Design of Compound Libraries for Fragment Screening

Peter W. Kenny

pwk.pub.2008@gmail.com | http://fbdd-lit.blogspot.com

Achtung!Spitfire!

A Brief History of Screening: In the Beginning…

Stuka on wikipedia

“Why can’t we pray for something good, like a tighter bombing pattern, for example? Couldn’t we pray for a tighter bombing pattern?” , Heller, Catch 22, 1961

… and then there was HTS

B52 on wikipedia

So, Maria, why do you think it is that the

Russians are so much better than the

Germans at tennis these days?

Actually we startedto beat them at their national sport almost

70 years ago and...

.... as Uncle Joe was so fond of saying, quantity

has a quality all of its own.

Even the stars of tennis have heard of HTS....

The HTS was YOUR idea so don’t try blaming me!

Unfortunately HTS is not a panacea

FBDD Essentials

Screen fragments

Synthetic

Elaboration

Target

Target & fragment hit

Target & lead

Linking

Fragment Elaboration Tactics

Merging

Growing

Why fragments?

• Access to larger chemical space

• Counter the advantage of competitors’ large

compound collections

• Ligands are assembled from proven molecular

recognition elements

• A smart way to do Structure-Based Design

• Control resolution at which chemical space is

sampled

PTP1B (Diabetes/Obesity): Fragment elaboration

Elaboration by Hybridisation: Literature SAR was mapped onto intial fragment hit

(green). Note overlay of aromatic rings of elaborated fragment (blue) and

difluorophosphonate (red). See Black et al BMCL 2005, 15, 2503-2507 |

http://dx.doi.org/doi:10.1016/j.bmcl.2005.03.068

Inactive at 200mM

15 mM

3000 mM3 mM

150 mM

(Conformational lock)

130 mM

(3-Phenyl substituent)

Fragment-based lead discovery: Generalised workflow

Target-based compound selection

Analogues of known binders

Generic screening library

Measure

Kd or IC50

Screen

Fragments

Synthetic elaboration

of hits

SARProtein

Structures

Milestone achieved!Proceed to next

project

A model for molecular complexity

This model is equally relevant to conventional and fragment-based screening. See Hann, Leach

& Harper J. Chem. Inf. Comput. Sci., 2001, 41, 856-864 | http://dx.doi.org/10.1021/ci000403i

Success landscape

Degree of substitution as measure of molecular complexity

The prototypical benzoic acid can be accommodated at both sites and, provided that binding can be

observed, will deliver a hit against both targets See Blomberg et al JCAMD 2009, 23, 513-525 |

http://dx.doi.org/10.1007/s10822-009-9264-5 | This way of thinking about molecular complexity is

similar to the ‘needle’ concept introduced by Roche researchers. See Boehm et al J. Med. Chem.

2000, 43, 2664-2774 | http://dx.doi.org/10.1021/jm000017s

Ligand Efficiency

LE= DGº/NonHydHopkins, Groom & Alex, DDT 2004, 9, 430-431

Binding Efficiency Index

BEI= pIC50/(MW/kDa)Abad-Zapaftero & Metz, DDT 2005, 10, 430-431

Ligand Lipophilicity Efficiency

LLE = pIC50 - ClogPLeeson & Springthorpe , NRDD 2007, 6, 881-890.

Measured binding is scaled Measured binding is offset

Binding Efficiency

Measures

Fragment screening requirements

• Assay capable of reliably quantifying weak (~mM)

binding

• Library of compounds with low molecular complexity

and good (~mM) aqueous solubility

•

Screening Library Design Requirements

• Precise specification of substructure

– Count substructural elements (e.g. chlorine atoms; rotatable bonds;

terminal atoms; reactive centres…)

– Define generic atom types (e.g. anionic centers; hydrogen bond

donors)

– SMARTS notation is particularly useful

• Meaningful measure of molecular similarity

– Structural neighbours likely to show similar response in assay

Measures of Diversity & Coverage

•• •

•

••

•

•

•

••

•

••

•

2-Dimensional representation of chemical space is used here to illustrate concepts of diversity and

coverage. Stars indicate compounds selected to sample this region of chemical space. In this

representation, similar compounds are close together. The title slide for this talk shows the optical range

finder that was salvaged from the pocket battleship Admiral Graf Spee and can be seen in Montevideo.

Coverage, Diversity & Library Design

••

• ••

•• •• •• •

•

Neighborhoods and library design

Hits, non-hits & lipophilicity: Survival of the fattest*

Mean Std Err Std Dev

Hits 2.05 0.08 1.10

Non-Hits 1.35 0.03 1.24

*Analysis of historic screening data & quote: Niklas Blomberg, AZ Molndal

Comparison of ClogP for hits and non-hits from

fragment screens run at AstraZeneca

Sample

AvailabilityMolecular

Connectivity

Physical

Properties

screening samples Close analogs Ease of synthetic

elaboration

Molecular

complexity

Ionisation Lipophilicity

Solubility

Molecular

recognition

elementsMolecular shape

3D Pharmacophore

Privileged

substructures

Undesirable

substructures

Molecular

size

3D Molecular

Structure

Fragment selection criteria

Why I don’t use the rule of 3: http://fbdd-lit.blogspot.com/2011/01/rule-of-three-considered-harmful.html

Acceptable diversity

And coverage?

Assemble library in

soluble form

Add layer to core

Incorporate layer

Yes

No

Select core

Core and layer library design

Compounds in a layer are selected to be diverse with respect to core compounds. The ‘outer’ layers

typically contain compounds that are less attractive than the ‘inner’ layers. This approach to library

design can be applied with Flush or BigPicker programs (Dave Cosgrove, AstraZeneca, Alderley Park)

using molecular similarity measures calculated from molecular fingerprints. See Blomberg et al

JCAMD 2009, 23, 513-525 | http://dx.doi.org/10.1007/s10822-009-9264-5

ClogP: Charged library compounds

ClogP: Neutral library compoundsNon-hydrogen atoms

GFSL05: Size and lipophilicity profiles

Rotatable bonds

61

1713

4 4

1

0

Breakdown of GFSL05 by charge type

Neutral

Anion Cation

Ionisation states are identified using AZ ionisation and tautomer model. Multiple forms are generated for

acids and bases where pKa is thought to be close to physiological pH. See Kenny & Sadowski Methods

and Principles in Medicinal Chemistry 2005, 23, 271-285 | http://dx.doi.org/10.1002/3527603743.ch11

GFSL05: Numbers of neighbours within library as function of

similarity (Tanimoto coefficient; foyfi fingerprints)

0.90 0.85 0.80

FBDD Blogs

These are ‘crosslinked’ and both will direct you to LinkedIn & facebook groups

Practical Fragments: http://practicalfragments.blogspot.com

FBDD Literature: http://fbdd-lit.blogspot.com

Spare slides

A (small) selection of literatureGeneral

• Erlanson, McDowell & O’Brien, Fragment-Based Drug Discovery, J. Med. Chem., 2004, 47, 3463-3482 | http://dx.doi.org/10.1021/jm040031v

• Congreve et al. Recent Developments in Fragment-Based Drug Discovery, J. Med. Chem., 2008 51, 3661–3680 | http://dx.doi.org/10.1021/jm8000373

• Albert et al, An integrated approach to fragment-based lead generation: philosophy, strategy and case studies from AstraZeneca's drug discovery programmes. Curr. Top. Med. Chem. 2007, 7, 1600-1629 | http://www.ingentaconnect.com/content/ben/ctmc/2007/00000007/00000016/art00006

• Hann, Leach & Harper, Molecular Complexity and Its Impact on the Probability of Finding Leads for Drug Discovery, J. Chem. Inf. Comput. Sci., 2001, 41, 856-864 | | http://dx.doi.org/10.1021/ci000403i

Screening Libraries• Blomberg et al, Design of compound libraries for fragment screening, JCAMD 2009, 23, 513-525 |

http://dx.doi.org/10.1007/s10822-009-9264-5 • Schuffenhauer et al, Library Design for Fragment Based Screening, Curr. Top. Med. Chem. 2005, 5,

751-762 | http://www.ingentaconnect.com/content/ben/ctmc/2005/00000005/00000008/art00003• Baurin et al, Design and Characterization of Libraries of Molecular Fragments for Use in NMR

Screening against Protein Targets, J. Chem. Inf. Comput. Sci., 2004, 44, 2157-2166 | http://dx.doi.org/10.1021/ci049806z

• Colclough et al, High throughput solubility determination with application to selection of compounds for fragment screening. Bioorg, Med. Chem. 2008, 16, 6611-6616 | http://dx.doi.org/doi:10.1016/j.bmc.2008.05.021

• Kenny & Sadowski, Structure modification in chemical databases. Methods and Principles in Medicinal Chemistry 2005, 23, 271-285 | http://dx.doi.org/10.1002/3527603743.ch11

Scheme for fragment based lead optimisation

20%10%

30%

40%

50%

log(S/M)

Aqueous solubility:

Percentiles for measured log(S/M) as function of ClogP

Data set is partitioned by ClogP into bins and the percentiles and mean ClogP is calculated for each. This way of plotting results is

particularly appropriate when dynamic range for the measurement is low. Beware of similar plots where only the mean or median

value is shown for the because this masks variation and makes weak relationships appear stronger than they actually are. See

Colclough et al Bioorg. Med. Chem. 2008, 16, 6611-6616 | http://dx.doi.org/doi:10.1016/j.bmc.2008.05.021

Measure solubility for

neutral (at pH 7.4)

fragments for which

ClogP > 2.2

GFSL05 project

• Strategic requirement:

– Readily accessible source of compounds for a range of fragment

screening applications (NMR, Biochemical Assay, HTS at 10 x

normal concentration)

• Tactical objective:

– Assemble 20k structurally diverse compounds with properties that

are appropriate for fragment screening as 100mM DMSO stocks

GFSL05: Overview

• Molecular recognition considerations

– Requirement for at least one charged center or acceptably strong

hydrogen bond donor or acceptor

• Substructural requirements defined as SMARTS

– Progressively more permissive filters to apply core and layer design

– Restrict numbers of non-hydrogen atoms (size) and extent of

substitution (complexity)

– Filters to remove undesirable functional groups (acyl chloride) and

to restrict numbers of others (nitro, chloro)

– ‘Prototypical reaction products’ for easy follow up

• Control of lipophilicity (ClogP) dependent on ionisation state

– Solubility measurement for more lipophilic neutrals

• Tanimoto coefficient calculated using foyfi fingerprints (Dave

Cosgrove) as primary similarity measure

– Requirement for neighbour availability in core and layer design

APGNMR07: Overview

• General

– Designed for NMR screening (especially 2D protein detect)

– 1200 Compounds

– Derived in part from existing AZ NMR libraries and GFSL05

– Molecules smaller on average than those in GFSL05

– Stock solutions: 200mM in d6-DMSO

• Partitioning of library for cocktailing

– Groups (200) of 6 compounds defined

– Allows screening in mixtures of 6 or 12

– Acid:Base:Neutral = 1:1:4

GFSL05

APGNMR07

Lipophilicity profiles for GFSL05 and APGNMR07

GFSL05: Numbers of available neighbours as function of similarity

(Tanimoto coefficient; foyfi fingerprints) and sample weight

>10mg

>20mg

0.90 0.85 0.80

0.90 0.85 0.80