Transcript week10

Computer Aided Drug/Vaccine
Design and Case Study
Dr. Haifeng Chen
Shanghai Jiaotong University
2013 Nobel Chemistry Prize
The prize was awarded for laying the foundation for the computer
models used to understand and predict chemical processes.
Reference Book



Xiaojie Xu and Tingjun Hou. Computer Aided Drug
Desing. Chemistry Industry Press, 2004.
Kaixian Chen. Computer Aided Drug Design:
Principal, Method and Appication. Shanghai Science
and Technology Press, 2000.
Textbook of Drug Design and Discovery. Edited by
Povl Krogsgaard-Larsen, Tommy Liljefors, Ulf
Madsen，published by Taylor & Francis，2002.
Reference Book


A. R. Leach. Molecular Modelling.
Principles and Applications. Addison
Wesley Longman, Essex, England, 2001.
Broad introduction to many aspects of
molecular modeling and computational
chemistry techniques, covering basic
concepts, quantum and molecular
mechanics models, techniques for
energy minimization, molecular
dynamics, Monte Carlo sampling, free
energy simulations, and drug design
applications
SCI Article of Undergraduate Student
1. Z. Li, J. Han, H. F. Chen*. Chem. Biol. Drug Des. 72:350-359, 2008.
（2005）（Chinese Academy of Sciences）
2. Z. Li, H. Zhang, Y. Li, J. Zhang, H. F. Chen*. Chem Biol Drug Des
77:63-74, 2011. （2005）
3. F. Qin, Y. Chen, Y. X. Li, H. F. Chen*. J. Chem. Phys. 131: 115103,
2009. （2005）（SJTU）
4. H. Zhang, F. Qin, W. Ye, Z. Li, S. Ma, Y. Xia, Y. Jiang, J. Zhu, Y. Li, J.
Zhang, H. F. Chen*. Chem Biol Drug Des 78:427-437, 2011. （2008）
（Duke University）
5. G.W. Yan, Y. Chen, Y. Li, H. F. Chen*. Chem Biol Drug Des 79:916-925,
2012. （2011）（Michigan State University）
6. S.Y. Ma, W. Ye, D. J. Ji, H.F. Chen*. Medicinal Chem. 9: 420 – 433, 2013.
(2009) （Purdue University）
7. K. Wu & H.F. Chen*. Chem Biol Drug Des 2014，in press. (2011)
（Carnegie Mellon University）
Ebola
15 October 2014
Pig Flu/A H1N1
From 1997- now
Avian Flu/H7N9
2002-2003
SARS
Human and animal infectious diseases










Plague (6th century)
- Death rate : 30%～100％
Cholera (18th century)
- Death rate : 30%～100%
Anthrax (19th century)
- Death rate : 20%
Ebola virus (1976)
- Death rate : 50%～90%
HIV (1980)
- Death rate : 61%
Mad cow disease (1985)
- Death rate: 100%
Avian flu (1997)
- Death rate : 33.3%
Pig flu (2009)……
H7N9 (2013)
11
Drug discovery of post genomics
Function
genomics
Target
discov
ery
Target
evaluat
ion
Lead
disco
very
Lead
optimiz
ation
Preclinic
test
Clinical
trail
Market
Drug development flowchart
New Chemical
Entity
Structure
optimization
Preclinical test
(ADMET)
New drug research
application
Clinical trail
(I II III)
New drug
application
Post market
research
Target Identification
Disease
analysis
Drug design flowchart
Clinic test
Drug screening， Potential
drug discovery, side effects
Drug Discovery Today 7: 315-323 (2002)
Virus analysis of avian flu
N Engl. J. Med. 2013,20:1888-1897.
Drug target
Receptor
Enzyme
Ion channel
Nucleic acid

Success target : 300－400
Science 2013, 341:84-87.
Success Cases

HIV-1 Protease Inhibitors in the
market:




Inverase (Hoffman-LaRoche, 1995)
Norvir (Abbot, 1996)
Crixivan (Merck, 1996)
Viracept (Agouron, 1997)
Drug discovery today 2, 261-272
(1997)
Merck HIV-1 protease drug (Crixivan)
Screening
L-F35524
Crystal
1987-1988
Clinical test for
4000 samples
1993
1989
1996.3.
1987
1989
Sequence
Function
Clone
1989-1992
Inhibitor
research
1993-1996
Human test
FDA approve:
42 days
18
Challenge of drug development
New chemical entity: Difficult
Time: long (10-15 years)
Cost : expensive (800 million US$)
Method: do not speed (Combine Chem. & HTS)
How to speed?
19
Challenge of drug development
20
Main methods of CADD
Statistics Math
Statistical mechanics
QM/MM
Enzyme catalyst
Quantum mechanics
Molecular mechanics
Conformer Search
Molecular Dynamics
Newton Second Law
Classic MD
Ab initio MD
Monte Carlo Simulation
21
21
22
22
Most used technologies
24
Computer aided drug design Method
24
25
QSAR，Quantitative Structural-Activity Relationship
Cl
Cl
Molecule
Structure

Hasch (1962): Hansch analysis

Richard Cramer III (1987): CoMFA

Gerhard Klebe (1994): CoMSIA

Lowis (1997): HQSAR

Vapnik (1992/2001):SVM
Tin Kam Ho (1995/2003):Random
Forest

Biological activity(Φ)
IC50, Ki
…
26
Molecules Are Not Numbers!
27
Molecular Descriptors
Hansch classic QSAR
HQSAR(Holograph QSAR)
Fragment size
Number of fragments
Atom types
Bond types
Atom hybridization
Stereocenters
PLS analysis result
QSAR Comb. Sci. 23:36-55, 2004.
Principle and application
of 3D-QSAR
Method of 3D-QSAR
Bioactivity
3D-QSAR Model
3D-QSAR


CoMFA (Comparative Molecular
Field Analysis )
CoMSIA (Comparative Molecular
Similarity Indices Analysis )
The hypothesis condition of CoMFA
All molecules
 Have same interaction mechanism with
the same kind of receptor (or enzyme,
ion channel,etc.)
 Have identical binding sites in the same
relative geometry.
3D-QSAR steps
Create a 3D database
Calculate charges
for each of compounds
(Gasteiger-Hückel)
Conformer
search
Minimize the structure
(Tripos force field)
Alignment
Training set and test set
(3:1)
Calculate the steric and electrostatic field energies
(Steric and electrostatic contributions were cutoff
to a value of 30 kcal/mol)
Do regression analyses
(partial-least squares (PLS))
Perform using full cross-validation
(leave one-out method)
r2 value
(q2)
Contour maps
Conformer Search




Gridsearch
Multisearch
Random search
System search
Alignment Rules

Pharmacophore-based alignment
Pharmacophore is a spatial arrangement of
atoms or functional groups which response for
bioactivity

Structure-based alignment
MCSS(Maximum Common Substructure) or
skeleton structure

Dock-based alignment
Active conformer could align together by the
results
of molecular docking
Grid and probe atom
Probe atom
Box must cover the structures.
The type of probe atom





Sp3 C+
Sp2 OSp3 N+
H+
Ca2+ …
Potential function of
CoMFA
Partial Least Square
Contour
Maps
Predictions
Bio
QSAR Table = SYBYL MSS
PLS
QSAR
equation
Construct CoMFA Model




PLS Component
Field contribution
Steric and electrostatic contour plots
Relationship between EA and PA
Interpretation of CoMFA
Green: bulk group
Yellow: small group
Red: negative charge
Blue: positive charge
For drug design, the most powerful tool is to find the relationship
between fields and bioactivity, then design
new lead compounds.
Advantages of CoMFA vs
Classical QSAR





Visualization
Higher predictive power
Truly three-dimensional, shape-dependant
nature of CoMFA descriptors
CoMFA analyzes the interaction energy of an
entire ligand rather than arbitrarily selected
substructure of the ligand
CoMFA has been accepted by many as the
ultimate solution to the problem of correlating
chemical structure and biological activity
Shortcomings of CoMFA




CoMFA parameters do not include
hydrophobicity
Need to specify initial “alignment rule”
and “active conformation”
Often fail when a few molecules are
very dissimilar from all others
The results from one CoMFA analysis
are not easily compared with another
one
Factors of influence CoMFA






Active conformer
Aligment rules
Probe atom
Lattice size
Orientation of alignment molecular set
Step size
CoMSIA





Steric fields
Electrostatic fields
Hydrophobic fields
Hydrogen bond donor fields
Hydrogen bond acceptor fields
Potential function of CoMSIA
CoMSIA is not sensitive to changes in orientation of
the superimposed molecules in the lattice.
Interpretation of CoMSIA
Yellow: hydrophobic
↑;
White: Hydrophobic
↓
Cyan: hydrogen bond donor ↑; Purple: hydrogen bond donor ↓
Magenta:hydrogen bond acceptor ↑; Red: hydrogen bond acceptor ↓
Shanghai High school-SJTU Join
Program
Drug Design of HIV-1
Protease Inhibitor
Student ： Guanwen Yuan（Shanghai
High School)
Supervisor： Haifeng Chen（SJTU）
Content
1
Backgrounds
2
Methods
3
Results
4
Discussions
About AIDS

Infect : 60,000,000
WHO

Death : 30,000,000

1,050,000 (2008) 23.3 billion $ China

Up to now, no bacterin
http://www.cmt.com.cn/xshy/gr/AIDS2010/AIDS2010/201007/t20100714_263494.html
Vaccine design
Envelope trimer
Science 2013,DOI: 10.1126/science.1245627
Computer Aided Vaccine
design
Computers Aid Vaccine Design
Michael Hagmann
Epitope：
Increase efficiency of discovery
epitope for 10~20 times.
Decrease 95% experiment.
Decrease cost.
Improve the speed of discovery.
Science 2000,290:80-82.
Computational B cell
vaccine design method
B cell Epitope：5~20 residues.
Sequence-based methods
Propensity scales
(high concentration of charged and polar residues, lacking in large
hydrophobic residues)
antigenic index/flexibility scale/hydrophobicity scale/ turn scale.
Machine learning techniques : ROC curve
Structure-based methods
(3D structures of antibody–antigen complexes)
Substructure search
Feature mapping
Computational T cell
vaccine design method
T cell Epitope：~9 residues.
Sequence-based methods
Binding motifs
Decision trees
Artifi cial neural networks
Support vector machines
Hidden Markov models
Structure-based methods



Protein threading
Homology modeling
Molecular docking
Vaccine design Case
T cell
Nature, 2014,507:201-206.
AIDS of China
Number of Death
14000
12000
10000
8000
6000
4000
2000
0
2006
2007
2008
Year
2009
2010
HIV-1 Life Cycle
Nature Medicine
5, 740 - 742 (1999).
Integrase Inhibitors
Integrated site
Science, 26 June 2014;
DOI:10.1126/science.1254194
HIV-1 drug target
CCR5 1IKY 2013
HIVRT 1HNI 1995
HIVIN 3L3V 2010
HIVPT 1HXW 1995
Therapeutic method
HIV RT inhibitor
Cocktail therapeutics
HIV Protease inhibitor
AIDS
HIV Integrase inhibitor
Cocktail Types
New anti-HIV inhibitor
Drug resistance
Combination of drugs


Dolutegravir（整合酶抑制剂）+
Abacavir （阿巴卡韦）（非核苷HIVRT
inhibitor）+ Lamivudine（拉米夫定）
（核苷类似物）（DTG-ABC-3TC）
EFV （非核苷HIVRT inhibitor）(泰诺
福韦)-TDF （核苷类HIVRT inhibitor）
(替诺福韦酯)-FTC（核苷类HIVRT
inhibitor）(恩曲他滨)
N Engl J Med. 7, 2013; DOI: 10.1056/NEJMoa1215541
Background

HIV Protease (HIVPR)

HIV-1 codes p55 and p60



HIVPR can break pre protein and activate
protein. Specific enzyme of virus.
HIVPR is the key enzyme of mature for
HIV-1 virus.
Drug target
Binding mode between
inhibitor and HIVPR

作用机理：
抑制剂与酶结合→使酶失去催化
活性→阻断HIV在体内的复制→
抗AIDS的药效
Simulation open and close of HIVPR
PNAS 109:20449–20454, 2012.
Motion with PCA
Research Methods

Computer Aided Drug Design

Molecular Dynamics (MD) Simulation

3D-Quantitative Structure-Activity
Relationship (3D-QSAR)
Data Set
Bioorg Med Chem Lett 2003, 13, 3601-3605.
Molecular dynamics simulation







Molecular dock: M17-HIVPR, M35-HIVPR
AMBER8.0 & Parm99SB force field
5000ps simulation - 298K
Result analysis
Hydrogen bond
Hydrophobic interaction
Binding free energy
Molecular Dock





Calculating the binding free energy
Finding the molecular mechanism
between ligand and receptor
Finding the relationship between
bioactivity and binding free energy
Using with other methods(Find active
conformer to build CoMFA model)
Virtual screening
The AutoDock Software

Developed by AJ Olson’s group in 1990.

AutoDock uses free energy of the docking molecules using 3D
potential-grids

Uses heuristic search to minimize the energy.

Search Algorithms used:

Simulated Annealing

Genetic Algorithm

Lamarckian GA (GA+LS hybrid)
Docking complex
Ligand
Receptor
Docking preparing - ligand





Assign charges
Align with tempale molecule
Define rotatable bonds
Merge non-polar hydrogens
Write .pdbq ligand file
Receptor preparing






Delete water and ligand in complex
Add polar hydrogen
Load charge (Kollman_all)
Minimize hydrogen atom
Add solvation parameters
Write .pdbqs protein file
Autodock result
Correlation between Binding Free
Energy and bioactivity
9
8
Log(1/EC50)
7
6
5
4
-16
-15
-14
-13
-12
-11
-10
-9
-8
-7
Binding free energy (kcal/mol)
pIC50=0.759 - 0.503*△G
(n=76, r=0.739, F1,75=89.217, SD=0.861)
3D-QSAR Study





Construct structure-bioactivity model,
predict drug bioactivity and virtual screen
Molecular alignment, then calculate force
parameter
PLS parameter represents the quality of
model. R2 ，accuracy
Test set can evaluate model
Contour plot
3D-QSAR Study




SYBYL7.0 program package
Training set & Test set (3:1)
(Random class)
Molecules Alignment
CoMFA & CoMSIA models
Results

Molecular Dynamics Simulation

The most important hydrophobic and
hydrogen bonding interactions

Action Mechanism of Inhibitors
Stability of complex
3.5
M17-HIV PR
M35-HIV PR
RMSF (Å)
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0
25
50
75 100 125 150 175
Residues
M35 < M17
Distance (Å)
Hydrogen bond
8
7
6
5
4
3
8
7
6
5
4
3
2
0
Ala28-O6(M17)
Asp25-O6(M17)
A
Asp30-F3(M35)
Asp25-O1(M35)
B
1000
2000
3000
4000
5000
Time (ps)
Bioactivity：M35>M17
M35/HIVPT: Two strong hydrogen bonds
M17/HIVPT: Two weak hydrogen bonds
Hydrophobic interaction
Population (%)
100
M17
M35
80
60
40
20
0
1 2 3 4 5 6 7 1 2 3 4 5 6 7 8 9 10
Native contact
Bioactivity：M35>M17
M35/HIVPT: 10
M17/HIVPT: 7
Binding free energy
Binding mode and key
residue
Common
residues：
Ile50(A)
Ile50(B)
Asp25(A)
Ile83(A)
Ala28(A)
Conclusion of MD simulation

Similar action mechanism

both systems have hydrogen bond with catalytic
residue of Asp25 of HIVPR

Key residues: Ile50(A), Ile50(B), Asp25(A), Ile83(A),
and Ala28(A)

Hydrogen bond offered by the OH

Strong hydrophobic interaction offered by the
benzene ring
Result of 3D-QSAR
Force Combination of
CoMSIA
F
1/SEE
R
2
Q
2
1.0
0.8
0.7
0.6
0.5
8
6
80
60
40
SE
SEH SED
SEA SEHA SEHD SEDA SEHDA
Force field
Best CoMSIA model：SEHA
Calculated activity
Prediction ability
3.5
3.0
2.5
2.0
4.0
3.5
3.0
2.5
2.0
1.5
1.5
CoMFA
Training set
Test set
CoMSIA
2.0
2.5
3.0
Experimental activity
r2 of test set : 0.939 (CoMFA)
0.825 (CoMSIA)
3.5
Contour plot analysis

CoMFA & CoMSIA
Field
+
—
Steric
green
yellow
Electro
static
blue
red


X: bulk & positive
charge
Y: small volume &
positive charge
Contour plot of CoMSIA
Field
+
Hydrophob Orange
ic
Hydrogen cyan
bond
acceptor


X: hydrophilic substitute
Y: hydrophobic & hydrogen bond donor
white
purple
Conclusion of 3D-QSAR
Y
O
X
O
O
O




O
X
O
X: bulk and positive groups
M30(4-COOCH3) > M28(4-CH3) > M29(4-CN) > M31(4-COOH);
M1(4-H) > M21(4-F)
Y: small and positive charge groups
M14(4-CH3) > M15(4-CN) > M13(4-CF3) > M16(4-COOCH3) >
M18(4-CH2OH) > M20(4-CONH2) > M17(4-COOH).
X：hydrophilic group
M32（4-CH2OH）>M30（4-COOCH3）; M32>M31（4COOH）; M32>M28（4-CH3）； M32 >M29(4-CN)
Y: hydrophobic and hydrogen bond donor group
M16(4-COOCH3)>M17(4-COOH); M14(4-CH3)> M19(4-CH2NH2)
Y
MD vs 3D-QSAR





MD: two hydrogen bonds between
M35(F3/O1)and Asp25/Asp30
3D-QSAR: hydrogen bond favour regions
F3/O1
MD: M35 has hydrophobic interactions with
Ile50(A), Ile50(B), Ile83(A), and Ala28(A).
3D-QSAR: Benzene is covered by hydrophobic
favour regions.
The result of MD is consistent with that of 3DQSAR.
Comparison with previous
work
PA  0.246 X _ p  0.198B1 X _ p  0.148 Y _ p  0.252mrY _ p  0.156 B5Y _ p
 0.191N X _ F  0.151N Y _ F  0.246 N X _ O  0.116 I Y _ p  0.461I Y _ Hbond _ do _ p
 0.189 I Y _ di _ F  0.219 I Y _ F _ O  2.997
Predicted activity
3.5
This work
2
r = 0.970
Previous work
2
r = 0.867
3.0
Quality: same
2.5
Quantity：better
2.0
1.5
1.5 2.0 2.5 3.0 3.5 4.0 2.0 2.5 3.0 3.5 4.0
Experimental activity
Conclusion




MD simulation suggests interaction
mechanism, key hydrogen bond and
hydrophobic interactions.
3D-QSAR method constructs robust
prediction model.
The results of MD are agreement with
those of 3D-QSAR.
Better than previous works.
Prize of this program
First prize of
Shanghai
Second prize of
China
Pulications



Insight into the Binding Mode of HIV-1 Protease
Inhibitor Using Molecular Dynamics Simulation and
3D-QSAR. Chem Biol Drug Des 2012. (IF=2.50)
Insight into the Stability of Cross-β Amyloid Fibril
from Molecular Dynamics Simulation. Biopolymers
93: 578-586, 2010. (IF=2.82)
Conformational Selection or Induced Fit for Brinker
and DNA Recognition. Physical Chemistry
Chemical Physics. 2011 (IF =4.06)
Forecoming research of
HIV-1 protease
Mutant residues：
V32I
I50V/L
I54M/V
I84V
L90M
A71V
Complex of HIV-1 PR and nelfinavir
Drug resistant？
Thank You!