the role of chemoinformatics in the design of a...

The Role of Chemoinformatics in the Design of a Comprehensive Drug Discovery Screening Collection

Edgar Jacoby

Center for Proteomic Chemistry

2nd Strasbourg Summer School on ChemoinformaticsVVF Obernai, France, 20-24 June 2010

| Obernai| 20.06.10 | Edgar Jacoby

NIBR compound collection enhancement project

External compound selection process

Combinatorial library synthesis – scaffold projects

Compilation of annotated knowledge-based sets

Molecular Information System for pharmaceutical ligands

Knowledge-based ligand design and virtual screening strategies

Chemoinformatics: Quo Vadis?

Overview


Essential Properties of Compounds


Essential Properties of Compounds - Continued

Jacoby et al. Chemogenomics 2007, 1-38


LDCStructuresDBOracle and ISIS databases in place

10.000.000+ Unique structures

13.000.000+ Catalogue entries

130+ Collections

50+ Vendors

LDCScaffoldDB Oracle and ISIS databases

18.000+ Unique scaffolds

15 Vendors

Externally Available Chemistry Space


Chemistry, Biology, Informatics Based Compound Selection

2. Completeselection by with compounds from container 2:(average structures)average density, similarity <= 0.88

descriptor space

3. Fi ll gaps with compounds from container 3 (penalized structures):lower density, similarity <= 0.80

descriptor space

0. The existing archive

descriptor space

1. Select from container 1:(target family related affinity)high density, similarity <= 0.95

descriptor space

2. Completeselection by with compounds from container 2:(average structures)average density, similarity <= 0.88

descriptor space

3. Fi ll gaps with compounds from container 3 (penalized structures):lower density, similarity <= 0.80

descriptor space

0. The existing archive

descriptor space

1. Select from container 1:(target family related affinity)high density, similarity <= 0.95

descriptor spaceSimilar to Know MDDRActives and Drugs

PenalizedDiverse Drug Like

Jacoby et al. Curr Top Med Chem. 2005, 5, 397-411


ID Structure Chemical Description

33 A

N

O carbonic acid nitriles Y Y N

34 A

OPPOO

* OPSOO

**OSSOO

* *

PO

OO

**S

O

OO

**O

OO

* *

O PO

O PO

OOO

not:

acyclic acid anhydrides including phoshonates and phosphinates, but not di- and triphospates

Y Y N

35 B

O PO

O PO

OOO

PO

OO

**

OPPOO

*

SO

OO

**

OPSOO

**

O

OO

* *

OSSOO

* *

not:

cyclic acid anhydrides including phoshonates and phosphinates, but not di- and triphospates

Y Y N

36 C O

NO

H

H acyclic alkyl carbamates Y N N

Definition of the Exclusion-Criteria:A: Don't store in MCSB: Don't produce as solution or store in SOLARC: Don' buy (but take for free)

Toxic ReactiveUnintersting

Archive Exclusion FlagUninteresting

Substructure/Fragment Filters


In Silico Substructure Tox Alerts


Y

Confirm availability and price

N

Sample/catalogue entry domain Structure domain

registered in corporate structure DB

not registered in corporate structure DB

Compound received?N

Register compounds (registered)

Y

Catalogue entry marked as selected NY

Div

ersi

ty

Order placed (ordered)

Decide to order

Don’t order (cancelled)

NY

Load Catalogue, add new structures (new), check availability for all compounds

Compound is not available

YN

Catalogue entry marked as rejected

Y

Identity by structure (smiles)

Structures processed in parallel acquisition processes

Structures of “archivecompounds”

N

Export new structures for evaluation (exported) Structure based evaluation of compoundsDrug-Likeness and Target Focus?Diversity?

Y

Confirm availability and price

N

Sample/catalogue entry domain Structure domain

registered in corporate structure DB

not registered in corporate structure DB

Compound received?N

Register compounds (registered)

Y

Catalogue entry marked as selected NY

Div

ersi

ty

Order placed (ordered)

Decide to order

Don’t order (cancelled)

NY

Load Catalogue, add new structures (new), check availability for all compounds

Compound is not available

YN

Catalogue entry marked as rejected

Y

Identity by structure (smiles)

Structures processed in parallel acquisition processes

Structures of “archivecompounds”

N

Export new structures for evaluation (exported) Structure based evaluation of compoundsDrug-Likeness and Target Focus?Diversity?

Cheminformatics Platform for Selection and Logistics

Schuffenhauer, A. et al. Comb. Chem. High Throughput Screen. (2004), 7, 771-782.


Schuffenhauer et al. J Chem Inf Model. 2007, 47, 47-58.

Scaffold Tree Example for HTS ResultsPubChem Pyruvate Kinase Data Set

Color intensity by fraction of actives


Library Evaluation

Pat

enta

bilit

y

Nov

elty

of S

caffo

ld

Nov

elty

of P

rodu

cts

Mod

ular

ity

Pla

nned

Siz

e

Rat

io o

f Low

est a

nd

Hig

hest

Nr.

of B

Bs

Com

plem

enta

rity

Rul

e of

Fiv

e

InS

ilico

Tox

BB

Ava

ilabi

lity

BB

Div

ersi

ty

Chi

ral P

urity

HT

Sol

ubili

ty

HT

logP

HT

Per

mea

bilit

y

CY

P 4

50 (M

etab

olis

m

HT

HE

RG

cha

nnal

as

say

Am

es te

st

Eas

e of

Rea

lisat

ion

Tim

e of

Dev

elop

men

t

Che

mic

al S

tabi

lity

1 1 1 0 0 -1 1 1 1 1 1 1 0 0 0 0 0 -1 0 1 1

Prototype CriteriaPreprototype Criteria

-1

0

1

Evaluation Scheme for Combinatorial Hit Discovery Libraries

noveltytargetfocus

syntheticcomplexity

potentialfor HLO

computedproperties

2004

2007

Jacoby et al. Curr Top Med Chem. 2005, 5, 397-411


The Ideal Size of Combinatorial Libraries


AlogP Distribution

0.0%

2.0%

4.0%

6.0%

8.0%

10.0%

12.0%

-5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10

AlogP

% o

f sam

ples CHC [%]

SOLAR [%]CMedC [%]

Molecular Weight Distribution

0.0%

1.0%

2.0%

3.0%

4.0%

5.0%

6.0%

7.0%

8.0%

9.0%

10.0%

0 100 200 300 400 500 600 700 800 900 1000

Molecular Weight

% o

f sam

ples

CHC [%]SOLAR [%]CMedC [%]

H-Bond Donors

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

30.0%

35.0%

40.0%

0 1 2 3 4 5 6 7 8 9 10

Count

% o

f sam

ples


H-Bond Acceptors

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Count

% o

f sam

ples


Monitoring of Compound Collections - Lipinski Profiles


0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

0 2 4 6 8 10

AlogP

-logS

n (ex

p, e

quili

briu

m)

logS vs AlogP, no saltlogS vs AlogP, saltGSE Limit -logS >= logP + 0.5Limit -logS >= logP

Prediction of Water Solubility

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

9.0

10.0

0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

-logSpH6.8(QMPRPlus, Corina 3D structure)

-logS

pH6.

8(ex

p,eq

uilib

rium

)

inside model rangeoutside model rangesaltsexact-1 log unit+1 log unit

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10-logSn (QikProp, Corina 3D structures)

-logS

n (ex

p, e

quili

briu

m)

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10

-logSpH6.8(calculated,ACD)

-logS

pH6.

8(ex

p, e

quili

briu

m)

non_saltsaltexact+1 log unit-1 log unit

14

126127

128129

130131

2132

136133

134

135137

138139

76

10810990

140

112142

141

142145143

84147

144148

85149

118150

146

151

152154153


The “Egan Egg”: Statistical Prospectively Useful Model

-20 0 20 40 60 80 100 120 140 160 180 200

-2

0

2

4

6

8

HigherPermeability

HigherSolubility

Poor Solubility

Poor Permeability& Poor Solubility Few marketed drugs lie

in this region

Poor Permeability Drugs in this region usually work via prodrugs of active transport

Good Hunting Grounds

Clo

gP

PSA WJ Egan et al, JMC, 2000

Access Egan Egg via IsisDrawInSilico Profile T2ABS plot


Chemist Triage Score and Naïve Bayesian Models

Yes/No decision from five individual triaging exercises taken to build one Bayesian models

All five models are used in a consensus score, fraction of models voting against the molecules

Chemist Triage Score

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

MCS SLIA ICLF PSEB MCSexternal

2003

MCSexternal

2004

Drugs NAPR HTS Hits2004

Random

Frac

tion

of c

ompo

unds

Series6Series5Series4Series3Series2Series1

1

0.8

0.6

0.4

0.2

0

ChemistTriageScore

Chemist Triage Score

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

MCS SLIA ICLF PSEB MCSexternal

2003

MCSexternal

2004

Drugs NAPR HTS Hits2004

Random

Frac

tion

of c

ompo

unds

Series6Series5Series4Series3Series2Series1

1

0.8

0.6

0.4

0.2

0

ChemistTriageScore

Histogram of NB Score

0%

2%

4%

6%

8%

10%

12%

14%

16%

-100 -80 -60 -40 -20 0 20 40 60 80 100%

of s

ampl

es

Training Set

Screening Collection

Primary Hit-list

Less attractive cpds.


Examples of Unwanted Molecules

NH2

NH

O

NH

O

NH

O

N+N

H

O

N+

O OH

NH

Cl

Cl

Cl

Cl N

O

Br

F

F

F

N

N+

O

O

1.0

0.8 0.8

S

OHOH

OH

O

O

F FF

ClCl

NHN

ON

OH

OH

O

0.6 0.6

0.4

OH

NH

OHNH

OH

OH

NH

O

OHO

OH

F

N

O

NON

S

0.0012

ChemistTriageScore: Excluded

ChemistTriageScore: Penalized

Similog density: Excluded

Similog density: Penalized

0.0017

0.0029

0.0036


The Challenge of Quality in Candidate Optimization

Biller S. et al. Biotechnology: Pharmaceutical Aspects, 2004, 413-429.


Safety Pharmacology Prediction


Which Annotated Compound Libraries Make Sense ?

1. Approved drugs – FDA orange book - SOSA

2. Other reference compounds (USAN, INN) with known bioactivity (Tocris, Prestwick, etc.)

3. Compounds processed by profiling

4. Primary metabolites, natural products and derivatives (KEGG, etc.)

5. Target family focused libraries

6. Diversity selection – DOS / BIOS

7. Protein mimetics library: β-turn/α-helix mimetics

8. Fragment library for NMR/Xray screening


Wess DDT 2002, 7, 533-535

Focusing : Optimizing the Overlap Between Chemical and Biological Spaces


Annotation ?No

YesKnownStructure ?

NoYes

Is binding site ‘Drug

like’ ?

No

Yes

Ligand-based

Yes No

Similaritybased list

Known Actives ?

Yes Select compounds

similar to known actives

Are Actives

‘Drug like’ ?

Yes

No

Structure-based

TARGET Select

compounds compatible to binding site

Structure based list

Diversity screening

Knowledge-Based Virtual Screening


Molecular Information System for Pharmaceutical Ligands of the Main Target Families

Target Compound Selections MIS

Ligand -Target Sequence

Ligand -Target Classification

Ligand - Interaction Classification

Ligand - Functionality

Reference Ligands

LigandsNovartisMDDR,...

TargetsSwiss-Prot

GPCRDB,...

Schuffenhauer, A. and Jacoby, E. BioSilico ( 2004), 2, 190-200


Databases of Known Pharmaceutical Ligands and Drugs

Lit & PatentBiologyIdentificationModelStructure

Structure

NN

NNN

O

O

O

Chiral185987

Generic nameValsartan < Rec INN; USAN >

Extreg

Preview

TrademarkValpression (Menarini, IT)Provas (Schwarz, DE)Varexan (Egis, HU)Di (N ti CA DE

185987Pref#

Launched

Act.invest?

SourceNovartis

Company.codeCGP-48933CGP-63170 (comb. with

MDDR-3D 2001.1(23.01)

Phase

Index Activity28000 Cardiotonic31000 Antihypertensive31432 Angiotensin II

AT1 Antagonist


Knowledge Based Approaches : GPCR IMDL Patent Database

Activity Based Search

38 Patent Families

1810 Reference Structures

Similarity Search Naïve Bayes Model

542 SOLAR in silico hits


Similarity Searching using Similog Keys

1NN 3NN avg cent

P. Floersheim’s Similog keyscounts of DABE triplets in 2D space

O

OO

H

0010-4-1100-6-0100-6-

4

6

6

0010

0100

1100

Schuffenhauer, A. et al. J. Chem. Inf. Comp. Sci. ( 2003), 2, 391-405.


NH

N

Cl

N

N

O

SNH

O O

N

Cl

O

O

N

NH

N

NN

Cl

N

OO

OH

HCl

NH

N

N

O

SNH

O O

N

Cl

NNH

SNH

N

OO

Cl

ON

O

NH

N O

FO

OH

N

OOH

NH

OHN

OH

NH2

O

NH

N

NS

NH

N

N

ClCl

NH

NH2

OH

N

N

N

O

N

O

Propranololβ antago.

8-OH-DPAT5HT1A part. ago.

Serotonin5HT ago.

Ketanserin5HT2A antago.

Janssen_1D4 antago.

Kissei_1D2 antago.

RO-16814β ago.

5HT7 antago.pKB = 7.25

Known Reference Architectures Novel Prototypes

WAY-1006355HT1A antago.

Design of TAM Libraries for Monoamine-Related GPCRs

Jacoby, E. Quant. Struct. Activity. Relations. (2001) 20, 115-123.


CAS DB

Deconvolute

Recompose

Databases of Active Fragments

NOVARTIS DB

New Leads + New Affinity Profiles

Deconvolute and recombine : Drug like molecules

Currently applied knowledge-based strategy

Jacoby et al. Drug News Perspect. (2003),16, 93-102.


Ligand selectivity in a subfamily with conserved molecular recognition

Do we understand and master the principles of profile composition in terms of molecular recognition ?

α2A α2B α2C β1 β2 D2.L D3 D4.2 5HT1A 5HT1D 5HT1E 5HT1F 5HT2A H1

Sumatriptan < 6 < 6 < 6 < 6 < 6 < 6 < 6 < 6 6.4 8.4 < 6 7.8 < 6

Alniditan 7.4 6.8 7.7 < 6 < 6 < 6 6.5 7.1 8.4 9.4 6.6 6.4 < 6 < 6

Haloperidol < 6 6 .3 6.2 < 6 < 6 8.7 8.1 7.9 < 6 6.6 < 6 < 6 6.7 6.1

Pipamperone 6.1 7.5 6.5 < 6 < 6 6.9 6.6 8.3 < 6 6.8 < 6 < 6 8.3 < 6

Clozapine 7.3 7.7 8.1 < 6 < 6 6.7 6.6 7.4 6.9 6.4 6.4 6.9 8.0 9.6

Olanzapine 6.3 6.7 6.7 < 6 < 6 7.5 7.3 7.6 < 6 6.2 < 6 6.5 8.6 9.2

pKi


0

1

2

3

4

5

60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9 1

Maximum Tanimoto Index

% C

ompo

unds

Focus of TAM library – Similarity Histograms

Monoamine GPCRs

Peptide-binding GPCRs

LGIC

R1N

R1

R2

'

Okram et al. Chemistry & Biology 2006, 13, 779–786

A general strategy for creating “inactive” conformation kinase inhibitors - ABL



0%

20%

40%

60%

80%

100%

BCG-001

BET-001

BET-02A

DIA-002

IMO-00

1

AAL-001

BET-02B

HOP-002

HOP-003

DIA-001

PUR-003

SYK-001

AAL-002

BZI-001

GUA-003

IMI-2

51

OXI-001

SRC-001

TAM-003

IAB-451

IPO-825

PUR-001

TAM-001

TAM-002

TSA-003

TAM-004

TAM-005

TSA-001

GPCRKinaseProteaseother EnzymesNURion channelPPIfunctional

Is Target Family Focus a Myth ?

Selectivity Analysis of Combinatorial Libraries


Extend Knowledge-based Approaches

Target family based approaches

BIOS – Protein domain based designs

Privileged Scaffolds

Protein Secondary Structure Mimetics: α-helix/β-turn

Co-factor mimetics: ATP, NADP, FAD, Co-A, SAM, etc.

Privileged structures / Scaffold target matrix

NH

O

N

ON

O

N

NH

O

N

N

Cl

N

N

O

O

N

N

N

O

NO

N

OOH

O

NN

N NH

NO

NH

O

NH2

NN

N NH

NH

NH

N

O

O

O NH

O

NH

O

NH

OH

RS504393 α1A / CCR2b antagonist

Diovan AT1 antagonist

Merck_2 GHr agonist

Meiji Seko µ1 agonist / D2 antagonist

Merck_1 NK1 antagonist

Pfizer_1 CCK antagonist

Naftopidil α1 antagonist

Servier_1 NK1 antagonist

5HT1

A5H

T1B

5HT1

D5H

T1F

5HT2

A5H

T2B

5HT2

C5H

T4A

1AA

2AB

1B

3D

1D

2D

3D

4M

1M

3H

2Fr

eque

ncy

PS1 0 0 0 0 2 0 0 0 0 4 0 0 1 11 0 0 0 0 0 4PS2 0 0 0 0 0 1 1 0 0 0 0 0 0 2 0 14 0 0 0 4PS3 6 0 0 0 3 3 3 0 0 0 0 0 0 0 0 0 0 0 0 4PS4 2 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 4PS5 2 0 0 0 0 0 0 0 1 0 0 0 0 2 1 0 0 0 0 4PS6 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 3 0 0 0 4PS7 2 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 0 0 0 4

N

N

N

N

N

N

NH

NH

N

NH

OO

O

NH

O

NH N

PS1

PS2

PS4

PS3

PS5

PS6

PS7


Privileged structures GPCRsThe 2-aryl-indole case


Co-factorsBioactive Motifs with Conserved Binding Sites in Enzymes


Ligands vs. domains and folds that bind themAnalysis of PDB – Cofactor Binding Sites Open Large Potential Target Space

Distribution patterns of small-molecule ligands in the protein universe and implications for origin of life and drug discoveryHong-Fang Ji , et al.Genome Biology 2007, 8:R176doi:10.1186/gb-2007-8-8-r176

Ligands Full name of ligands Number of

domains Number of

folds

PLP Pyridoxal-5'-phosphate 17 10 TDP ThiaminDiphosphate 36 10 SAM S-adenosylmethionine 41 13 COA CoenzymeA 29 14 GTP Guanosine-5'-triphosphate 26 15 AMP AdenosineMonophosphate 29 15 NAP NadpNicotinamide-adenine-dinucleotidePhosphate 73 16 FMN FlavinMononucleotide 122 16 NDP NadphDihydro-nicotinamide-adenine-dinucleotidePhosphate 44 18 ANP PhosphoaminophosphonicAcid-adenylateEster 59 19 FAD Flavin-adenineDinucleotide 152 21 NAD Nicotinamide-adenine-dinucleotide 149 27 GDP Guanosine-5'-diphosphate 86 29 ADP Adenosine-5'-diphosphate 137 31 ATP Adenosine-5'-triphosphate 97 35


Sam: S-AdenosylmethionineJack of all Trades and Master of Everything?

Loenen WA.Biochem Soc Trans. 2006 Apr;34(Pt 2):330-3

NH2

O

OH

O

OH

OHN

NN

N NH2

S+

Conformational Analysis of 252 SAM PDBs3 main Clusters – Pharmacophorfamilies (J. Priestle MLI)



MDM4

HIT

SET

46511

HTD Homology model based on SPIRO-OXINDOLE and NUTLIN MDM2 structures Constrained SP Glide docking Selection of top 20000 Compounds

Compounds From In-house MDM2 Assays and Similars Hit-finding and Hit-to-lead assays - compounds with IC50 < 50 µM (650) FCFP4 similarity search (1873) Hopfen QSAR (5000)

Literature and Patent Mining WOMBAT Literature Data and 14 Patent families from MDL Patent Database FCFP4 similarity search (675) and NB model (5000) FEPOPS (1000) similarity search, Substructure search (1707)

UNITY Searching Homology model based on SPIRO-OXINDOLE and NUTLIN MDM2 structures Constrained search: Hydrophobic centers and HBonds (49305) Constrained SP Glide docking Selection of top 20000 Compounds

Knowledge-Based Virtual Screening – MDM4/2


FEPOPS 3D similarity searching

4 feature centroids and distances

Jenkins et al. J. Med. Chem. (2004), 47, 6144-6159.

2 4 2 4

Balanol PKC inhibitorATP

Balanol low 2D similarity to ATP

High similarity in FEPOPS as in Xray

P

O

OH

OH

NO

N

N

OH OH

O

N

NH2

P

OH

O O

PO

O

OH

NH

NH

O

O

O

OH

OH

O

OH

OO

OH

Computational Procedure

1. Generate conformers

2. Calculate FPs and properties

3. Compute fingerprints

4. Search based on fingerprints

HNH

H

O NH

N

O

H

ClF

O

NH

Cl

ON

H

H

Cl

N

ON

NH

O

Cl

O

Pharmacophore Screening – MDM4/2



Knowledge-Based Virtual Screening – MDM4/2

Comparing HTS and Virtual Screening – MDM4/2

Virtual Screen HTS


Comparing HTS and virtual screening – MDM4/2


Chemoinformatics: Quo vadis?Towards prediction paradise?

Van de Waterbeemd, H. et al. Nat Rev Drug Discov. 2003,192-204


Field of Informatics

Primary Data

Similarity Relationships

Key Applications

Bio- Protein

Sequences Sequence Alignments Function Inference

Protein Structures Structure Alignments Function Inference

Binding Sites Site Alignments

Target Hopping, Cross Reactivity,

Selectivity, Opportunity Mining

Structural

Sites+Ligands Binding Mode Alignments

Enhanced Screening, Scaffold Hopping, Novel Scaffolds

Chem- Small Molecules Molecular Similarities Screening

Structure Determination

Small Molecule Docking

Site Annotation

Chemoinformatics: Quo vadis?Molecular Informatics

Courtesy of Dr. D. Debe, Eidogen-Sertanty


Chemoinformatics: Quo vadis?Chemical Systems Biology – Compound-pathway maps



Acknowledgements

Ansgar Schuffenhauer

Maxim Popov

Kamal Azzaoui

Joerg Muehlbacher

Jeremy Jenkins

Meir Glick

Bill Egan

Peter Ertl

Bernard Faller

Hans-Joerg Roth

Peter Fuerst

Recent Publications


the role of chemoinformatics in the design of a...

Documents