[Methods in Enzymology] Fragment-Based Drug Design - Tools, Practical Approaches, and Examples...

C H A P T E R O N E

M

IS

SL

ethods

SN 0

tructu.L.C.,

Designing a Diverse High-Quality

Library for Crystallography-Based

FBDD Screening

Brett A. Tounge and Michael H. Parker

Contents

1. In

in

076

ral BSpr

troduction

Enzymology, Volume 493 # 2011

-6879, DOI: 10.1016/B978-0-12-381274-2.00001-7 All rig

iology and Medicinal Chemistry, Johnson & Johnson Pharmaceutical Research and Ding House, Pennsylvania, USA

Else

hts

eve

4

2. L
ibrary Requirements for Different Screening Methods 6
2
.1. T raditional biophysical screening methods 7
2
.2. N MR screening 7
2
.3. X -ray screening 7
3. L
ibrary Design for X-Ray Screening 8
3
.1. P roperty filters 8
3
.2. F ragment ranking—FBDD Score 9
3
.3. D iversity 11
4. Im
plementation 13
4
.1. X -ray primary screening library 13
4
.2. Q uantity and purity 13
4
.3. C lustering for plating 17
5. C
onclusions 17
Refe
rences 19
Abstract

A well-chosen set of fragments is able to cover a large chemical space using a

small number of compounds. The actual size and makeup of the fragment set is

dependent on the screening method since each technique has its own practical

limits in terms of the number of compounds that can be screened and require-

ments for compound solubility. In this chapter, an overview of the general

requirements for a fragment library is presented for different screening plat-

forms. In the case of the FBDD work at Johnson & Johnson Pharmaceutical

Research and Development, L.L.C., our main screening technology is X-ray

crystallography. Since every soaked protein crystal needs to be diffracted and

a protein structure determined to delineate if a fragment binds, the size of our

vier Inc.

reserved.

lopment,

3

Figureand Datoms

4 Brett A. Tounge and Michael H. Parker

initial screening library cannot be a rate-limiting factor. For this reason, we have

chosen 900 as the appropriate primary fragment library size. To choose the best

set, we have developed our own mix of simple property (“Rule of 3”) and “bad”

substructure filtering. While this gets one a long way in terms of limiting the

fragment pool, there are still tens of thousands of compounds to choose from

after this initial step. Many of the choices left at this stage are not drug-like, so

we have developed an FBDD Score to help select a 900-compound set. The

details of this score and the filtering are presented.

1. Introduction

The typical collection, or deck, used for high throughput screening(HTS) is comprised mostly of compounds that have �15 nonhydrogenatoms. As an example, the size distribution of our compound collection atJohnson & Johnson Pharmaceutical Research and Development, L.L.C.shows an average nonhydrogen atom count of �30 (Fig. 1.1). Compoundssmaller than this, fragments, tend to have lower absolute affinities, and thusare difficult to detect in routine HTS campaigns. However, over the pastdecade, there has been a considerable amount of effort put into making useof these smaller compounds (Chen and Hubbard, 2009; Chessari andWoodhead, 2009; de Kloe Gerdien et al., 2009; Fischer and Hubbard,2009; Murray and Rees, 2009; Orita et al., 2009b; Schulz and Hubbard,2009; Wang et al., 2009). These efforts have been driven by the recognitionthat fragments offer several unique properties relative to typical “drug-size”molecules. Since smaller compounds are less complex, they have a higher

0

20,000

40,000

60,000

80,000

100,000

1 6 11 16 21 26 31 36 41 46 51 56 61

Cou

nt

HA

1.1 The size distribution of the Johnson & Johnson Pharmaceutical Researchevelopment, L.L.C. compound collection is centered on �30 nonhydrogen(HA). The typical HTS screen uses compounds with �15 nonhydrogen atoms.

FBDD Library Design 5

probability of binding, tend to bind with higher ligand efficiencies, covergreater chemical space with fewer compounds, and, depending on how thefragments are chosen, have no built-in bias for a particular target.

As one moves from large complex drug-size molecules to smaller frag-ments, there are fewer constraints that need to be met for a compound tobind to a target protein. For example, for a protein to accommodate threehydrogen bonding partners while maintaining good van der Waals contactsrequires a very specific binding pocket and ligand geometry. However,many proteins can accommodate a single hydrogen bond acceptor attachedto a phenyl ring. As a result, the probability of binding goes up as the size ofthe ligand goes down (Hann et al., 2001).

Support for this theory can be seen in the high ligand efficiencies foundfor smaller molecules. Ligand efficiency is generically defined asDDGbinding/HA, where HA is the nonhydrogen count. Numerous papershave been published showing that smaller molecules bind with higherligand efficiencies (Fig. 1.2; Abad-Zapatero, 2007; Bembenek et al., 2009;Hopkins Andrew et al., 2004; Nissink, 2009). This is directly related to theargument made above for binding probability. Smaller, less complex mole-cules have fewer constraints that need to be met when binding and thus theytypically have a better “fit” to the protein.

The efficiency of chemical space coverage afforded by using fragments isillustrated in Fig. 1.3. There are �33 unique six-membered rings in theComprehensive Medicinal Chemistry database (Bemis and Murcko, 1996,

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 10 20 30 40 50 60 70

LE

HA

Figure 1.2 Ligand efficiency (LE) shows a precipitous decline between 10 and 25nonhydrogen atoms (HA).We have extracted the affinity data used in this plot from theBindingDB database developed at the University of Maryland Biotechnology Institute(Liu et al., 2007).

Ideal hit

Fragmentscreen

Ligandscreen

1:561

1:33 1:33

HN

Figure 1.3 By breaking compounds into smaller substituents, one is able to coverchemical space more efficiently. In this case, by screening the piperidine and phenylrings as fragments, the ideal two-ring system could be mapped by screening just 33compounds. Enumerating all 33 possible six-membered rings into a two-ring systemwould result in 561 compounds that would need to be screened.


1999). For just a simple two-ring system, that would result in 561 uniquecombinations (assuming a single attachment point for each ring). As a result,to find the “ideal” hit, all 561 compounds would need to be screened. If onetakes the fragments individually, only 33 compounds would need to bescreened to find the best two rings.

Finally, one of the biggest advantages for fragment screening is that youare not starting with compound libraries that are already biased for specifictarget classes. For example, many corporate compound collections areheavily populated with ligands designed to hit kinases. The biased ligandswill likely not work well for very divergent protein classes. In contrast, bybreaking ligands into their component fragments, the resulting library willhave broader applicability in screening.

2. Library Requirements for Different

Screening Methods

For all these reasons, fragment-based drug discovery (FBDD) hasgreatly expanded over the past decade. In particular, there has been consid-erable effort put into developing screening technologies suited to detectingweak binders. The methodologies used fall into three basic categories:traditional biophysical screening, NMR, and X-ray crystallography(Barker et al., 2007; Blaney et al., 2006; Dalvit, 2009; Danielson, 2009;


Hartshorn et al., 2005; Jhoti, 2007; Jhoti et al., 2007; Orita et al., 2009b;Perspicace et al., 2009). Each of these methods represents tradeoffs betweenthroughput and structural information content. As a result, the librarydesign criteria are screening method-dependent.

2.1. Traditional biophysical screening methods

Traditional biophysical methods (e.g., surface plasma resonance, Thermo-fluorÒ) represent the high end in terms of throughput. For most of thesemethods, running�100,000 compounds for a screen is routine.However, thevery high throughput methods represent the other end of the spectrum interms of binding site information content. Traditional biophysical methodssimply tell you whether or not a binding event occurs, but not where on theprotein it occurred. As a result, these methods are often used as a prescreen toselect ligands for NMR or X-ray crystallography. In addition, compoundsolubility is a crucial aspect of the library design for these methods.

2.2. NMR screening

Ligand-detected NMR-based screening falls in the middle in terms of bothstructural information and throughput. Typical fragment screening librariesfor NMR are 20,000–50,000 in size. Ligand-detected NMR methods offerlimited information in terms of binding. At best, the pharmacophore for theligand can be defined using saturation transfer difference-based methodswhich tell you what part of the ligand is in contact with the protein (Mayerand Meyer, 1999). Protein chemical shift perturbation-based NMR meth-ods offer more structural information, but are lower throughput (�10,000compounds; Hajduk et al., 1999). In addition, since the backbone resonanceassignments must be done, chemical shift perturbation techniques are lim-ited to proteins �40 kDa in molecular weight. Finally, as in the methodsmentioned above, compound solubility is crucial for NMR-basedscreening.

2.3. X-ray screening

While X-ray screening represents the low end in terms of typical through-put (�1000) and it can only be applied in cases where a robust X-raystructure can be produced, it offers the highest information content. Once ahit is found, the exact binding location and orientation is known, whichallows for more direct follow-up chemistry. In addition, we have found thatin practice, compound solubility is not a limiting factor. As long as we canobtain a sufficient concentration in the buffer solution for soaking, anyprecipitate simply provides a source for additional compound as fragmentssoak into the crystal lattice.


3. Library Design for X-Ray Screening

For various reasons that are outlined in other chapters of this book,X-ray crystallography was chosen as the main screening platform for ourFBDD program. As mentioned above, the low throughput of this screeningmethod forces one to balance screening coverage with time to screen. Inpractice, we settle on a 900-compound screening set. This allows for suffi-cient chemical space coverage and allows the screen to be accomplished in areasonable time frame. Various methods have been published for makingsuch a selection. We will review some of the more common techniques aswell as present our unique metric for final compound selection.

3.1. Property filters

The overall process for compound selection is summarized in Fig. 1.4. Ourinitial pool of candidate fragments is drawn from both our in-house libraryand commercial sources. This rather large list can be quickly reduced byapplying simple property filters. The most commonly used filter is the“Rule of 3” (Congreve et al., 2003). For the initial property filtering step,we use a modified rule of three set,

1. 5 � nonhydrogen atoms � 152. Hydrogen bond acceptors � 33. Hydrogen bond donors � 34. 1 � number of rings � 35. Number of unspecified stereo centers ¼ 0

ACD 600,000 compounds

80,000

50,000

FBDD library

Diversity

Drug-like

Property filtering

Figure 1.4 Overview of the general filtering process used to build FBDD libraries.The compound counts after each stage is based on a filtering of the ACD.


coupled with a substructure search to remove unwanted functional groups(Table 1.1).

As an example, the above filtering is applied to the Available ChemicalsDirectory (ACD; Symyx Solutions, Inc., 2007a). After running the prop-erty filtering on the initial set (632,790 compounds), 101,952 compoundsremain. The substructure filtering then brings the list down to 77,078.

3.2. Fragment ranking—FBDD Score

At this point, several methods are available to rank and select fragments(Baurin et al., 2004; Blomberg et al., 2009; Brewer et al., 2008; Colcloughet al., 2008; Jacoby et al., 2003; Mercier et al., 2006; Orita et al., 2009a;Schuffenhauer et al., 2005). While diversity metrics could be used, they tendto pick out non-drug-like compounds since the rough property filteringoutlined above does not remove these compounds. Instead, it is importantat this stage to introduce a method to select the more desirable compounds.For this, we use a metric we call FBDD Score that captures size, chemicalcomplexity, and drug-likeness.

For drug-likeness, we have adapted a previously published method usedto rank reagents for combinatorial chemistry libraries (Tounge andReynolds, 2004). For each fragment molecule, we calculate a subsimilarityto each compound in a drug-like database (the Comprehensive MedicinalChemistry database) and keep the highest subsimilarity score (SymyxSolutions, Inc., 2007b). The score (CMCSubSim) is computed using thefollowing general formula:

CMCSubSim ¼ Total number of keys in the target that match keys in the probe

Keysp

ð1:1Þ

where Keysp is the total number of unique keys in the probe molecule (i.e.,the fragment). In our implementation, the Extended Connectivity with apath length of 4, ECFP_4, descriptor keys are used (Rogers and Hahn, 2010).The CMCSubSim score ranges from 1 to 0. A fragment molecule whoseentire structure is found in one or more compounds from the referencedatabase is considered drug-like and will have a CMCSubSim ¼ 1.

The next element of the score is designed to capture the complexity ofthe molecule. The goal with this aspect of the score is to bias our set offragments to those with a simple pharmacophore (e.g., limit the number offunctional groups (amines, acids, halogens, etc.) on a given fragment). Thiskeeps the probability of binding high since fewer constraints need to be met.To capture this, we use the simple metric of calculating the percentage ofheteroatoms in the fragment. This component of the score is simply

Table 1.1 A substructure filtering of the initial fragment pool is run to eliminate all compounds that have the functional groups listed

1-2 Diketones Aldehydes Itrazole-like Diazo Azides

Acetylene Alkyltriene Long chain aliphatics (C7þ) Nitroxide Nitro

Nitrile Alkyl halide Carbamoyl halides Peroxide S-Oxide

Acetals a-Halo carbonyls Carbodiimides Hypoiodate-like Phosphorhalide

Acid chlorides Anhydrides Disulfide oxides Isocyanates Diphosphide

Acrylates Anthracenes 1,4-Pentadiene-3-diphenylmethylene Isothiocyanates Sulfonyl halides

Acyl halides Aziridines Fullerenes Methylenehydrazines Thiocarbamoyl halides

Acylhydrazinones Oxiridines Halogen-heteroatom Diamine (N_chain_N) 1,1,2-Trimercaptoethylene

Adamantanes Miconazole-like Hemes N-hydroxy Thioyl halides


PercHetAtom ¼ HA�Number of carbon atoms

HAð1:2Þ

where HA is the nonhydrogen atom count.The final element of the score is added to bias our selections to smaller

molecules. This is accomplished by adding a penalty score which is defined as

SizePenalty ¼ HA� 6

9ð1:3Þ

This score applies no penalty when the atom count is 6 and –1 forfragments with 15 nonhydrogen atoms. This aspect of the score can beadjusted depending on the size bias one wants to apply.

All these components are combined into the final FBDD Score as follows:

FBDDScore ¼ CMCSubSim� SizePenalty

2� PercHetAtom

2ð1:4Þ

In this combination of the terms, both the size and complexity termshave been divided by 2 in order to give more weight to the drug-like term.An example of the score can be found in Fig. 1.5. Once applied to the entireproperty-filtered fragment set, the score allows one to pick out drug-like,small compounds with simple pharmacophores. In practice, all compoundswith an FBDD Score � �0.0 are kept (Fig. 1.6). For example, scoring andfiltering the remaining 77,078 compounds from the above ACD filteringusing a cutoff of �0 leaves 53,515 compounds from which to do the finallibrary selection.

3.3. Diversity

Once the FBDD Score filtering is done, diversity can be used to select thefinal number of fragments needed for the screen. At this point, manydifferent algorithms are available. The details of the final selection of ourfragment library for X-ray screening set are outlined in Section 4 of thischapter.

CMCSubsim

0.94O

FBDD score=0.71

–0.17 –0.06

9 atoms 1heteroatom

Figure 1.5 The FBDD Score is composed of three terms and helps to rank fragments.In the example shown, this fragment ranks high for drug-likeness (CMCSubSim) andhas low penalties for size (–0.17) and complexity (–0.06).

O

ON

O

ON

N

N

N

N

O

O

O

Cl

Cl

0.67

0.67

0.67

0.63

0.62

– 0.47

– 0.41

– 0.44

– 0.40

– 0.36

F

F

FN

N

S

OO

Cl

S

Cl

Cl

Cl

ClCl

N

FF

F

Br F

N

O

FF

F

Br F

S

N

S

N

N

A B

Figure 1.6 The FBDD Score provides a robust scoring algorithm to select out moredesirable fragments. The examples shown are a selection of high (A) and low (B)scoring fragments from the ACD.



4. Implementation

4.1. X-ray primary screening library

Given our choice of X-ray crystallography as our primary screening plat-form, a non-target-biased initial screening set of 900 compounds needs tobe selected (Primary Screening Library). The initial pool of candidatecompounds is a mix of commercial sources (ACD) and our corporatecollection. In total, this entailed �2.5 million compounds. The first stepis to apply the property and substructure filtering outlined above. This initialfiltering leaves a pool of �110,000 compounds.

After the property filtering, the FBDD Score is calculated. The scoreranges from –0.47 to 1.0 for this set, with �74,000 compounds having ascore greater than zero. In practice, most of the compounds chosen for theprimary screen deck have an FBDD Score of greater than zero. However,some compounds that scored lower than 0.0 are kept. These compoundsoffer a unique functionality in terms of 3D shape. For example, somebridged and spiro systems that had an FBDD Score of approximately –0.3are included.

The final step is to use a diversity metric to select the final set of 900compounds. To ensure we cover the desired functional groups, this selec-tion is done in two steps. First, substructure filtering is used to divide the setinto six groups (Hartshorn et al., 2005):

1. Carboxylic acids2. Amines3. Amidines4. Alcohols5. Amides6. Other

Second, from each of these groups, a diverse set of fragments are chosen.This is done using the “Diverse Molecule” selection tool in Pipeline Pilot(ECFP_4 fingerprint; Accelrys, 2009). For each subgroup, �150 com-pounds are chosen to establish the final screening deck of 900 fragments.A represented set of this final selection can be found in Fig. 1.7.

4.2. Quantity and purity

Once the final selection is complete, the compounds are all ordered as neatsamples from either vendors or from our internal collection. Before enteringinto the FBDD library, quality control of purity is run on each sample toensure a purity of�95%. All compounds are stocked at�50 mg providing asupply to cover multiple years of screening.

O

N

N

N

N

N

O

OS

NO

N

NN

N

N

N

N

N

N NN

N

O

O

Cl

Cl

N

N

N

NN

N

NN

0

S

NN

O

O

OO

O

O

O

O

O

O

OO

O

O

O

O

OO

O

O

O

O

O

O

O

O

N

N

NN

Br

N

N O

F

N

F

F F

OO

OO

NN

Br

OO

FF

F

N

N

N

N

N

N N O

N NN

N

NN

N

N

NN N

N

NN

N

N

N

NN

N

N

N

OO

OO

O

OOO

O

O

OOS

SS

S

NN

Cl

N

O

O

O

O

O

F

Br

NN

F N

F

O

O O

O

O

N N

O

O

N O

N

O

O

N

N

N

NS N

NN

N

O

O

O

O

NChiral

S

N

O

O

S

N

N

N

N

S

O

Cl O

O

N N

O

N

N

O

O

O

O

S OS

O

O

N

N

N

N

O

N

N

NN

N

N

N

N

O

NO

O

O

O

O

O

O

O

O

O

N

N

N

N

N

O

O

N

OO

OO

O

N

F

F

F

O

O

O OS

O

O

Br

O

O

N

O

OO

O

O

N

N

N N

N

N

O

OO

O

O

N

N

N

Br

N

N

N

N

N

N

N

N

N

N

N

N

N N

O

O

O

S N

N

OO

O

S

N

N N

O

N

Br

N

N

O

NN

S

O

O

O

O

O

O

OO

N

N

N

N

N

N

N

Figure 1.7 (Continued).

OO

O

Br

O

O

OO

N

N

N

N

N

NO

O

Cl NN

N

N

N

O

NN

O

NS

O

O

N

N

N O

N N

O

N

N

OO

NN

O

O

O

O

S

N

N

O

OF

Cl

N

N

O O

O

N

N

N

O

O

O

N

N

NO

NN

N

F

O

O

O

Cl

O

O

O

NCl

Cl

N N

O

N

N

F

FF

O

N

N

O

O

N

OBr

O

O

N

NN

O

NN

N

N

NO

N

O

N

Br

O

N

N

NN

O

O

O

O

NN O

Cl

O

O

O

OO

N

N OON

N

N

N

N

N

Cl

N O

O O

N

N

N

O

O

N

N

N

I

O

N O

O

N

BrN

N

N

N

Figure 1.7 A representative sample of the final FBDD Primary Library.


4.3. Clustering for plating

In order to increase throughput for the crystallography studies, all theprimary library compounds are grouped into clusters of five for screening.In order to avoid the need for deconvolution after finding a hit well, thecompounds are grouped into shape-similar sets. This differs from the moreroutinely used method of grouping by dissimilarity. A detailed discussion ofthe ideas behind the shape-similar grouping choice can be found in Chapter13 (Spurlino, 2011). In brief, one of its main advantages is that it allows us toquickly determine the pharmacophore of the binding fragments.

The clustering of the compound is done using the Pipeline Pilot plat-form (Fig. 1.8). The clustering algorithm first splits the fragments intocompounds with neutral, negative, and positively charged groups. Withinthe groups, the compounds are clustered into groups of approximately fiveusing the “Cluster Molecules” tool to determine the cluster center mole-cules. Since the “Cluster Molecules” component only allows you to definean average cluster size, this step is used only to find the most dissimilarcompounds that can then be used as the seed molecules for the nextcomponent, “Equipartition Molecules.” This component finds the fourclosest molecules to the seed and groups them to form the final clusters offive. For all these steps, the following descriptors are used: Pipeline Pilotfunctional class fingerprints (path length 4), Pipeline Pilot extended con-nectivity class fingerprints (path length 4), calculated logP (AlogP; Ghoseet al., 1998), number of hydrogen bond acceptors, number of hydrogenbond donors, MDL public keys, number of atoms, and number of rings.

5. Conclusions

As outlined above, the requirements for an FBDD library in terms ofsize, and in some cases, physical properties such as solubility, differ depend-ing on the screening methods being used. For X-ray crystallography-basedscreening, throughput is low, so the library size is small. This puts tighterconstraints on how a library is chosen. In particular, it is important to selectmolecules that have no more than 15 nonhydrogen atoms, have a simplepharmacophore, and are drug-like. While very useful, “Rule of 3” typeproperty filtering does not capture enough chemical information to fullydefine this set. To get to the final selection, the FBDD Score was developed.It provides a robust scoring metric to bias fragment selections to simple, interms of pharmacophore, and drug-like compounds.

One criterion that must be met independent of the screening methodis a strict purity quality control cutoff. All compounds must be �95%pure before becoming part of the screening deck. Poor purity could lead

1

SD file Ionize

Pos Chg

Neg Chg

Clustermolecules

Clustersize=5

Clustersize=5

Clustersize=5

Clustermolecules

Clustermolecules

Clustermolecules

SD writer

Equipartitionmolecules




E

DCB

A

Figure 1.8 Pipeline Pilot protocol for creating the shape-similar wells of five compounds. The compounds are sorted by charge state (step A)and then piped into a clustering component (step B) where they are grouped into clusters of approximately five to find the cluster centers. Toenforce the fixed cluster size of five, the compounds are next piped to the equipartition component (step C). This component finds the fourclosest molecules to the seed (from step B) and groups them to form the final clusters of five. Finally, a check is done to make sure allcompounds are grouped into wells of five (steps D and E).


to false-positive hits that are close to impossible to track down. This canwaste valuable time and money. In the end, one needs a high-qualityfragment pool at the start of the FBDD process.

REFERENCES

Abad-Zapatero, C. (2007). Ligand efficiency indices for effective drug discovery. ExpertOpin. Drug Discov. 2, 469–488.

Accelrys (2009). Pipeline Pilot, 7.5. Accelrys, San Diego, CA.Barker, J., Hesterkamp, T., Schade, M., and Whittaker, M. (2007). Fragment screening:

Biochemical assays versus NMR. Innov. Pharm. Technol. 23, 19–22.Baurin, N., Aboul-Ela, F., Barril, X., Davis, B., Drysdale, M., Dymock, B., Finch, H.,

Fromont, C., Richardson, C., Simmonite, H., and Hubbard, R. E. (2004). Design andcharacterization of libraries of molecular fragments for use in NMR screening againstprotein targets. J. Chem. Inf. Comput. Sci. 44, 2157–2166.

Bembenek, S. D., Tounge, B. A., and Reynolds, C. H. (2009). Ligand efficiency andfragment-based drug discovery. Drug Discov. Today 14, 278–283.

Bemis, G. W., and Murcko, M. A. (1996). The properties of known drugs. 1. Molecularframeworks. J. Med. Chem. 39, 2887–2893.

Bemis, G. W., and Murcko, M. A. (1999). Properties of known drugs. 2. Side chains. J. Med.Chem. 42, 5095–5099.

Blaney, J., Nienaber, V., and Burley, S. K. (2006). Fragment-based lead discovery andoptimization using X-ray crystallography, computational chemistry, and high-through-put organic synthesis. Meth. Princ. Med. Chem. 34, 215–248.

Blomberg, N., Cosgrove, D. A., Kenny, P. W., and Kolmodin, K. (2009). Design ofcompound libraries for fragment screening. J. Comput. Aided Mol. Des. 23, 513–525.

Brewer, M., Ichihara, O., Kirchhoff, C., Schade, M., Whittaker, M. (2008). Assemblinga fragment library. In “Fragment-Based Drug Discovery: A Practical Approach,”(E. Zartler, M. Shapiro, eds.), pp39–62. John Wiley & Sons, Ltd, United Kingdom.

Chen, I. J., and Hubbard, R. E. (2009). Lessons for fragment library design: Analysis ofoutput from multiple screening campaigns. J. Comput. Aided Mol. Des. 23, 603–620.

Chessari, G., andWoodhead, A. J. (2009). From fragment to clinical candidate—A historicalperspective. Drug Discov. Today 14, 668–675.

Colclough, N., Hunter, A., Kenny, P. W., Kittlety, R. S., Lobedan, L., Tam, K. Y., andTimms, M. A. (2008). High throughput solubility determination with application toselection of compounds for fragment screening. Bioorg. Med. Chem. 16, 6611–6616.

Congreve, M., Carr, R., Murray, C., and Jhoti, H. (2003). A ‘rule of three’ for fragment-based lead discovery? Drug Discov. Today 8, 876–877.

Dalvit, C. (2009). NMR methods in fragment screening: Theory and a comparison withother biophysical techniques. Drug Discov. Today 14, 1051–1057.

Danielson, U. H. (2009). Fragment library screening and lead characterization using SPRbiosensors. Curr. Top. Med. Chem. (Sharjah, United Arab Emirates) 9, 1725–1735.

deKloeGerdien,E.,Bailey,D.,Leurs,R., anddeEsch Iwan, J.P. (2009).Transforming fragmentsinto candidates: Small becomes big in medicinal chemistry.Drug Discov. Today 14, 630–646.

Fischer,M., andHubbard,R.E. (2009). Fragment-based liganddiscovery.Mol. Interv.9,22–30.Ghose, A. K., Viswanadhan, V. N., and Wendoloski, J. J. (1998). Prediction of hydrophobic

(lipophilic) properties of small organic molecules using fragment methods: An analysis ofAlogP and CLogP methods. J. Phys. Chem. A 102, 3762–3772.

Hajduk, P. J., Gerfin, T., Boehlen, J.-M., Haeberli, M., Marek, D., and Fesik, S. W. (1999).High-throughput nuclear magnetic resonance-based screening. J. Med. Chem. 42,2315–2317.


Hann,M.M., Leach, A.R., andHarper, G. (2001).Molecular complexity and its impact on theprobability of finding leads for drug discovery. J. Chem. Inf. Comput. Sci. 41, 856–864.

Hartshorn, M. J., Murray, C. W., Cleasby, A., Frederickson, M., Tickle, I. J., and Jhoti, H.(2005). Fragment-based lead discovery using X-ray crystallography. J. Med. Chem. 48,403–413.

Hopkins Andrew, L., Groom Colin, R., and Alex, A. (2004). Ligand efficiency: A usefulmetric for lead selection. Drug Discov. Today 9, 430–431.

Jacoby, E., Davies, J., and Blommers, M. J. J. (2003). Design of small molecule libraries forNMR screening and other applications in drug discovery. Curr. Top. Med. Chem.(Hilversum, Netherlands) 3, 11–23.

Jhoti, H. (2007). Fragment-based drug discovery using rational design. Ernst Schering Found.Symp. Proc. 3, 169–185.

Jhoti, H., Cleasby, A., Verdonk, M., andWilliams, G. (2007). Fragment-based screening usingX-ray crystallography and NMR spectroscopy.Curr. Opin. Chem. Biol. 11, 485–493.

Liu, T., Lin, Y., Wen, X., and Jorissen, R. N. (2007). BindingDB: A web-accessibledatabase of experimentally determined protein-ligand binding affinities. Nucleic AcidsRes. 35, D198–D201.

Mayer, M., and Meyer, B. (1999). Characterization of ligand binding by saturation transferdifference NMR spectroscopy. Angew. Chem. Int. Ed. 38, 1784–1788.

Mercier, K. A., Germer, K., and Powers, R. (2006). Design and characterization of afunctional library for NMR screening against novel protein targets. Comb. Chem. HighThroughput Screening 9, 515–534.

Murray, C. W., and Rees, D. C. (2009). The rise of fragment-based drug discovery. Nat.Chem. 1, 187–192.

Nissink, J. W. M. (2009). Simple size-independent measure of ligand efficiency. J. Chem. Inf.Model. 49, 1617–1622.

Orita, M., Ohno, K., and Niimi, T. (2009a). Two golden ratio’ indices in fragment-baseddrug discovery. Drug Discov. Today 14, 321–328.

Orita, M., Warizaya, M., Amano, Y., Ohno, K., and Niimi, T. (2009b). Advances infragment-based drug discovery platforms. Expert Opin. Drug Discov. 4, 1125–1144.

Perspicace, S., Banner, D., Benz, J., Muller, F., Schlatter, D., and Huber, W. (2009).Fragment-based screening using surface plasmon resonance technology. J. Biomol. Screen.14, 337–349.

Rogers, D., and Hahn, M. (2010). Extended-connectivity fingerprints. J. Chem. Inf. Model.50, 742–754.

Schuffenhauer, A., Ruedisser, S., Marzinzik, A. L., Jahnke, W., Blommers, M., Selzer, P.,and Jacoby, E. (2005). Library design for fragment based screening. Curr. Top. Med.Chem. (Sharjah, United Arab Emirates) 5, 751–762.

Schulz, M. N., and Hubbard, R. E. (2009). Recent progress in fragment-based leaddiscovery. Curr. Opin. Pharmacol. 9, 615–621.

Spulino, J. (2011). Fragment Screening Purely with Protein Crystallography. In “FragmentBased Drug Design Tools, Practical Approaches, and Examples,” (L. Kuo, ed.), Vol. 493.Elsevier, San Diego.

Symyx Solutions, Inc. (2007a). Available Chemicals Directory. Symyx Solutions, Inc,San Diego, CA.

Symyx Solutions, Inc. (2007b). Comprehensive Medicinal Chemistry. Symyx Solutions,Inc, San Diego, CA.

Tounge, B. A., and Reynolds, C. H. (2004). Defining privileged reagents using subsimilaritycomparison. J. Chem. Inf. Comput. Sci. 44, 1810–1815.

Wang, X., Yang, Q., and You, Q. (2009). Fragment-based drug discovery. Zhongguo YaokeDaxue Xuebao 40, 289–296.

[Methods in Enzymology] Fragment-Based Drug Design - Tools, Practical Approaches, and Examples...

Documents

Transcript of [Methods in Enzymology] Fragment-Based Drug Design - Tools, Practical Approaches, and Examples...