have a “hit” and a validated “lead series:
QSAR
Pharmacophore Modelling
Active Site Modelling (to be discussed another time)
Docking Experiments
‘Virtual Screening’ or ‘vUltra-HTS’
13. What is QSAR? (Quantitative Structure Activity Relationships) Activity = function (structural properties)
Example*:
Relate biological activity to electronics and hydrophobicity
log(1/C) = k1logP – k2(logP)2 + k3s + k4
C = concentration of a compound that gives a response
P = partition coefficient between water & 1-octanol
K1,k2,k3,k4 = constants
s = Hammett substituents parameter
*C. Hansch. (1969) Accounts of Chemical Res. 2:232-239
14. QSAR Objective Correlate 3D structure of a ligand with its biological activity
Problems:
Frame of reference
Unknown steric interactions
Multiple binding modes
Underdetermined system
15. QSAR Process Synthesize & test biological activity for diverse set of ligands (including actives and inactives).
Clever experimental design maximizes information content from a series.
Regression techniques to fit an equation to the data.
Which properties are correlated with activity?
Choose/calculate Descriptors
Cross-validate.
Predict, wary of extrapolation vs. interpolation
16. Assumptions/Requirements of QSAR Models All analogs belong to congeneric series.
All analogs have the same mechanism of action.
All analogs bind in a similar fashion.
The effect of isosteric replacement can be predicted.
Binding affinity is correlated with interaction energy (e.g., ionic effects are approx. const.)
Biological activity is correlated with binding affinity (e.g., not with transport properties).
17. Types of QSAR Models What distinguishes various QSAR methods from one another is how the descriptors are estimated and how the equations are solved.
“2-D QSAR” methods tend to use 2-D molecular information.
“3-D QSAR” methods tend to use 3-D molecular information.
Linear regression methods (LS, PLS, PCA, Neural nets,…)
Nearest-neighbor estimation (Clustering, B-cuts)
18. What is a Descriptor? Molecular Descriptors:
Calculated:
Solution to a mathematical procedure that transforms chemical information into a number
surface areas (polar, non-polar), dipole moment, volume
Experimental:
the result of some standardized experiment to measure a molecular attribute
melting point, partition coeficients, refractive index, etc.
19. Why do we need Descriptors? Why do we need descriptors?
Describe different aspects of molecules
Compare different molecular structures
Compare different conformation of same molecule
Database storage
Relate structure to activity (QSAR)
Find best set of descriptors relevant to activity
20. Types of Descriptors 0D (zero dimensions)
Chemical formula
Atom counts
1D
Fragment count
2D
Topological representation
3D
Geometrical representation
4D
3D + scalar field
21. Types of Descriptors Counts of features: For example HBAs, HBDs, aromatic ring systems, substructures/fragments ( e.g. , carbonyl groups, basic nitrogens, carboxyl groups,…),etc.
Physicochemical Properties: LogP, solubility, MW, MP, BP, heat of sublimation, molar refractivity, Hammett parameters, etc.
Topological Indices: Wiener index, branching indices, kappa shape indices, electrotopological state indices, atom-pairs, topological torsions, etc.
BCUTs (3-D, 2-D, 2-T): Electrostatic, charge, and polarizability (hydrophobic).
Others: Volsurf, polar surface area, etc.
22. Dimensionality Reduction Too many descriptors! Need to reduce to manageable size!
PCA – Principle Component Analysis
Choose those descriptors that correlate with activity
23. Lipinski’s “Rule of Five” LogP < 5
MW < 500
Number of HBD < 5
Number of HBA < 10
24. Why QSAR Fails
25. Pharmacophore Models Paul Erlich, early 1900:
“a molecular framework that carries (phoros) the essential features responsible for a drug’s (pharmacon) biological activity”
Peter Gund, 1977
“a set of structural features in a molecule that is recognized at a receptor site and is responsible for that molecule’s activity”
Wikipedia, today
A pharmacophore is a 3-dimensional substructure of a molecule that carries (phoros) the essential features responsible for a drug’s (pharmacon) biological activity.
Alternatively described as an ensemble of interactive functional groups with a defined geometry.
Basically, one tries to talk the protein language by finding the “structural and chemical complementaries” (aka pharmacophore hypothesis) to target receptors
26. Pharmacophore Modeling Analog-based pharmacophores
Unknown receptor
Information from active and inactive molecules
3D-QSAR
Receptor-based pharmacophores
Known (and characterized) receptor
Docking
27. Pharmacophore Modeling Obtain activity data on an appropriate set of compounds (range of activities)
Carry out conformational analysis (multiple conformations, ~5.0 kcal mol -1 energy window)
Identify “ pharmacophoric features”
Identify common feature patterns(2-point, 3-point, 4-point,… pharmacophores; multiple solutions)
Search compound collections with pharmacophore “models”
Screen selected compounds
“Structure-activity relationships without chemistry”
Pharmacophore methods also provide guidance in synthesis & design
28. Visual Pattern Recognition
29. Visual Pattern Recognition Visual identification of common structural and chemical features among active molecules and those features that are missing in the inactive ones
Measurement of the 3D aspects of the common features, w. r. t. each other
Development of a draft pharmacophore
Validation that the pharmacophore fits the active compounds and fails to fit the inactive ones
Refinement of the model by applying it to a database of compounds with known activity, until the desired result is reproduced
30. Visual Pattern RecognitionOne aromatic ring: phenyl or pyridyl
31. Visual Pattern RecognitionSecond 5- or 6- member ring
32. Visual Pattern RecognitionUrea group or amide functionality in 2nd ring
33. Visual Pattern RecognitionRing systems are side-by-side
34. Measurements From the amide nitrogen to the center of the aromatic ring is approx 5-6 Ĺ
The aromatic ring and amide group are within 0.5 Ĺ RMSD (Root Mean Square Difference) from planarity
35. Developing a pharmacophore model Phenyl or pyridyl ring
5- or 6- member ring with link option
Additional nitrogen, various bond types
Distance and planarity constraints
36. Phramacophore Features
37. ‘3’-Point Pharmacophore
38. ‘4’-Point Pharmacophore
39. Active Compounds in a Pocket
40. Search a Compound Library for ‘Hits’ – e.g. matches to Pharmacophore Model
41. Pharmacophore Based Screening Hits
42. 3D-QSAR: CoMFAComparative Molecular Field Analysis Select test and training sets of comparable diversity.
Generate 3-D structures (single conformations).
Structurally Align and place in a box with grid
43. 3D-QSAR: CoMFA1st need to structurally align 2-D Alignment Methods
Maximum common substructure based methods
Feature-Based Vector Methods
Discrete feature values (“Bit Strings”)
Continuous feature values
3-D Alignment Methods
Field-based methods
Structure-based methods
(generalized RMSD approaches)
Key Assumption
All molecules bind to the same region of the target binding site in essentially the same way!
44. 2-D & 3-D AlignmentHIV-1 RT Inhibitors
45. 2-D & 3-D AlignmentHIV-1 RT Inhibitors
46. 3D-QSAR: CoMFAComparative Molecular Field Analysis Calculate property fields for each molecule at every grid point (training set)
Property value at each grid point is equivalent to a descriptor value in 2-D QSAR
Grid points with low variance may be neglected; nevertheless this may result in hundreds of grid points
Many more descriptor values than experimental data points, thus “traditional” least-squares approach cannot be used
Perform partial least squares (PLS) analysis
Validate model (test set)
Predict activities of new molecules
47. 3D-QSAR: CoMFA
48. Angiotensin Converting Enzyme (ACE)* Overall pressor (pressure-raising) effect
Converts deca-peptide Angiotensin I to Angiotensin II
AII = potent vasoconstrictor
Stimulates release of steroid aldosterone
Results in increased sodium retention
Hydrolyzes C-terminal peptide from the hypotensive (vasodilator) nona-peptide bradykinin
Stimulates antidiuretic hormone (ADH) release
ACE action increases both vascular resistance And blood volume
Inhibiting ACE results in antihypertensive effect
Slides from Modeling Biosystems I Lectures by Dr. Dan Kuster (Marshall lab) Washington State University @ St. Louis
49. Angiotensin Converting Enzyme (ACE) ACE “makes” Angiotensin II in vivo
50. Angiotensin Converting Enzyme (ACE)
51. How to inhibit ACE? Bradykinin-Potentiating Peptide (BPP)
Naturally occurring nonapeptide in Bothrops jararaca venom
BPP5a = Glu-Lys-Phe-Ala-Pro
Phe-Ala-Pro also inhibits ACE
Many studies to deduce structure-activity relationship with Phe-Ala-Pro mimics
Complicated by pKa at titratable groups, etc
Simple computer models
Several inhibitors emerge as potent
Good peptide-analog correlation for captopril
Poor peptide-analog correlation for enalapril
How to Inhibit ACE?
52. Captopril Designed w.r.t. BPP5a
First ACE inhibitor drug
Orally available
Potent
Undesirable clinical side effects
Loss of taste
Rashes
At therapeutic doses, does not interact with nervous system
Similar sulfhydryl moeity as penicillamine, similar side effects
53. Find an Active Conformation Apply the active analogue hypothesis
Choose a potent inhibitor (IC50<50nm)
Scan systematically through all torsions
Generate many conformations
Reject those with improbably large energy (VdW criterion)
Multiple-Minima Problem
# of conformers = (360/angle) N-rejected
N= # of rotatable bonds
Small angles or large # of bonds ? Combinatorial explosion!
Create distance map (DMAP) relating structural features relevant to binding site
54. Distance MAPs (DMAPs) Du = Zn
DMAP valid for all active analogs
DMAP can become many dimensional for some systems
55. Grids How know if two DMAPs are equal?
Define a 3D grid around molecule
Two conformers evaluate to same grid points?
Effective resolution of predicted DMAP
~0.1- 0.5 angstrom
Fine grid rejects more conformations
Balance angle resolution with grid spacing!
56. Can this molecule place the right atoms in the necessary spacing?
57. Analogue Search Add fields and pharmacophore features to grid
Screen for best fits.
58. 3D-QSAR: An ‘Active’ Hit
59. 3D-QSAR: An ‘InActive’ Hit
60. Summary Problems finding new drugs efficiently and cost-effectively!
Computers may help the process!
QSAR
Pharmacophore Models
CADD
But still not enough known- the best you can do is enhance ‘hit-rate’
The experiments still need to be done – no one will ever take or pay for a ‘virtual drug’!!!!
61. Acknowledgements Slides/Figures from:
Dr. Dan Kuster (Marshall lab)
Washington State University @ St. Louis
Professor Gerry Maggiora
Department of Pharmacology & Toxicology,College of Pharmacy College of Pharmacy University of Arizona University of Arizona