Transcript Lect 10:Computer aided drug design: structure-based

SMA5422: Special Topics in
Biotechnology
Lecture 10: Computer aided drug
design: structure-based approach.
Chen Yu Zong
Department of Computational Science
National University of Singapore

Drug design overview.
 Introduction of methodology.
 Examples: drug resistance,
toxicity prediction.
Traditional Drug Design
Methods: Random screening


Long design cycle: 7-12 years.
High cost: $350 million USD per
marketed drug.
Drug Discovery Today 2, 72-78 (1997)
Too slow and costly to meet demand.
Strategies for improving design
cycle:

Smart screening:
• High-throughput robotic screening.

Diversity of chemical compounds:
• Combinatorial chemistry.
Nature 384 Suppl., 2-7 (1996)
High expectation.
Alternative approach?

Current situation:
• Molecular mechanism of disease processes,
structural biology.
• Rising cost of experimental equipment and
resources.
• Computer revolution (low cost, high power).
• Software development.
Computer approach?
Strategies for improving design
cycle:

Computer-aided drug design:
• Receptor 3D structure unknown:
• QSAR.
Pharm. Res. 10, 475-486 (1993).
• Receptor 3D structure known:
• Ligand-protein docking.
Science 257, 1078-1082 (1992)
Is ligand-protein docking
practical?

3D structure of proteins and small molecules:
• 15,000 protein entries in PDB,
growth rate: ~100-200 per month.
• 250,000 small molecules in ACD.

Computation time:
• 100,000 small molecules per week.
Nature 384 Suppl., 23-26 (1996)

Computer cost:
• Decreasing dramatically.
Success Stories:

HIV-1 Protease Inhibitors:
•
•
•
•
Inverase (Hoffman-LaRoche, 1995)
Norvir (Abbot, 1996)
Crixivan (Merck, 1996)
Viracept (Agouron, 1997)
Drug discovery today 2, 261-272 (1997)
Examples of Other Drugs Designed by
Structure-Based Methods:

Human renin inhibitor
Antihypertension.



Collagenase and stromelysin inhibitor
Anticancer and antiarthritis.
Purine nucleotide phosphorylase inhibitor
Antidepressant.
Thymidylate synthase inhibitor
Antiproliferation.
Nature 384 suppl, 23-26 (1996)
Favourable Conditions for
Application of Ligand-Protein Docking
Human Genome Project
Protein Crystallography
Functional Genomics
 Ligand-Protein Docking
Pharmacogenomics
Molecular Biology
Modeling Technology
Information Technology
Computer-aided drug design in Industry
and Premier Universities

Pharmaceutical Giants:
• Merck, Abbott, Bristol-Myers Squibb, Pfizer,
Glaxo-Welcome.

Biotech New and Emerging Stars:
• Agouron, Arris, Chiron, ISIS, MetaXen, Vertex.

Major Universities:
• Harvard, UCSF, UC Berkeley, Washington U,
Cambridge, Columbia.
Computer-aided drug design in Industry

Structure-based design viewed as having
competitive edge:
• An indication: Companies are withholding 3D
structures of key proteins.

Modeling group viewed as a key component in
drug discovery team:
• Many companies have setup modeling group.

Investment in computer equipment:
• An indication: Glaxo-welcome bought 100 SGI
workstations in 1996.
Ligand-Protein Docking is the
Most Rational Approach:
Reason: Based on receptor structure

Mechanism of drug action:
Mechanism of drug action:
Mechanism of drug binding:
Ligand binding mechanism
.
.
.
.
Scoring Functions in
Ligand-Protein Docking
Potential Energy Description:
Scoring Functions in
Ligand-Protein Docking
Potential Energy Description:
Scoring Functions in
Ligand-Protein Docking

Potential Energy Description:
• van der Waals interactions
• Electrostatic interactions
V = ligand atoms [ Aij1/2 Arec- Bij1/2 Brec+ qiQrec ]
Modelling Strategy for
Ligand-Protein Docking
Rceptor Cavity Model
Ligand-Protein Docking Algorithm
Utilization of geometric features
Docking Evaluation (Chemical Complementarity)
Local energy minimisation to release bad contacts
More realistic potential functions
Average CPU time:
5,000 small molecules per week
The Use of Molecular Mechanics Energy
Functions in Docking Evaluation

Potential Energy Description:
•
•
•
•
Hydrogen bonding
van der Waals interactions
Electrostatic interactions
Empirical solvation free energy (energy evaluation only)
V = H bonds [ V0 (1-e-a(r-r0) )2 - V0 ] +
non bonded [ Aij/rij12 - Bij/rij6 + qiqj /r rij] +
atoms i Dsi Ai
Example 1: Study of Drug Resistant
Mutations by Ligand-Protein Docking
Enzyme-inhibitor
PDB Id
Mutation introduced
HIV-1 protease + MK 639
1HSG
HIV-1 protease + Saquinavir
HIV-1 protease + SB 203386
HIV-1 protease + VX 478
HIV-1 protease + U89360e
1HXB
1SBG
1HPV
1GNO
V82A, V82F, V82I, I84V, V82f/I84V, M46I/L63P,
V82T/I84V, M46I/L63P/V82T/I84V
V82F, V82I, I84V, G48V, V82F/I84V, V82T/I84V
I32V/V47I/I82V
M46I/L63P, V82T/I84V, M46I/L63P/V82T/I84V
V82D, V82N, V82Q, D30F
HIV-1 RT + Nevirapine
HIV-1 RT + TIBO R82913
1VRT
1TVR
L100I, K103N, V106A, E138K, Y181C, Y188H
L100I, K103N, V106A, E138K, Y181C, Y188H
J. Mol. Graph. Mod. 19, 560-570 (2001).
Quality of Modelled Structures
Wild type X-ray structure:
Blue
Modelled mutant:
Red
Mutant X-ray structure:
Green
Mutation induced energy change compared
with observed drug resistance data
J. Bio. Chem.271, 31947 (1996)
AIDS 12: 453 (1998)
Biochemistry 37, 8735 (1998)
Modelling Strategy
for Ligand-Protein Induced Fit:
Generation of multiple conformations
Identification of key rotatable bonds
constructive to conformation change
Algorithm for collective rotation
of backbone bonds
Geometric constraint that mimics
chain restoring forces
Adjustment of side chain orientation
minimisation of side chain packing along pathway
Energy barrier along pathway
Based on molecular mechanics energy functions and force fields
Example 2: Prediction of toxicity, side effect,
pharmacokinetics and pharmacogenetics
by a receptor-based approach
Annu. Rev. Pharmacol Toxicol 2000,
40:353-388
1997, 37:269-296
Importance of prediction of side effect, toxicity,
pharmacokinetics in early stages of drug discovery



Drug Candidates
in Different Stages of Development
Majority of Candidates Fail to Reach
Market
Clin Pharmacol Ther. 1991; 50:471
Most drug candidates fail
to reach market
Pharmacokinetics (60%),
side-effect and toxicity
(40%) are the main reason.
Large portion of money
(USD$350 million) and
time (6-12 years) spent on
a clinical drug has been
wasted on failed drugs.
Drug Discov Today 1997; 2:72
Strategy
Existing Methods:
New Method:
Given a Protein,
Find Potential Binding Ligands
From a Chemical Database
Given a Ligand,
Find Potential Protein Targets
From a Protein Database
Compound Database
Protein Database
Compound 1
...
Compound n
Protein 1
...
Protein n
Protein
Ligand
Successfully Docked Compounds
as Putative Ligands
Successfully Docked Proteins
as Putative Targets
Science 1992;257: 1078
Proteins 2001;43:217
Feasibility
Proteins
 Database: >14,400 3D structures in PDB.
 Protein diversity: 17% in PDB with unique sequence.
 Advance in structural genomics: 10,000 unique proteins within 5 years.
Ann. Rev. Biophys. Biomol. Struct. 1996; 25:113
Nature Struct. Biol. 1998; 5:1029
Method
 Ligand-protein docking docking algorithms capable of finding binding
conformations.
Proteins. 1999; 36:1 Proteins 2001; 43:217
Additional information
 Rapid accumulation of knowledge in proteomics, pathways, protein
functions.
Computer resources
 Increasing power and decreasing cost (Linux PC, Multi-processor machines)
Automated Protein Targets Identification
Software INVDOCK
Ligand
\|/
Automated Process to inversely dock the Lignad to each entry in
a Built-In Biomolecular Cavity Database (10,000 Protein and Nucleic Acid Entries)
\|/
Step 1: Vector-based docking of a ligand to a cavity
Step 2: Limited conformation optimization on the ligand and side chain of biomolecule
Step 3: Energy minimization for all atom in the binding site
Step 4: Docking evaluation by molecular mechanics energy functions and comparison with other ligands
Successfully Docked Proteins and Nucleic Acids
as Potential Targets of a Ligand
|
\|/
Potential Applications:
\|/
Protein function, Proteomics, Ligand transport, Metabolism
Therapeutic Targets, Side-Effects, Toxicity
Pharmacokinetics (ADME)
INVDOCK Testing on Toxicity Targets
Compound
Number of
experimentally
confirmed or
implicated
toxicity targets
Number of
toxicity
targets
predicted by
INVDOCK
Number of
toxicity
targets
missed by
INVDOCK
2
Number of
toxicity targets
without 3D
structure or
involving
covalent bond
4
Number of
INVDOCK
predicted
toxicity targets
without
experimental
finding 2
Aspirin
15
9
Gentamici
n
17
5
2
10
2
Ibuprofen
5
3
0
2
2
Indinavir
6
4
0
2
2
Neomycin
14
7
1
6
6
Penicillin
G
7
6
0
1
8
Tamoxifen
2
2
0
0
4
Vitamin C
2
2
0
0
3
Total
68
38
5
25
29
J. Mol. Graph. Mod., 20, 199-218 (2001).
Toxicity and side effect targets of Aspirin
identified from INVDOCK search of protein database
PDB
Protein
1a42
Carbonic anhydrase II
1a6a
HLA-DR3
1a7c
Plasminogen activator
inhibitor
1d6n
Hypoxanthine-guanine
phosphoribosyltransferase
1hdy
Alcohol dehydrogenase
Experimental
Finding
Target
Status
Toxicity/Side
Effect
Ref
Activate
enzyme
activity that
may lead to
increase in
plasma
bicarbonate
Change in
concentration.
HLA level
Implicated
Metabolic
alkalosis
(hypoventilation).
Puscas I
Implicated
Aspirin-induced
asthma
Dekker
JW
Tissuedependent
response of
protein.
Implicated
Hypertension,
thrombolysis
Smokoviti
sA
Excess uric acid
in serum*
Inhibition of
activity
Confirmed
Increased blood
alcohol level
Gentry RT
J. Mol. Graph. Mod., 20, 199-218 (2001).
Conclusion:

Structure-based computer aided drug design
is a promising approach.

Revolution in molecular biology and
computer technology sets the stage for this
approach.

Much remains to be done.