DepartmentAuthor

Lead-like Properties, High-throughput Screening and Combinatorial Library Design

Andy Davis, Simon Teague, Tudor Oprea, John Steele, Paul Leeson

Teague, Leeson, Oprea, Davis, Angew Chem 1999, 38, 3743

DepartmentAuthor

Fastest - first and best

Target HTSHit

EvaluationHit toLead

LeadOptimisation

DESIGN AND

SYNTHESIS

Potency Efficacy

Selectivity

compounds

information

Kinetics Metabolism Enzymology

compounds

LeadHTS + Combichem

DepartmentAuthor

Fisons History

• Early lit work - largely peptidic

• Approaches available to us• solid phase ?

• Solution phase ?

• Singles or mixtures ?

DepartmentAuthor

Design Criteria

• Library Design Buzzwords and Concepts• “Diverse“

• “Universal !”

• Pharmacophore mapping libraries

• focussed libraries

DepartmentAuthor

“Universal” Library

SNH2

R

R160CO2H

Solidphase

SHNH

R160

O

R40X

Solutionphase

SNH

R160

O

R40

i) Solid phase R80CO2H

ii) Cleave

NH2R N

HR

80

O

Br

STEP 2

NuNH

R80

O

Nu

STEP 1

Approach 1

Approach 2

Walters and Teague Tet Lett. 2000, 41, 2023

DepartmentAuthor

Charnwood “Universal” Library

Distribution of ACDlogP's in Universal Library vs PDR Drugs

0

5

10

15

20

-5 -2 1 4 7

10

13

16

ACDlogPs

% C

ou

nt

PDR ACDlogP's

% Universal logP's

Distribution of MWt in Universal Library vs PDR Drugs

0

5

10

15

20

25

10

0

25

0

40

0

55

0

70

0

85

0

10

00

MWt

%C

ou

nt

PDR MWt

% universal MWt

55,000 member library

DepartmentAuthor

Distribution of Ns and Os in PDR and GPCR Libraries

0

5

10

15

20

25

30

35

40

0 4 8

12

16

20

Ns and Os

%C

ou

nt

% Ns Os PDR

% Ns and Os GPCR

Distribution of donors in PDR and GPCR Libraries

0

5

10

15

20

25

30

35

0 1 2 3 4 5 6

Mo

re

donors

% C

ou

nt

%dons PDR

%dons GPCR

Distribution of ACDlogPs in PDR and GPCR Libraries

0

5

10

15

20

25

-5 -2 1 4 7

10 13 16

ACDlogP

% o

ccu

r PDR ACDlogP

GPCR ACDlogP

Distribution of Mwt in PDR and GPCR Libraries

0

5

10

15

20

25

301

00

20

0

30

0

40

0

50

0

60

0

70

0

80

0

90

0

10

00

Mwt

% O

cc

ur

PDR MWt

GPCR Mwt

Early GPCR Library

DepartmentAuthor

The Age of Lipinski

• HTS lead generation biases chemistry

alertsHTS

DepartmentAuthor

Design Criteria

• Library Design Buzzwords and Concepts• “Diverse“

• “Universal”

• Pharmacophore mapping libraries

• Drug-like properties– Lipinski etal Adv Drug Del. Rev. 1997, 23, 3-25

– Sadowski, J. Med. Chem, 1998. 41, 3325.

– Ajay etal, J.Med.Chem, 1998, 41, 3314

• focussed libraries etc etc.

DepartmentAuthor

Our experiences ??• by 1998

• 75%+ screening bank Combi derived

• applied current design criteria• focussed upon “drug-like libraries”

• we are looking for drug-like potency - • do we find it ??

0

5

10

15

20

4.5 5

5.5 6

6.5 7

7.5 8

pIC50

%

cou

nt

3000 hits 1e6 screen points

DepartmentAuthor

Charnwood Confirmed HTS Hits

• In > 1e6 screen tests - not 1 nM hit• probability of a nM hit < 1e-6

• But hits are already drug-like size

0

5

10

15

20

4.5 5

5.5 6

6.5 7

7.5 8

pIC50

%

cou

nt

0

5

10

15

20

150

250

350

450

550

650

750

850

950

MWt

%

cou

nt

3000 hits 1e6 screen points

DepartmentAuthor

Bang for your Buck• Andrews analysis (J Med Chem 1984, 27, 1648.)

• scoring without a protein– analysed 200 good ligands for their receptor

– assume all interactions are optimally made

– apply fn group counts = regression vs potency

G (kcal/mol) = -14 -0.7n DOF + 0.7 n Csp2 + 0.8 n Csp3 +11.5nN++1.2n N +8.2n CO2- + 10n PO4- + 2.5n OH + 3.4 n C=O +1.1 n O,S +1.3n hal

D Williams GHB = 0.5-1.5 kcal/mol Glipo = 0.7 kcal/mol -CH3

Grot= 0.4 - 1.4 kcal/mol

Williams etal Chemtracts, 1994, 7, 133

DepartmentAuthor

y = 1.1993x - 2.4771R2 = 0.3168

-10

-5

0

5

10

15

20

25

0 5 10 15 20

obsd pKi

An

dre

ws

pK

i

Andrews Analysis Training set

• Significant ,model incl by 2 outliers

Biotin

DepartmentAuthor

Andrews - 2

DepartmentAuthor

Andrews - Coloured by Charge

• Multiply charged compounds overpredicted• oral targets 0,1 charge

DepartmentAuthor

Final Model - 0,1 charges

DepartmentAuthor

0

10

20

30

40

50

60

70

80

90

-2 0 2 4 6 8 10 12 14 16 18 >18pIC 50

y /%

0

2

4

6

8

10

12

14

16

-2 0 2 4 6 8 10 12 14 16 18 >18pK i

y /%

Andrews predictions

HTS Obsd activities

HTS screening Hits

• probabilities• predicted

– p(<10nM) = 22%

• obsd– p(<10nM) <e-8%

Many hits underperform

DepartmentAuthor

HTS Screening Hits• Drug-like hits

– potency of many underperform– binding via weak non-specific interactions

– not all interactions made or very suboptimal

– would explain “flat SAR” in Hit-to-Lead activities

– small M leads easier to optimise than large M

• “easy” and “difficult” hit-to-lead projects• easy to increase Mwt/logP - increase potency

– easy to demonstrate SAR, increase potency 10x

• difficult because of flat SAR– difficult to reduce Mwt and logP maintaining potency

–

DepartmentAuthor

HtL Examples - GPCR Project

N

CONH2SR

SOH

O

NH

O

R

acid

IC50 = 4.6 MMwt 268ClogP 3.4

IC50 = 0.55 MMwt 350clogP 3.7

IC50 = 0.18 MMwt 380ClogP = 4.5

DepartmentAuthor

GPCR Hit-to-Lead

Many analoguessame or loss potency

CONH2SR

SOH

O

NH

O

R

Many analoguessame potency

• Both series dropped -

DepartmentAuthor

GPCR Hit-to-Lead

• Rapid Hit-to-Lead optimisation• clear SAR

• drug-like series with good DMPK

• patentable

N

acid

Cl

Cl

acid

IC50 = 4.6 MMwt 268ClogP 3.4

IC50 = 0.02 MMwt 336ClogP 5.3 (:-<)

DepartmentAuthor

22

MWt Distribution of PDR Drugs and Renin Inhibitors

0

5

10

15

20

251

00

20

0

30

0

40

0

50

0

60

0

70

0

80

0

90

0

10

00

110

0

12

00

MWt

% C

ou

nt

PDR MWt

renin

“Difficult” Project - 2 Renin Inhibitors

No renin inhibitor went passed PII

all failed due to poor bioavailability, high cost

DepartmentAuthor

0

5

10

15

20

25

100 200 300 400 500 600 700

M r

y / %

Process Lead Optimisation

0

5

10

15

20

25

100 200 300 400 500 600 700

M r

y / % b

0

5

10

15

20

25

100 200 300 400 500 600 700

M r

y / %

PDROutside drug space old Combi Library

Lead-like

• Optimisation Hypothesis

DepartmentAuthor

Bang for your Buck - 2Would a lead-like compound “hit” in HTS ?

• Andrews analysis of leads• estimated pKi for “leadlike” ligand

• 15,000 “random” drugs designed

• random numbers of “features bounded by oral drugs

filtered by est Mwt - and 0,1 charge

G (kcal/mol) = -14 - 0.7n DOF (n = 1-8) + 0.75 n Csp2+sp3 (n=4-18)

+ 11.5n N+ (n=0,1) + 1.2n N (n=0-4) + 2.5n OH (n=0,1) + 3.4 n C=O

(n=0-2) + 1.1 n O,S (n=0-2) + 1.3n hal (n=0,1)

DepartmentAuthor

Leadlike Bang for your Bucks

• HTS screening environment• Small leads probably need 1 charge @10M

Distribution of Andrews predicted pKi for neutral and basic leads Mwt <300

0

100

200

300

400

500

600

700

-9 -7 -5 -3 -1 1 3 5 7 9 11 13 15predicted pKi

Co

un

t

N+

N

DepartmentAuthor

100 lead - drug pairs

DepartmentAuthor

1998: less than 600 solid compounds with mwt <250 and clogP <2

1999: 3000 added by purchase. Synthesis added >30000

1998: less than 600 solid compounds with mwt <250 and clogP <2

1999: 3000 added by purchase. Synthesis added >30000

Lead-like Profile• Mwt 200-350

• optimisation adds ca. 100

• logP 1-3• optimisation may increase by 1-2 logunits

• single charge• positive charge preferred

• secondary or tertiary amine

DepartmentAuthor

Distribution of Ns and Os in PDR and GPCR Libraries

0

5

10

15

20

25

30

35

40

0 4 8

12

16

20

Ns and Os

%C

ou

nt

% Ns Os PDR

% Ns and Os GPCR

Distribution of donors in PDR and GPCR Libraries

0

5

10

15

20

25

30

35

0 1 2 3 4 5 6

Mo

re

donors

% C

ou

nt

%dons PDR

%dons GPCR

Distribution of ACDlogPs in PDR and GPCR Libraries

0

5

10

15

20

25

-5 -2 1 4 7

10 13 16

ACDlogP

% o

ccu

r PDR ACDlogP

GPCR ACDlogP

Distribution of Mwt in PDR and GPCR Libraries

0

5

10

15

20

25

301

00

20

0

30

0

40

0

50

0

60

0

70

0

80

0

90

0

10

00

Mwt

% O

cc

ur

PDR MWt

GPCR Mwt

Early GPCR Library

DepartmentAuthor

0

5

10

15

-5 -2.5 0 2.5 5 7.5 10

ACDlogP

cou

nt % PDR99

leadlike

Mitsunobu Library

05

1015202530354045505560

0 2 4 6 8 10

Donors

cou

nt % PDR99

leadlike

0

5

10

15

20

25

30

35

40

45

0 4 8 12 16 20 24

NsOs

cou

nt % PDR99

leadlike

0

5

10

15

20

25

30

35

40

100 200 300 400 500 600 700 800 900

Mwt

cou

nt % pdr99

leadlike

DepartmentAuthor

Lead Continiuum -

350Mwt >500 Mwt <200

Drug-likeLeadlike HtL problems ?

Topical target ?

HTS screeningNon-HTS

Shapes (Vertex )Needles(Roche)MULBITS(GSK)Crystallead(Abbott)

DepartmentAuthor

Screening File Split• Step taken by some companies - drivers

• logical conclusion of leadlike paradigm

• cost/feasibility some HTS technologies

Screening file

Good oral file Bad - topical/desperate file

DepartmentAuthor

Summary• HTS

• starting points are crucial to speed throughout process

• screening file should reflect what chemists can easily work upon• ideally we all want to find drugs in our screening file

– but generally a HTS finds only leads not drugs

• file-size isnt everything = quality is equally important

• Libraries• Many approaches - targeted libraries v successful

– kinase libraries - 4x hit rate - screening file

• libraries should reflect what you wish to find– leads not drugs

Teague, Leeson, Oprea, Davis, Angew Chem 1999, 38, 3743

DepartmentAuthor