a presentation by Dr David Lloyd Trinity College Dublin ...Compound Number IC50 in MCF-7 MTT MDG-ER-001 8.23E-07 MDG-ER-002 8.00E-06 MDG-ER-003 2.02E-05 MDG-ER-004 5.59E-04 MDG-ER-005

Focused Virtual Screening

a presentation by Dr David Lloyd

Trinity College DublinDaylight EuroMUG 2004

Lead discovery in the Human Estrogen Receptor

1592

750 AD

Biochemistry in TCD – largest Department in Country

Significant research output

Centre for Structural Biology and Molecular Design

building on PRTLI and SFI investment in structural biologybuilding on PRTLI and SFI investment in structural biology

Molecular Design Group

Established 2004

Ireland’s first protein X-ray facility

Integrated Drug Discovery

Chemistry Biology

Computation

Structure Based Design – looking in the ER


Curr Med Chem 2003, Frontiers Med Chem 2005 (in press)

Significance of ER

• Estrogens regulate cell growth, differentiation & development of reproductive tissues in men and women.

• Maintain bone density preventing osteoporosis.

• Exerts anti-atherosclerotic effects which lowers Cholesterol levels.

• Involved in many CNS effects (Parkinson's) and implicated in Alzheimer's.

ER as a target

• 60% of primary breast cancers contain ER- alpha

• Estrogens are mitogenic for ER-positive breast cancer cells.

Target: Target: EstrogenEstrogen Receptor Receptor ��

Structure Based Design – Docking in Nuclear Receptors

Docking AlgorithmsDocking Algorithms Scoring FunctionsScoring Functions

Structure Based Design – Docking in Nuclear Receptors

FREDFRED

FlexXFlexX

Discover3Discover3

eHitseHits

In houseIn house

PLPPLP

ChemScoreChemScore

CScoreCScore

PLCPLC


Building on knowns : using receptor structural knowledge– semi-rational design

Traditional scaffold hopping –human de novo rational design

J Med Chem2001

Anti Cancer Drug Design

2001

Traditional scaffold hopping –human de novo rational designComputer-enhanced!!

Benzoxepin antiestrogens

N

aminoalkylation

aminoalkylation

BBr3

PhZnCl

Pd(PPh3)4

boronic acids

Pyr.HCl

PyHBr3H+n-BuLi

OMe

Br

1110

987

para CN (19)para Me (15)

meta Me (16)para Cl (17)meta NO2 (18)

para OMe (12)ortho OMe (13)meta OMe (14)R2 =

21-2820

R1=

NN

N

O

N

6 7-115

Suzuki Route

Heck Route

4

3

R2

O

O

N

O

R1O

O

HO

Br

O

MeO

O

HO

R2

2

O

HO

O

MeO

Br

O

MeO

O

HO

OMe

O

O

Computer-enhanced human de novo rational design

Ortho-ring substitutionis tolerated - meta is not -

elcectic binding mode

Computer-enhanced human de novo rational design

J Med Chem2004

Haystack built from 880 ‘drug-like’ compounds from WDI

�40 Cox-2 inhibitors�40 Estrogen Receptor Modulators �40 Histamine ‘modulators’

Active ‘needles’ introduced from a separate validated ligand set

Let the computer decide : Virtual Screening

Virtual Screening

vHTS – Performance Measures – Validation

Enrichment = hit rate observed in subsethit rate in database (random)

Enrichment Subset Size (%)

1 5 10 15 20

Ligands

10

50

100

150

200

Max Actives

10

40

40

40

40

Best Possible

Value

25

20

10

6.7

5.0

e.g. 1% sampled = 10 compounds.Subset - 10 actives = hit rate of 10/10 = 1.0, Hitrate in database is 40/1000= 0.04 : enrichment = 1 / 0.04 = 25

Target Database

Docking Protocol

Rescoring

Generation of Hits

Active set of compounds for development

-Remove waters & Calculate centre of bound ligand.-Use multiple structures

PreProcessing

• Samples search space and generates a set of binding poses for each ligand conformer.

• Docked positions have their respective hydrogen bond lengths optimized to allow for refinement of the final structure.

• CF (Complementarity Function) evaluates fit

• Ranks these modes/ligand positions

• Provides a numerical score allowing for ‘hit’identification

In-house docking protocol

Actual Hits Retrieved

0

5

10

15

20

25

30

35

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

% Sample Database

% H

its R

etri

eved

F_Score

G_Score

PMF_Score

D_Score

ChemScore

Xscore

PLP_Score

Fresno

Screenscore

Hammerhead

Theoretical Maximum HitsRetreived

Chemscoreperforms best of scoring functions.

Accounts for:Hydrogen bond contacts, Lipophiliccontacts, entropic penalty.

G-score focuses on hydrogen bonding interactions only for example.

Getting it right – Scoring Functions

-38.208COX20080-94.16(Drug_401)

-39.083ESTR0067-94.24(Drug_472)

-39.112Drug_61-94.43(Drug_751)

-39.176Drug_259-94.5(Drug_476)

-39.624Drug_421-94.53(Drug_466)

-39.874Drug_265-95.72(ESTR0085)

-40.123ESTR0068-96.07(ESTR0024)

-40.154Drug_315-96.18(ESTR0045)

-40.389Drug_257-96.33(Drug_161)

-40.687Drug_217-96.64(ESTR0043)

-40.991Drug_249-97.1(Drug_163)

-41.031Drug_416-100.14(ESTR0034)

-41.635Drug_823-101.03(ESTR0046)

-41.647Drug_219-102.62(Drug_474)

-42.169Drug_440-103.16(Drug_160)

-42.209Drug_353-103.35(Drug_633)

-42.485Drug_344-103.42(Drug_159)

-44.427ESTR0072-104.85(Drug_158)

-46.568ESTR0079-107.73(ESTR0025)

FREDCHEMSCOREName_IDSybyl6.91CHEMSCOREName_ID

Getting it right – Scoring Functions

Hit Retrieval

0

5

10

15

20

25

30

35

40

1 7 13 19 25 31 37 43 49 55 61 67 73 79 85 91 97

% Database Screened

% H

it R

etri

eved

In House Protocol

FlexX

FRED

Best Value

Getting it right – early method validation

Getting it right – Ligand Pre Processing

Virtual High-Throughput Screening

<1 sec per compound – rigid/rigid system

CorinaCorina OmegaOmega StergenStergenRubiconRubicon QuacPacQuacPac

0.4740_RUBICON_LEVEL1

0.5740_CATALYST_LEVEL1

0.4840_OMEGA_LEVEL1

0.5440_CORINA_LEVEL1

X



0.6640_OMEGA_LEVEL2


Quac-X



0.7140_OMEGA_LEVEL3


Quac-X-10 Confs



0.7040_OMEGA_LEVEL4


Quac-Ster-X



0.6940_OMEGA_LEVEL5


Quac-Ster-X-10 Confs


Random screening – 40 actives in 1000 –each active returns a score – the bigger the difference between the active and inactive scores, the better the method

Preprocessing can increase the cutoff value for ligand consideration – reducing the subset we must consider in order to find our active ligands.


6.58.7512.516.2520LEVEL1_RUBICON

67.59.1712.517.5LEVEL1_CATALYST

78.751013.7522.5LEVEL1_OMEGA

1011.881516.2522.5LEVEL1_CORINA

Avgenrichment4321Subset %

11.513.1314.213.7510LEVEL2_RUBICON

55.6256.668.7515LEVEL2_CATALYST

9.511.2514.218.7522.5LEVEL2_OMEGA

7.511.2518.32025LEVEL2_CORINA

910.62514.1618.7522.5LEVEL3_RUBICON

910.62513.332025LEVEL3_CATALYST

9.511.251521.2525LEVEL3_OMEGA

9.511.8751521.2525LEVEL3_CORINA

1113.7514.1718.7522.5LEVEL4_RUBICON

6.56.8758.331017.5LEVEL4_CATALYST

9.510.62513.3316.2520LEVEL4_OMEGA

9.511.251521.2525LEVEL4_CORINA

1213.7516.6622.525LEVEL5_RUBICON

911.2513.3318.7525LEVEL5_OMEGA

89.37514.172025LEVEL5_OMEGA

9.511.251521.2525LEVEL5_CORINA

18 purchased and assayed

Does it really work ?– Validate, Validate, Validate

Compound Number IC50 in MCF-7 MTTMDG-ER-001 8.23E-07MDG-ER-002 8.00E-06MDG-ER-003 2.02E-05MDG-ER-004 5.59E-04MDG-ER-005 6.06E-04TAMOXIFEN 5.51E-06

Screen ligands, prepare ranked hitlist

cluster hits – 20 clusters

5 Hits 5 Hits µµµµµµµµm rangem range4 Chemical Classes4 Chemical Classes3 novel Chemotypes3 novel Chemotypes

MW 450MW 450--550550LOGP 4.8LOGP 4.8--6.56.5

What else do we need?

Familial scoring functions

Validation Validation Validation

Flexible systems – dynamics in dockingChemical intelligence in fragment assembly

Tiered Discovery – integration of technologies

System Simulation

Acknowledgements

The ER collaboratorsDr Mary Meegan – School of Pharmacy TCDDr Vladimir Sobolev – Weismann Institute, IsraelProf James Sexton – Trinity Centre for High Performance ComputingProf Clive Williams – Biochemistry TCDDr Daniela Zisterer – Biochemistry TCDDr Amir Khan – Biochemistry TCD

The workers The facilitatorsAndy Knox Dermot FrostYidong YangGiorgio CartaValeria OnnisGeorgia Golfis

a presentation by Dr David Lloyd Trinity College Dublin ...Compound Number IC50 in MCF-7 MTT MDG-ER-001 8.23E-07 MDG-ER-002 8.00E-06 MDG-ER-003 2.02E-05 MDG-ER-004 5.59E-04 MDG-ER-005

Documents