Copyright © 2006 Turquoise Consulting. All Rights Reserved Pharmacophore Perception and Use in Computer-Aided Drug Design Osman F. Güner, Ph.D.
Jan 29, 2016
Copyright © 2006 Turquoise Consulting. All Rights Reserved
Pharmacophore Perception and Use in Computer-Aided Drug Design
Osman F. Güner, Ph.D.
Copyright © 2006 Turquoise Consulting. All Rights Reserved2
Outline
Pharmacophores: definition Examples: Cardiotonic drugs ACE Inhibitors
Function vs. Topological Query Concept (Angiotensin II blockers)Pharmacophore model and hit list analysisPeptidomimetic design strategy (HIV-2 inhibitors, endothelin)Training set selection (PDE IV inhibitors)Model refinement - Clustering and merging (5HT3 inhibitors)Predictive models – (Antimalarial agents, FPT inhibitors)Receptor-based pharmacophores - Dealing with multiple-binding modes (DHFR inhibitors, HIV-1 Protease inhibitors)Bibliography Reviews and published success stories
Copyright © 2006 Turquoise Consulting. All Rights Reserved3
Pharmacophore, Definition
Copyright © 2006 Turquoise Consulting. All Rights Reserved4
Pharmacophore – Value!
“Perceiving a pharmacophore is the most important first step towards understanding the interactions between a receptor and a ligand.” From the Preface of “Pharmacophore Perception, Development,
and Use in Drug Design,” Güner, O.F. ed., International University Line, 2000, La Jolla, p xv.
With a pharmacophore model, you can: Search databases to retrieve compounds that match the model Design and enhance compounds to better fit the models Align molecules that match a common pharmacophore framework Develop a sense of the significant receptor-ligand interactions Understand the different binding mechanisms Develop predictive (3D-QSAR) models
Copyright © 2006 Turquoise Consulting. All Rights Reserved5
Example: Cardiotonic Drugs
Observe the following four drugs
Copyright © 2006 Turquoise Consulting. All Rights Reserved6
Cardiotonics Pharmacophore Model
Two ring systems: aromatic (phenyl or pyridyl); 5-6 membered lactam with possible nitrogen and double bonds in the ring; with distance and planarity constraints
from: Güner, O. F. “Manual Pharmacophore Generation: Visual Pattern Recognition,” in Pharmacophore Perception and Development for Drug Design, 2000, 17-20
Copyright © 2006 Turquoise Consulting. All Rights Reserved7
Cardiotonic Hit from MDDR-3D
One of the active compounds retrieved
Copyright © 2006 Turquoise Consulting. All Rights Reserved8
Semi-Automatic ProcessesActive analog approach
Mayer, D.; Naylor, C. B.; Motoc, I.; Marshall, G. R. “A unique geometry of the active site of angiotensin-converting enzyme consistent with structure-activity studies,” J. Comput.-Aided Mol. Des. 1987, 1(1), 3-16.
Pharmacophore model from: Haraki, K. S.; Sheridan, R.P.; Venkataraghavan, R.; Dunn, D.A.; McCulloch, R. Tetrahedron Comp. Meth. 1990, 6C, 565-573
Copyright © 2006 Turquoise Consulting. All Rights Reserved9
ACE Inhibitor Hit from Derwent’s World Drug Index
Captopril retrieved with a conformation that maps onto the query
Copyright © 2006 Turquoise Consulting. All Rights Reserved10
Outline
Pharmacophores: definition Examples: Cardiotonic drugs ACE Inhibitors
Function vs. Topological Query Concept (Angiotensin II blockers)Pharmacophore model and hit list analysisPeptidomimetic design strategy (HIV-2 inhibitors , endothelin)Training set selection (PDE IV inhibitors)Model refinement - Clustering and merging (5HT3 inhibitors)Predictive models – (Antimalarial agents, FPT inhibitors)Receptor-based pharmacophores - Dealing with multiple-binding modes (DHFR inhibitors, HIV-1 Protease inhibitors)Bibliography Reviews and published success stories
Copyright © 2006 Turquoise Consulting. All Rights Reserved11
Function vs. Topological Query
Chemical function-based queries represent the 3D arrangement of the biologically relevant functions. The search is more “exhaustive” in nature It attempts to minimize “false negatives” in the hit list,
compromising the selectivity Hit list contains compounds with high diversity
Structural topology-based queries represent 3D arrangement of the functional groups The search is more “suggestive” in nature It attempts to increase the selectivity in the hit list,
minimizing “false positives” Hit list contains compounds with high topological similarity
Copyright © 2006 Turquoise Consulting. All Rights Reserved12
Structural Topology-based 3D Query
Pharmacophore model for angiotensin II blockers (developed by Erich Vorpagel via Apex-3D)
Copyright © 2006 Turquoise Consulting. All Rights Reserved13
Hit Retrieved from Derwent WDI
Zolasartan, an angiotensin antagonist
Copyright © 2006 Turquoise Consulting. All Rights Reserved14
Chemical Function-based 3D Query
Hypothesis for angiotensin II blockers by Peter Sprague (from Catalyst tutorial)
Copyright © 2006 Turquoise Consulting. All Rights Reserved15
Hit Retrieved from Derwent WDI
Zolasartan, an angiotensin antagonist
Copyright © 2006 Turquoise Consulting. All Rights Reserved16
In Derwent’s WDI, there are 448 compounds listed as “angiotensin-antagonists”
Function-based query retrieves higher percentage of known active compounds;Topology-based query retrieves hit lists with higher yield of known actives
Function- vs Topology-based Results
Query Type
No. of Hits
No. of Actives
% Yield
% Actives
GH-Score
Function Query
1,403 165 11.8 36.8 0.162
Topology Query
30 29 96.7 6.5 0.741
Copyright © 2006 Turquoise Consulting. All Rights Reserved17
Outline
Pharmacophores: definition Examples: Cardiotonic drugs ACE Inhibitors
Function vs. Topological Query Concept (Angiotensin II blockers)Pharmacophore model and hit list analysisPeptidomimetic design strategy (HIV-2 inhibitors, endothelin)Training set selection (PDE IV inhibitors)Model refinement - Clustering and merging (5HT3 inhibitors)Predictive models – (Antimalarial agents, FPT inhibitors)Receptor-based pharmacophores - Dealing with multiple-binding modes (DHFR inhibitors, HIV-1 Protease inhibitors)Bibliography Reviews and published success stories
Copyright © 2006 Turquoise Consulting. All Rights Reserved18
Database Domain
Schematic representation of a typical database and a hit list that contains some known active compounds
Database D
Actives AHa
Hits Ht
Copyright © 2006 Turquoise Consulting. All Rights Reserved19
Analyzing Hit Lists
Different metrics can be used to evaluate the quality of a hit list Enrichment (E): indicates how many time more richer the hit list is
than the original database with respect to the yield of actives
where Ht is the total number of compounds and Ha is the number of know actives in the hit list, A is the active compounds in the database, and D is the number of compounds in the database.
AH
DH
DAH
H
Et
at
a
Copyright © 2006 Turquoise Consulting. All Rights Reserved20
Analyzing Hit Lists
Selectivity (%Y) the percentage of known actives in the hit list
Coverage (%A) the percentage of known active compounds retrieved from the database
100% t
a
H
HY
100% A
HA a
Copyright © 2006 Turquoise Consulting. All Rights Reserved21
GH-Score
2
21
A
Hw
H
Hw
GH
a
t
a
t
ta
AH
HwAwHGH
221
Copyright © 2006 Turquoise Consulting. All Rights Reserved22
GH-Score
with w1 = 1.5 and w2 = 0.5, and correction for false positives
From: Güner, O. F. and Henry, D. R. “Metric for Analyzing Hit Lists and Pharmacophores,” in Pharmacophore Perception, Development, and
Use in Drug Design, IUL Biotechnology Series, 2000, La Jolla, 195-211.
2
21
A
Hw
H
Hw
GH
a
t
a
t
ta
AH
HwAwHGH
221
AD
HH
AH
HAHGH at
t
ta 14
3
Copyright © 2006 Turquoise Consulting. All Rights Reserved23
Ht = D – AHa = 0
Testing the Functions Against the Best and the Worst Hit Lists
The “Best” Hit List
D D
A AA = Ha = Ht
The “Worst” Hit List
Copyright © 2006 Turquoise Consulting. All Rights Reserved24
Functions Perform Evenly Against Typical Cases
where D = 50,000, A = 100Good hit list is 80 active compounds in a hit list of 200
Bad hit list is 100 active compounds in a list of 50,000
Case Enrichment %Y %A GH-ScoreBest 500 100 100 1Good 200 40 80 0.5Bad 25 5 50 0.2
Worst 0 0 0 0
Copyright © 2006 Turquoise Consulting. All Rights Reserved25
Against Extreme Cases?
where D = 50,000, A = 100Good hit list is 80 active compounds in a hit list of 200
Bad hit list is 100 active compounds in a list of 50,000
ExtemeY is a single hit that is active
ExtremeA retrieves all actives together with the rest of the database
Case Enrichment %Y %A GH-ScoreBest 500 100 100 1
ExtremeY 500 100 1 0.75Good 200 40 80 0.5Bad 25 5 50 0.2
ExtremeA 1 0.2 100 0Worst 0 0 0 0
Copyright © 2006 Turquoise Consulting. All Rights Reserved26
Published Successes with GH-Score Raymond, J.W. and Willett, P., “Effectiveness of graph-based and
fingerprint-based similarity measures for virtual screening of 2D chemical structure databases,” J. Comp.-Aided Molec. Des., 2002, 16(1), 59-71.
“They found that the cumulative recall and the “Goodness of Hit List” or Güner-Henry (GH) score were among the most successful of those tested for measuring the effectiveness of similarity retrieval.”
Klinger, S.; Austin, J., “Chemical Similarity Searching Using a Neural Graph Matcher,” ESANN’2005 proceedings – European Symposium on Artificial Neural Networks, Bruges (Belgium), 27-29 April 2005, 479-484.
“The position in the ranking corresponding to the maximum GH score is used as the cut-off point and subsequent structures in the list are removed from further consideration.”
Chang C.; Bahadduri, P.M.; Polli, J.E.; Swaan, P.W.; Ekins, C., “Rapid Identification of P-glycoprotein Substrates and Inhibitors,” Drug Met. and Disp. 2006, 34(12), pp 1976-1984
“Specifically, the implementation of the GH score here can be used as a determinant of the effectiveness of the model in retrieving true and false-positive …”
Cai, W. , Xu., J.; Shao, X.; Leroux, V.; Beautrait, A.; Maigret, B., “SHEF: a vHTS geometrical filter using coefficients of spherical harmonic molecular surfaces,” J. Mol. Model. 2008, 14(5), 393-401.
“The important measurement to indicate the effectiveness of the filtering is the Güner-Henry score (GH), which suggests that the performance of SHEF is …”
Copyright © 2006 Turquoise Consulting. All Rights Reserved27
Outline
Pharmacophores: definition Examples: Cardiotonic drugs ACE Inhibitors
Function vs. Topological Query Concept (Angiotensin II blockers)Pharmacophore model and hit list analysisPeptidomimetic design strategy (HIV-2 inhibitors, endothelin)Training set selection (PDE IV inhibitors)Model refinement - Clustering and merging (5HT3 inhibitors)Predictive models – (Antimalarial agents, FPT inhibitors)Receptor-based pharmacophores - Dealing with multiple-binding modes (DHFR inhibitors, HIV-1 Protease inhibitors)Bibliography Reviews and published success stories
Copyright © 2006 Turquoise Consulting. All Rights Reserved28
Vector-based 3D Searching
Objective is to replace a flexible backbone of a peptide with a rigid frame; then synthetically attach the functional groups
Lauri, G.; Bartlett, P.A. J. Comput.-Aided Molec. Des. 1994, 8, 51-66.
Example: a peptide in its bound conformation to HIV-2 protease (from Brookhaven PDB 2phv)
Copyright © 2006 Turquoise Consulting. All Rights Reserved29
Development of vector query
1 Identify the C bonds that you would like to exchange with exocyclic bonds
2 Connect the C atoms with distance constraints
3 Delete the rest of the molecule
4 Connect the C atoms with distance constraints C = can be part of an
aromatic or aliphatic ring C = any atom not on a ring
Copyright © 2006 Turquoise Consulting. All Rights Reserved30
A Hit from ACD
8-Bromoadenosine (CAS No. 2946-39-6), a commercially available chemical
The final step is to synthetically modify the compound to attach the desired functional groups to the identified exocyclic bonds
Copyright © 2006 Turquoise Consulting. All Rights Reserved31
Search for Peptidomimetic Endothelin AntagonistsCyclic peptide endothelin antagonist, and a hit
from ACD retrieved by a vector-based search
Güner, O. F.; Hempel, J. C.; Lie G. C. in The Collection of Theses, China Int. Symposium on Biotechnology and Pharmaceutical Industry, 1996, 545-547.
Copyright © 2006 Turquoise Consulting. All Rights Reserved32
Synthetic Process
A synthetic transformation is needed to attach the desired amino acid functional groups to the vector tips from the original query
Copyright © 2006 Turquoise Consulting. All Rights Reserved33
From Peptide to Peptidomimetic
Note the excellent overlay of the important functional groups at the surface
Copyright © 2006 Turquoise Consulting. All Rights Reserved34
Outline
Pharmacophores: definition Examples: Cardiotonic drugs ACE Inhibitors
Function vs. Topological Query Concept (Angiotensin II blockers)Pharmacophore model and hit list analysisPeptidomimetic design strategy (HIV-2 inhibitors, endothelin)Training set selection (PDE IV inhibitors)Model refinement - Clustering and merging (5HT3 inhibitors)Predictive models – (Antimalarial agents, FPT inhibitors)Receptor-based pharmacophores - Dealing with multiple-binding modes (DHFR inhibitors, HIV-1 Protease inhibitors)Bibliography Reviews and published success stories
Copyright © 2006 Turquoise Consulting. All Rights Reserved35
Significance of Training Set Selection
Diverse set of patented phosphodiesterase IV inhibitors selected based on cluster analysis of topological descriptors
Similar set is selected based on the compounds most similar to rolipram
Copyright © 2006 Turquoise Consulting. All Rights Reserved36
Nine Most Diverse PDE IV Inhibitors
Copyright © 2006 Turquoise Consulting. All Rights Reserved37
Nine Most Similar PDE IV Inhibitors
Copyright © 2006 Turquoise Consulting. All Rights Reserved38
Top Scoring Hypotheses
On the left is the top hypothesis obtained from the diverse training set; on the right is the one from the similar set
Copyright © 2006 Turquoise Consulting. All Rights Reserved39
Rolipram Mapped to the Hypotheses
Copyright © 2006 Turquoise Consulting. All Rights Reserved40
Diverse vs Similar Training Set Search Results
Comparison of the results obtained from searching with the top scoring hypotheses from diverse and similar training sets
from: Güner, O. F.; Waldman, M.; Hoffmann, R.; Kim, J.-H. “Strategies for Database Mining and Pharmacophore Development,” in Pharmacophore Perception and Development for Drug Design, 2000, 213-231
Query # Actives(Ha)
# Hits(Ht)
%Y %A Enrichment(E)
GH score
Database 207 10,318 2.01 100.0 1.0 0
Diverse 73 1,589 4.59 35.3 2.3 0.105Similar 51 986 5.17 24.6 2.6 0.091
Copyright © 2006 Turquoise Consulting. All Rights Reserved41
Outline
Pharmacophores: definition Examples: Cardiotonic drugs ACE Inhibitors
Function vs. Topological Query Concept (Angiotensin II blockers)Pharmacophore model and hit list analysisPeptidomimetic design strategy (HIV-2 inhibitors, endothelin)Training set selection (PDE IV inhibitors)Model refinement - Clustering and merging (5HT3 inhibitors)Predictive models – (Antimalarial agents, FPT inhibitors)Receptor-based pharmacophores - Dealing with multiple-binding modes (DHFR inhibitors, HIV-1 Protease inhibitors)Bibliography Reviews and published success stories
Copyright © 2006 Turquoise Consulting. All Rights Reserved42
Use of Query Clustering and Merging
Two strategies can be applied for query merging to improve the ratio of known active compounds in the hit list
- increase selectivity to maximize the number of active compounds in the hit list -
increase coverage
Use query clustering to identify similar and diverse models
Scenario: Hypotheses generated using a topologically diverse set of
23 5-HT3 with an activity range of 0.2 to 1,400 nm. The top hypothesis has an r correlation of 0.8275 with respect to predicted vs actual activities
Copyright © 2006 Turquoise Consulting. All Rights Reserved43
Clustering the Hypotheses
Cluster analysis results for the top ten 5-HT3 hypotheses
Hierarchical “average linkage” clustering
Number of Clusters 2 3 4 5 6 7 8 9
5HT3.1 1 1 1 1 1 1 1 15HT3.2 2 2 2 2 2 2 2 25HT3.6 2 2 2 2 2 2 2 35HT3.9 2 2 2 2 2 2 2 35HT3.5 2 2 3 3 3 3 3 45HT3.3 2 3 4 4 4 4 4 55HT3.7 2 3 4 4 4 5 5 65HT3.10 2 3 4 4 4 5 6 75HT3.4 2 3 4 5 5 6 7 85HT3.8 2 3 4 5 6 7 8 9
Copyright © 2006 Turquoise Consulting. All Rights Reserved44
Hypotheses Considered for Merging
5HT3.1 on the left and 5HT3.5 on the right
Copyright © 2006 Turquoise Consulting. All Rights Reserved45
New Merged Hypothesis
5HT3.1 and 5HT3.5 were merged by using 1.2 Å tolerance
Copyright © 2006 Turquoise Consulting. All Rights Reserved46
Merging Similar Hypotheses
HT3.6 and HT3.9 aligned before merger on the left, and following merger on the right
Copyright © 2006 Turquoise Consulting. All Rights Reserved47
Search Results with Merged Queries
Comparison of the results with 5HT3.1 and the two merged hypotheses
from: Güner, O. F.; Waldman, M.; Hoffmann, R.; Kim, J.-H. “Strategies for Database Mining and Pharmacophore Development,” in Pharmacophore Perception and Development for Drug Design, 2000, 213-231
Query # Actives(Ha)
# Hits(Ht)
%Y %A Enrichment(E)
GH score
Database 225 10,318 2.18 100.0 1.0 0
HT3.1BEST
64 1,889 3.39 28.4 1.6 0.079
Merged(1&5)
53 1,667 3.18 23.6 1.5 0.070
Merged(6&9)
174 3,772 4.61 77.3 2.1 0.147
Copyright © 2006 Turquoise Consulting. All Rights Reserved48
Merged Queries Conclusions 5HT3.1 represents the activities
of the training set well (r2=0.8275 vs for the merged[6&9] query, r2=0.6088), but not necessarily accommodate the diverse sets of active compounds in the entire database
The merged[6&9] query has a much better coverage but also surprisingly retrieved a list with improved selectivity as well
On the right, a known 5-HT3 antagonists patented by Eli Lilly is displayed. Note how well the features of the compounds maps on the the merged[6&9] query.
Copyright © 2006 Turquoise Consulting. All Rights Reserved49
Outline
Pharmacophores: definition Examples: Cardiotonic drugs ACE Inhibitors
Function vs. Topological Query Concept (Angiotensin II blockers)Pharmacophore model and hit list analysisPeptidomimetic design strategy (HIV-2 inhibitors, endothelin)Training set selection (PDE IV inhibitors)Model refinement - Clustering and merging (5HT3 inhibitors)Predictive models – (Antimalarial agents, FPT inhibitors)Receptor-based pharmacophores - Dealing with multiple-binding modes (DHFR inhibitors, HIV-1 Protease inhibitors)Bibliography Reviews and published success stories
Copyright © 2006 Turquoise Consulting. All Rights Reserved50
Use of Variable Weights in Predictive Models
Taken from the literature (M. Grigorov, et al., J. Chem. Inf. Comput. Sci., 1997, 37, 124).
Data (synthetic 1,2,4-trioxanes) artemisinin yingzhaosu
IC90 values against the Plasmodium falciparum in vitro. Ranging from 0.4 - 1184. (16 training set, 4 prediction set)
Pharmacophores developed to estimate Antimalarial Activity and to mine for new leads.
Copyright © 2006 Turquoise Consulting. All Rights Reserved51
Pharmacophore (Constant Weights)
Antimalerial agents
Three feature pharmacophore: Hydrogen Bond Acceptor 2 Hydrophobic groups
Correlation: 0.88
Each Weight=2.49
Copyright © 2006 Turquoise Consulting. All Rights Reserved52
Pharmacophore (Variable Weights)
Weight 3.15
Weight 3.15
Weight 3.15
Weight 2.36
Antimalerial agents
Four feature pharmacophore: 2 Hydrogen Bond
Acceptors 2 Hydrophobic groups.
Correlation: 0.95
Copyright © 2006 Turquoise Consulting. All Rights Reserved53
Predicting the Test Set
RActualActivity
PredictedActivity
(Cons. Wts)
PredictedActivity(Var. Wts)
H inactive 290 1200
Me 1135 260 1300
Ph 568 260 110
Bu 30 200 37
Copyright © 2006 Turquoise Consulting. All Rights Reserved54
Predictive Pharmacophore Model Development Example: Farnesyl Protein Transferase
Primary therapeutic target in cancer research
This enzyme farnesylates cysteine-186 on RAS-encoded proteins renders protein lipophilic enough to associate with cell
membranes critical step for the expression of cell transforming activity
responsible for unregulated cell grown found in several carcinomas
Two binding domains, one for farnesyl group and so-called CaaX box
Copyright © 2006 Turquoise Consulting. All Rights Reserved55
Available SAR Data
Tetrapeptide structures (2D) with associated activity data
“What compounds should I make next?”
CompoundActivity (µM)CVFM 0.01CLIM 0.08CVIM 0.15CVLM 0.2
N-AcCVIM 0.25CCVQ 0.35CKIM 0.7CGIM 3CVIA 4CVIP 5
CompoundActivity (µM)CVAM 6CPIM 7CVIL 10.5
CVGM 20CVIE 40
CVEM 70CAIL 100CVIG 100
S-AcmCVIM 1000CVKM 1000
Copyright © 2006 Turquoise Consulting. All Rights Reserved56
Initial Pharmacophore Model from Automated Process
CVFM maps all the features of the lowest cost hypothesis (activity = 10 nM, estimated activity = 32 nM)
Hydrophobic
Hydrophobic
NegativeIonizable
Acceptor
Acceptor
Copyright © 2006 Turquoise Consulting. All Rights Reserved57
Which Groups are Not Needed?
L-731,735 (IC50 = 18 nM) is as active in vitro as CVFM
Hydrophobic
Hydrophobic
NegativeIonizable
Acceptor
Acceptor
Copyright © 2006 Turquoise Consulting. All Rights Reserved58
Do We Need to Accommodate Additional Sites?
3-S diastereomer predicted more active, but distal phenyls are not described by the hypothesis model
Copyright © 2006 Turquoise Consulting. All Rights Reserved59
Or Adjust Constraints?
Doubling the tolerance permits 3AMBA-M to map all features, aligning estimated activity with experimental value
Copyright © 2006 Turquoise Consulting. All Rights Reserved60
FTP Inhibitors Research(from James Kaminski, Schering-Plough)
In vivo pharmacologic profile of “tricyclic” farnesyl transferase (FPT) inhibitors optimized via a systematic SAR study directed by chemical synthesis
N
ClBr
BrH
N
O N
O
NH2
Sch 66336 R - (+) - enantiomer in vitro FPT IC50 = 0.002 µM (2 nM)
Clinical candidate - Phase III
N
Cl
N
O
N
Sch 44342in vitro FPT IC50 = 0.25 µM (250 nM)
Copyright © 2006 Turquoise Consulting. All Rights Reserved61
Top Scoring FTP Hypothesis
RMS = 0.84
Copyright © 2006 Turquoise Consulting. All Rights Reserved62
Hypothesis Explains Observations Related to Stereochemistry
in vitro FPT IC50 (µM)
Experimental R - (+) : 0.49
Estimated (R - isomer) : 0.65N
NNO
ClN
R
N
NNO
ClN
S
in vitro FPT IC50 (µM)
Experimental S - (-) : 0.14
Estimated (S - isomer) : 0.014
Copyright © 2006 Turquoise Consulting. All Rights Reserved63
New Classes of Active Compounds Identified
Class of “azole” antifungal agents identified that exhibits in vitro FPT inhibition, IC50 5 µM
S
O
NN
Cl
N
O
OCl
S
O(CH3)3C
NN
NN
S NNCOCH3
in vitro FPT IC50 = 4.8 µM
FPT Ras / TCA IC50 = 5.3 µM
GGPT IC50 > 39 µM
Selectivity > 8
in vitro FPT IC50 = 2.5 µM
FPT Ras / TCA IC50 = 3.1 µM
GGPT IC50 > 1.4 µM
Selectivity > 0.5
Copyright © 2006 Turquoise Consulting. All Rights Reserved64
Azole Lead Mapped onto the Hypothesis
cis isomer
in vitro FPT IC50 (µM)
Experimental : 4.8(cis isomer)Estimated : 3.8(cis isomer)
S
CH3
CH3CH3
ON NN
N
SNN
O
CH3
Copyright © 2006 Turquoise Consulting. All Rights Reserved65
Selective Cis-trans Isomerism
trans isomer
in vitro FPT IC50 (µM)
Experimental : 4.8(cis isomer)
Estimated : 0.18(trans isomer)
S
CH3
CH3CH3
ON NN
N
SNN
O
CH3
Copyright © 2006 Turquoise Consulting. All Rights Reserved66
Most Active Analog
in vitro FPT IC50 (µM)
Experimental : 0.2Estimated : 0.9
S
O
N
N
ClS
Copyright © 2006 Turquoise Consulting. All Rights Reserved67
Non Selectivity in Cis-trans Isomerism
cis - isomer (±) - enantiomersin vitro FPT IC50 = 0.45 µM
S
O
NN
CH3
ON
Cl
(+) - enantiomer , []D = +30.3 ° in vitro FPT IC50 = 0.5 µM
S
O
NN
CH3
ON
Cl
(-) - enantiomer , []D = -30.1 ° in vitro FPT IC50 = 0.4 µM
S
O
NN
CH3
ON
Cl
Copyright © 2006 Turquoise Consulting. All Rights Reserved68
Explained Well with the Hypothesis
in vitro FPT IC50 = 0.5 µM in vitro FPT IC50 = 0.4 µM
S
O
N
N
CH3ON
Cl
R
R
S
O NN
CH3ON
Cl
S S
Copyright © 2006 Turquoise Consulting. All Rights Reserved69
This Work Described in Publications
Kaminski, J. J.; Rane, D. F.; Snow, M. E.; Weber, L. Rothovsky, M. L.; Anderson, S.; Lin, S.L., “Identification of novel farnesyl protein transferase inhibitors using three-dimensional database searching methods,” J. Med. Chem. 1997, 40, 4103-4112.
Kaminski, J. J.; Rane, D. F.; Rothovsky, M. L., “Database mining using pharmacophore models to discover novel structural prototypes,” in Pharmacophore Perception, Development, and Use in Drug Design, Güner, O. F. Ed., IUL Biotechnology Series, 2000, La Jolla, pp 251-267.
4-5 Years Later !!!
Copyright © 2006 Turquoise Consulting. All Rights Reserved70
Crystal Structure of FTP Became Available
FPT - and - Subunits Interact to Form a Large Active Site Cavity
Copyright © 2006 Turquoise Consulting. All Rights Reserved71
How Does the Tri-cyclic Lead Dock?
Sch 44342in vitro FPT IC50 = 0.25 µM (250 nM)
N
Cl
N
O
N
Copyright © 2006 Turquoise Consulting. All Rights Reserved72
Bound vs Predicted Conformation
X - Ray Conformation
Catalyst Conformation
Copyright © 2006 Turquoise Consulting. All Rights Reserved73
Matching the Conformations to Pharmacophore
X - Ray Conformation
Catalyst Conformation
Copyright © 2006 Turquoise Consulting. All Rights Reserved74
Functional Water Interference
Sch 44342in vitro FPT IC50 = 0.25 µM (250 nM)
Copyright © 2006 Turquoise Consulting. All Rights Reserved75
Functional Water Interaction is Critical!
Sch 44342in vitro FPT IC50 = 0.25 µM (250 nM)
in vitro FPT IC50 = 16 µM (16,000 nM)
N
Cl
N
O
N
N
Cl
NN
O
Phe 360
Copyright © 2006 Turquoise Consulting. All Rights Reserved76
How Does the Azole Lead Dock?
SO
N
N
CH3
O
N
Cl
R
R
(+) - enantiomerin vitro FPT IC50 = 0.5 µM
Copyright © 2006 Turquoise Consulting. All Rights Reserved77
Different Binding Modes!!!
N
Cl
NO
N S
O
N
N
CH3ON
Cl
R
R
Copyright © 2006 Turquoise Consulting. All Rights Reserved78
Inside the Active Site
N
Cl
N
O
N
Sch 44342in vitro FPT IC50 = 0.25 µM (250 nM)
(+) - enantiomerin vitro FPT IC50 = 0.5 µM
SO
N
N
CH3
O
N
Cl
R
R
Copyright © 2006 Turquoise Consulting. All Rights Reserved79
Outline
Pharmacophores: definition Examples: Cardiotonic drugs ACE Inhibitors
Function vs. Topological Query Concept (Angiotensin II blockers)Pharmacophore model and hit list analysisPeptidomimetic design strategy (HIV-2 inhibitors, endothelin)Training set selection (PDE IV inhibitors)Model refinement - Clustering and merging (5HT3 inhibitors)Predictive models – (Antimalarial agents, FPT inhibitors)Receptor-based pharmacophores - Dealing with multiple-binding modes (DHFR inhibitors, HIV-1 Protease inhibitors)Bibliography Reviews and published success stories
Copyright © 2006 Turquoise Consulting. All Rights Reserved80
Receptor-Based Pharmacophores
The active site of DHFR-methotrexate complex (PDB 4dfr)
Copyright © 2006 Turquoise Consulting. All Rights Reserved81
Ligand-based Query
Pharmacophore based on the bound conformation of methotrexate and the features involved in binding
Copyright © 2006 Turquoise Consulting. All Rights Reserved82
Merged Shape/Pharmacophore Query
Copyright © 2006 Turquoise Consulting. All Rights Reserved83
Shape vs Pharmacophore vs Merged Query Results
Comparison of results with shape, Pharmacophore, and merged query
from: Güner, O. F.; Waldman, M.; Hoffmann, R.; Kim, J.-H. “Strategies for Database Mining and Pharmacophore Development,” in Pharmacophore Perception and Development for Drug Design, 2000, 213-231
Query # Actives(Ha)
# Hits(Ht)
%Y %A Enrichment (E) GH score
Database 80 10,318 0.78 100.0 1.0 0
Shape 13 2,244 0.58 16.3 0.8 0.035Pharmacophore 23 1,144 2.01 28.8 2.6 0.077Merged 4 20 20.00 5.0 25.8 0.163
Copyright © 2006 Turquoise Consulting. All Rights Reserved84
Hits and Leads
On the left, folate retrieved as a false positive since it was not listed as folate antagonists
On the right is a commercially available chemical retrieved from ACD
Copyright © 2006 Turquoise Consulting. All Rights Reserved85
TGF- Inhibitor from Biogen-Idec
Using the X-ray structure of a weak inhibitor of TRI, they developed a combined shape and pharmacophore model. They then built a multi-conformational database of commercially available compounds and screened the database with the pharmacophore+shape model. They retrieved a highly potent inhibitor from this screening
200,000 compounds
87 compounds
Drug Target (eg. TRI)
N
NHN
N
IC50 27nM
Copyright © 2006 Turquoise Consulting. All Rights Reserved86
Same Compound Discovered by Eli Lilly via Experimental HTS Methods
Eli Lilly Work published at J. Med. Chem. (December 2003)
Biogen-Idec work published at Bioorg. Med. Chem. Lett. (December 2003)
The story of two simultaneous discoveries created additional press coverage Nature Rev. Drug. Disc. (December 2003) Bio-It World (February 2004)
Finally Biogen and Lilly scientists jointly publish the story Singh et el., Curr. Opin. Drug Disc. Dev. 2004, 7(4), 437-445
Copyright © 2006 Turquoise Consulting. All Rights Reserved87
Outline
Pharmacophores: definition Examples: Cardiotonic drugs ACE Inhibitors
Function vs. Topological Query Concept (Angiotensin II blockers)Pharmacophore model and hit list analysisPeptidomimetic design strategy (HIV-2 inhibitors, endothelin)Training set selection (PDE IV inhibitors)Model refinement - Clustering and merging (5HT3 inhibitors)Predictive models – (Antimalarial agents, FPT inhibitors)Receptor-based pharmacophores - Dealing with multiple-binding modes (DHFR inhibitors, HIV-1 Protease inhibitors)Bibliography Reviews and published success stories
Copyright © 2006 Turquoise Consulting. All Rights Reserved88
Pharmacophore Review Articles
Gund, P. “Three-dimensional Pharmacophoric Pattern Searching,” in Progress in Molecular and Subcellular Biology, vol. 5, Hahn, F. E. Ed.; Springer-Verlag, 1979, Berlin, pp. 117-143.
Gund, P. “Pharmacophoric Pattern Searching and Receptor Mapping,” Ann. Reports Med. Chem. 1979, 14, 299-308.
Humblet, C. and Marshall, G. R. “Pharmacophore Identification of Receptor Mapping,” Ann. Reports Med. Chem. 1980, 15, 267-276.
Kurogi, Y. and Güner, O. F. “Pharmacophore Modeling and Three-dimensional Database Searching for Drug Design Using Catalyst,” Curr. Med. Chem. 2001, 8, 1035-1055.
Güner, O. F. “History and Evolution of the Pharmacophore Concept in Computer-Aided Drug Design,” Curr. Top. Med. Chem., 2002, 2, 1321-1332.
Copyright © 2006 Turquoise Consulting. All Rights Reserved89
Published Successes with Pharmacophore-based 3D Searching
"… this hitlist identified five compounds representing three structurally novel classes, that exhibited in vitro FTP inhibitory activity (IC50 5 M)."
J. J. Kaminski, et al. J. Med. Chem., 1997, 40, 4103
"… the excluded volumes reduced the size of the hit list… The single remaining compound was subsequently shown to bind to THR-a with an IC50 of 69 M."
P. A. Greenidge, et al. J. Med. Chem. 1998, 41, 2503.
Pharmacophores selectively separating sub-type antagonism Bremner, J.B. et al., Bioorg. Med. Chem. 2000, 8, 201-214
Predictive metabolism with pharmacophores Ekins S. et al., J. Pharm. Exp. Therap., 1999, 290(1), 429-438
Predictive model from diverse structures Tronchet, J.M.J. et al., Eur. J. Med. Chem., 1997, 32, 279-299
Common pharmacophore for two different classes Liao, N. et al., Chin. Chem. Lett., 1999, 10(0), 755-758
Dynamic receptor-based pharmacophores Carlson, H.A. et al., J. Med. Chem. 2000, 43, 2100-2114
Copyright © 2006 Turquoise Consulting. All Rights Reserved90
Hypoglycemic Agents
Glu, Ins and TG in Dex-induced rat at 100 mg/kg poY.Kurogi, J.Synth.Org.Chem., 2000, 676.Y.Kurogi, O.Güner, Curr. Med. Chem., 2001, 1035-1055.
One Hit (1/6)Novel CompoundLow MW (274)
OT-5226OT-5226Glu: 47 %Ins : 37 %TG : 34 %
RF05274RF05274Glu: 41 %Ins : 11 %TG : 41 %
CatalystTM
N
N
Cl
O
P
O
OO
NOF3C
NH
O O
Copyright © 2006 Turquoise Consulting. All Rights Reserved91
MC Proliferation Inhibitors (MCPI)
High hit ratio (100 %)Structural diversityNovel MCPIs
OP
NH
N O
Cl
O
O O
N
O
HN
HN
OO
O
O
Cl
Cl
Cl
CatalystTM
MCPI = 69 %a MCPI = 90 %a
a% inhibition at 100 nMMCPI:Mesangial Cell Proliferation InhibitionNCPI:Normal Cell Proliferation Inhibition
Increases Safety
NCPI = 30 %a NCPI = 0 %a
Y.Kurogi, et. al., J. Med. Chem., 2001: 44, 2304-2307
Copyright © 2006 Turquoise Consulting. All Rights Reserved92
Novel Fragrance Development through Olfactophores
Looking for “odoriferous compounds,” sense of smell via olfactory lobes olfactophores Scientists at Givaudan Roure have used Catalyst to design
fragrances
O
O
OR
R'Ambrox … to … New “ambery” odorant
Baigrowitz et al., Enantiomer, 2000, 5, 225-
234
Kraft, P. et al., Angew. Chem. Int. Ed. Engl.
2000, 39, 2980-3010
Copyright © 2006 Turquoise Consulting. All Rights Reserved93
Pharmacophores in Patents
WO 98/04913 Filed in 24 July 1997 by Juswinger Singh and co-workers,
this was an application by Biogen for pharmacophore model for VLA-4 inhibitors
WO 98/46630 Filed in 16 April 1998 by Terance Hart and co-workers, this
was an application by Peptide Therapeutics Limited for pharmacophore model for Hepatitis C NS3 Protease Inhibitors.
US 2002/0013372 Filed on March 12, 2001 by Sean Ekins, this is an
application by Pfizer Inc. for identification of CYP2D6 inhibitors.
Copyright © 2006 Turquoise Consulting. All Rights Reserved94
Acknowledgments
James Kaminski – Schering-Plough
Luke Fisher - Accelrys