Development and experimental validation of a docking strategy for the generation of kinase-targeted libraries

Article abstract of DOI:10.1021/jm701367r

A high-throughput docking strategy for the filtering of in silico compounds and the generation of kinase-targeted libraries is described. Systematic docking and scoring in three kinase crystal 3D structures of 123 structurally diverse kinase ligands led to the determination of six thresholds for each kinase. These thresholds were used as filters for the virtual screening of two collections of compounds: a collection of more than 2500 drugs and drug-like compounds (negative control) and a kinase-targeted library of 1440 compounds. This strategy was then experimentally validated by testing 60 compounds from the kinase-targeted library on 41 kinases from five different families. The 60 compounds were split into those passing all the thresholds and the others (30 compounds in each group). The overall hit enrichment was 6.70-fold higher in the first group, validating our approach for the generation of kinase-targeted libraries and the identification of scaffolds with high kinase inhibitory potential.

Full text of DOI:10.1021/jm701367r

3124
J. Med. Chem. 2008, 51, 3124–3132
Development and Experimental Validation of a Docking Strategy for the Generation of
Kinase-Targeted Libraries
Rafael Gozalbes,*,† Laurence Simon,‡ Nicolas Froloff,† Eric Sartori,† Claude Monteils,† and Romuald Baudelle†
Cerep, 19 aVenue du Que´bec, Courtaboeuf-1, Villebon sur YVette, 91951 Courtaboeuf Cedex, France, and
Cerep, Le bois l’EVeˆque, 86600 Celle l’EVescault, France
ReceiVed October 31, 2007
A high-throughput docking strategy for the filtering of in silico compounds and the generation of kinase-
targeted libraries is described. Systematic docking and scoring in three kinase crystal 3D structures of 123
structurally diverse kinase ligands led to the determination of six thresholds for each kinase. These thresholds
were used as filters for the virtual screening of two collections of compounds: a collection of more than
2500 drugs and drug-like compounds (negative control) and a kinase-targeted library of 1440 compounds.
This strategy was then experimentally validated by testing 60 compounds from the kinase-targeted library
on 41 kinases from five different families. The 60 compounds were split into those passing all the thresholds
and the others (30 compounds in each group). The overall hit enrichment was 6.70-fold higher in the first
group, validating our approach for the generation of kinase-targeted libraries and the identification of scaffolds
with high kinase inhibitory potential.
Introduction
fold consisting of an N-terminal and C-terminal domain
connected through a short strand (the “hinge region”), with
adenosine triphosphate (ATP) binding in the cleft between the
two domains. Even when there is a low overall sequence identity
in the ATP binding sites, such as for the tyrosine kinase Src
and the serine threonine kinase CDK2,a the backbone structure
is very similar.10 It is then possible to identify a common pattern
for most known kinase inhibitors, which consists of a heteroaro-
matic core including a hydrogen bond acceptor that interacts
with the backbone NH of the “hinge” region and a variable
hydrophobic pocket responsible for selectivity.
Intense activity of structural biologists in the field of kinases
has given access to hundreds of crystal structures available for
docking studies. Different approaches have been used for the
identification of new promiscuous kinase scaffolds leading to
the synthesis of kinase-targeted libraries, particularly structure
and pharmacophore-based approaches. However, most published
examples describe libraries designed for a single target and are
not suited for the generation of class targeted libraries.7a,11
We describe here the development of a high-throughput
docking and scoring strategy adapted to the filtering of in silico
compounds and the subsequent generation of kinase targeted
libraries.
Phosphorylation of proteins, catalyzed by a kinase, was
discovered in 1954 by Burnett et al.,1 and since then, there has
been strong interest in the role of protein phosphorylation in
regulating protein function. Protein kinases play a crucial role
in cellular proliferation, differentiation, or inflammation.2 In
addition, the most common regulatory mechanisms of protein
kinases include protein localization, ligand-coupled allosteric
activation or inhibition, and reversible conformational changes
at the catalytic site that are controlled by phosphorylation or
dephosphorylation.3 Their improper activation or inhibition can
be highly disruptive to the cell and be a major cause of cancer,
and as a consequence, protein kinases have emerged as an
extremely important set of targets for drug development.4
Initially, a large number of natural compounds were found to
be rather potent inhibitors of protein tyrosine kinases. Although
many showed initial promise, they were found to be highly
promiscuous and toxic;5 therefore, many efforts have been
devoted to the discovery of more specific compounds. Develop-
ment and approval of imatinib (for chronic myeloid leukemia)
and of gefitinib and erlotinib (for nonsmall cell lung cancer)
have provided proof-of-principle that small molecule kinase
inhibitors can be effective drugs.6
Combinatorial chemistry has now gained general acceptance
as a tool to speed up drug discovery. To increase its efficiency,
interest in library design has been recently shifted toward the
generation of target-class focused libraries.7 The possibility to
design kinase targeted libraries is based on the existence of
numerous docking data and on common structural features
present in most kinase inhibitors.8
The first published crystal structure of a protein kinase was
the catalytic site domain of the cyclic-AMP-dependent kinase.9
All kinases with known structures adopt essentially the same
Material and Methods
Crystal Structures of the Receptors. Three kinase crystal 3D
structures cocrystallized with a kinase inhibitor in the ATP site
were selected from the Protein Data Bank (PDB)12 for docking
(Figure 1):
a Abbreviations: Abl, Abelson tyrosine kinase; AGC, protein kinases A,
G, and C (cyclic nucleotide regulated and phospholipid-regulated kinases
and ribosomal S6 kinases); BSA, bovine serum albumin; CaMK, Ca2+
/
calmodulin-dependent kinases; CDK, cyclin dependent kinases; CGMC,
cyclin-dependent and mitogen-activated kinases; CTK, cytosolic tyrosine
kinases; DIEA, N,N-diisopropylethylamine; DTT, 1,4-dithiothreitol; EDTA,
ethylene-diamine-tetracetic acid; EGFR, epidermal growth factor receptor
kinase; Fak, focal adhesion kinase; HTRF, homogeneous time-resolved
fluorescence assays; Lck, lymphocyte-specific protein tyrosine kinase; PDB,
Protein Data Bank; PDGFR, platelet derived growth factor receptor; PLP,
piecewise linear potential; PMF, potential of mean force; RTK, receptor
tyrosin kinases; Src, sarcoma tyrosine kinase; TBTU, O-(benzotriazol-1-
yl)-N,N,N′,N′-tetramethyluronium tetrafluoroborate; THF, tetrahydrofuran.
* To whom correspondence should be addressed. Current address:
Structural Biology Laboratory, Medicinal Chemistry Department, Centro
de Investigacio´n Pr´ıncipe Felipe (CIPF), Avda. Autopista del Saler 16, 46012
VALENCIA, Spain. Phone: +34 963289681. Fax: +34 963289701. E-mail:
rgozalbes@cipf.es.
†
Cerep, Courtaboeuf Cedex, France.
Cerep, Celle l’Evescault, France.
‡
10.1021/jm701367r CCC: $40.75
2008 American Chemical Society
Published on Web 05/15/2008

Docking Strategy for Kinase-Targeted Libraries
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 11 3125
Figure 1. Kinase inhibitors cocrystallized with the three kinase crystal 3D receptors selected from the Protein Data Bank.
Figure 2. Some examples of diaminopyrimidine and related kinase inhibitors.
(a)Abelson tyrosine kinase (ABL). PDB-code: 1FPU. Cocrys-
tallized ligand: N-[4-methyl-3-[(4-pyridin-3-ylpyrimidin-2-yl)ami-
no]phenyl]pyridine-3-carboxamide (a variant of imatinib).13
(b)Epidermal growth factor receptor kinase (EGFR). PDB-code:
1M17. Cocrystallized ligand: [6,7-bis(2-methoxy-ethoxy)quinazo-
line-4-yl]-(3-ethynylphenyl)amine (erlotinib).14
(a) A set of 123 old and recent kinase inhibitors described in
publications and patents (see Supporting Information) and selected
to exemplify a large variety of chemotypes which served to
determine the best scoring function values.
(b) The BioPrint collection,17 containing more than 2500
marketed drugs, reference compounds, and compounds that have
failed in clinical trials which were selected for the purpose of our
BioPrint pharmaco-informatics platform.17 The hit rate of BioPrint
compounds on a panel of kinases (inhibition >40% @ 10 µM) is
around 2% (unpublished experimental results). This collection was
used as a negative control for our molecular modeling study.
(c)Cyclin dependent kinase 2 (CDK2). PDB-code: 1CKP. Coc-
rystallized ligand: purvalanol B.15
These three kinases are validated major therapeutic targets,
belonging to three different families according to a generally
accepted classification derived from the “kinome” analysis by
Manning16 cytosolic tyrosine kinase (CTK) for Abl, receptor
Tyrosine kinase (RTK) for EGFR, cyclin-dependent and mitogen-
activated kinase (CMGC) for CDK2.
Collections of Compounds. Our strategy was based on the
systematic docking of three sets of compounds in three kinase
crystal 3D structures:
(c) A kinase-targeted library of 1440 compounds synthesized
in-house based around 27 original scaffolds. This library was
designed around a diaminopyrimidine central core, which can be
considered as a privileged and a general structure for interacting
with the ATP binding site of kinases. Some examples of diami-
nopyrimidines and related structures reported as kinase inhibitors

3126 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 11
Gozalbes et al.
Table 1. (a) For Each Known Ligand of a Kinase, the Highest Scoring Value (Including All Conformations) Is Selected for Each Scoring Function,a
and (b) For Each Kinase (ABL, CDK2, EGFR) and for Each Scoring Function, a Threshold Is Chosen in a Way That All Known Ligands of a Kinase
Have a Score above This Threshold
a This value becomes the score for each ligand and for each scoring function.
are shown in Figure 2: N-aryl-2,4-diaminopyrimidine derivatives,
which have been described as inhibitors of Lck kinase18a and Fak
(focal adhesion kinase);18b N-aryl-4,6-diaminopyrimidine deriva-
tives, which have been reported to inhibit EGFR and PDGFR
kinases,18c Rho kinase,18d and src and abl kinases;18e and diami-
nopurine derivatives such as Purvalanol A, which are inhibitors of
cyclin dependent kinases (CDKs).19
The R-groups for the 27 diaminopyrimidine scaffolds were
selected taking into account some of the most common groups
described in scientific literature to be present in inhibitors of kinases,
and the R-groups for partners were selected in order to respect
Lipinski rules,20 particularly concerning molecular weight.
Docking and Scoring Approach. A systematic docking of the
three above-mentioned collections was performed with the three
kinase structures extracted from the PDB by using the module
“LigandFit” from Cerius2 (v. 4.10.).21 This program is designed
to dock molecules into a protein binding site. For each kinase, the
residues around the cocrystallized ligands were selected, the
cocrystallized ligands and water molecules were deleted, and
the hydrogens of the site residues were optimized. For each of the
123 kinase inhibitors, 10 conformations were generated. They were
selected for their adaptation to the shape of the site and then oriented
into the binding site, and the energy was optimized. During the
docking, the protein is rigid while the ligand remains flexible,
allowing different conformations to be searched and docked. For
each pose, six scoring functions implemented in LigandFit (PLP1,
PLP2, LigScore1, LigScore2, PMF, Jain)21 were calculated. For
the known ligands of a kinase, the highest score of the six functions
were saved. The lowest value of each function was then chosen as
the threshold for this kinase (see Table 1). Following the same
procedure for the three kinases, the 18 thresholds (six scoring
functions per kinase) were then used as filters for the docking of
BioPrint and the kinase-targeted library. A molecule was considered
as predicted active on a kinase if the six scoring values were
superior to the six thresholds.
Kinase Assays. To assay inhibitor activities on 41 kinases
(selected from Cerep kinase platform so as to sample the diversity
of kinase targets), homogenous time-resolved fluorescence (HTRF)
assays were performed in 96 half-well plates. The following were
combined in the reaction mixture: 4 µL of inhibitor (diluted in 100%
DMSO at 50 µM), 8 µL of enzyme, and 8 µL of ATP/substrate
mix (final concentrations around Km values). Enzymes, ATP, and
substrate were previously diluted in kinase buffer to get final
concentrations in the well: 50 mM Hepes/NaOH (pH 7.4), 40 µM
Na3VO4, 0.005% Tween-20, 1 mM DTT, and various concentrations
of MnCl2 or MgCl2. The assay was initiated by the addition of the
ATP/substrate mix. The reaction was incubated for different times
at room temperature. The reaction was stopped by addition of 55

Docking Strategy for Kinase-Targeted Libraries
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 11 3127
Scheme 1. List of the 27 Building Blocks
Scheme 2. Synthesis of Diaminopyrimidine Scaffoldsa
µL of revelation mix. The revelation mix was composed by EDTA
(0.33 M), specific europium-cryptate antibody antiphospho-
substrate, and XL-665 with various tags (CisBio International). The
mix was prepared in revelation buffer with final concentrations in
the well of 50 mM Hepes/NaOH (pH 7.4), 0.1% BSA, and 150
mM KF. The reaction was incubated for 1 h or more at room
temperature and was read with Rubystar (BMG, Germany).
Kinase activities were expressed as a percentage of maximal
activity (without inhibitor), resulting from n ) 2 experiments.
Chemistry. Twenty-seven pyrimidine building blocks were
selected and synthesized for the design of the kinase targeted library
(Scheme 1).
The three main series (building blocks 1a-g, 2a-h, and 3a-h)
were obtained in three steps as represented in Scheme 2 from the
appropriate dichloropyrimidine and sequential reaction with two
electrophiles, one of them bearing a methyl ester or a Boc-amine
to generate a functionalized building block. Using conditions well
described in literature,22 the selective substitution of the halides
was achieved for the three scaffolds. In the second series (starting
from 2,6-dichloropyrimidine), phenol was also used as a nucleophile
for the first step. The protecting groups of the building blocks were
then removed by hydrolysis with sodium hydroxide (for methyl
esters) or treatment with methanolic hydrochloric acid (for Boc
protection) to afford the 18 acid and the 5 secondary amine building
blocks.
a Reagents: (a) appropriate amine or phenol, butan-1-ol, 120 °C; (b)
appropriate amine, butan-1-ol, 120 °C; (c) NaOH, MeOH; and (d) HCl,
MeOH.
mamidine, followed by chlorination with POCl3 and substitution
with the appropriate amine, and then catalytic hydrogenolysis or
Boc-deprotection gave two secondary amine building blocks for
the library.
Compound 5 was obtained using a Suzuki coupling on the
substituted chloropyrazine 11 as depicted in Scheme 4.24 Finally,
compound 6 was obtained by first condensing 2-pyridylacetonitrile
on 2,6-dichloropyrimidine to obtain 12.25 Halide substitution with
excess piperazine gave the desired molecule.
Four other scaffolds (4a, 4b, 5, and 6) were also included in the
library to increase its chemical diversity and its interest for the
discovery of new kinase inhibitors. Compounds 4a and 4b were
obtained in four steps starting from commercially available 1-ben-
zyl-3-carbethoxy-4-piperidone (Scheme 3). Cyclization with for-
For the library synthesis, each acid building block was reacted
with a diverse set of 80 amine partners in classical conditions

3128 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 11
Scheme 3. 23 Synthesis of 4a and 4ba
Gozalbes et al.
a Reagents: (a) formamidine, NaOEt, EtOH, reflux; (b) POCl3, reflux (c) meta-anisidine, propanol, reflux; (d) H2, Pd/C, MeOH; (e) Boc-piperazine,
propanol, reflux; and (f) HCl, MeOH.
Scheme 4. Synthesis of 5 and 6a
As expected, the virtual hit rate is clearly higher for the
diaminopyrimidine library when compared to BioPrint. Fur-
thermore, this ratio is higher when the number of kinases
considered increases: 6.6 for one kinase (88.5/13.5), 18.8 for
two kinases (62.2/3.3), and 65.7 for the three kinases (26.3/
0.4). For Bioprint, 0.4% of the compounds are able to pass all
the filters for the three kinases. For the diaminopyrimidine
library, more than 88% of the compounds were successfully
docked in at least one kinase, but only 26.3% of the library
a Reagents: (a) ethanol, reflux; (b) 3-carboxyphenylboronic acid, Na2CO3,
Pd(P(Ph)3)4, ethanol/toluene, reflux; (c) NaNH2, THF; (d) piperazine,
propanol, reflux.
was docked in the three kinases. We observed that the docking
results were related to the building blocks (for example, 81%
of the compounds built around scaffold 2e were docked in the
three kinases, versus only 5% for 1g, scheme 1) but also to the
partners reacted with the building blocks to generate the final
compounds of the library. Even for kinase-oriented building
blocks, the choice of the partners is crucial to obtain final
compounds with kinase activities, and we expect our model to
be able to correctly select them. For the experimental validation,
in order to increase the hit rate and decrease the number of
false positives, we decided to select the more drastic criterion
and to consider as predicted active the compounds passing all
the filters for the three kinases.
Experimental Validation. Sixty compounds from the di-
aminopyrimidine library were selected according to their
docking scores: 30 compounds that passed the three filters were
Figure 3. Percentages of compounds from each collection predicted
to hit at least one, two or the three kinases. The BioPrint compounds
are the black bars, and kinase-targeted compounds are white.
considered as predicted active, and 30 not passing the three
filters were considered as predicted inactive. Compounds for
experimental validation were selected (1) so as to represent all
building blocks in the library, (2) so the number of compounds
related to each building block was proportional to the number
of these compounds predicted as active or inactive respectively,
and (3) so the final compounds (building blocks coupled to a
reagent) were randomly selected. The selection of compounds
for the experimental screening was done randomly to avoid any
bias in the selection. As a consequence, the selected compounds
represent all the families included in the library (see Table 3 in
the Supporting Information with the 60 structures of the
compounds included in the screening). Experimental percentages
of inhibition at 10 µM (n ) 2) were determined for the 60
compounds on 41 kinases from five different families including
the families related to ABL, CDK2, and EGFR (Table 2). By
all means, the proportion of hits was found to be clearly much
larger in the predicted active compounds group: the number of
points of activity (percentage of inhibition superior to 40%) was
108 for the predicted active group against 22 for the predicted
inactive group. Twenty compounds (67%) did inhibit at least
(TBTU, DIEA, DMF) to obtain the amides. Each amine scaffold
was reacted with a diverse set of 80 acid partners in the same
conditions and was also reacted with a diverse set of 40 halides in
DMF in the presence of DIEA at 90 °C to obtain alkylated
compounds. The compounds were extracted in ethyl acetate/Na2CO3
to remove salts and excess acidic reagents. All the compounds were
analyzed by LC/MS. At this step, only the compounds with purities
above 85% were kept for the final library. The compounds with
insufficient purities were systematically purified by preparative LC/
MS. The Tanimoto index26 for this library using Isis keys as
descriptors27 was 0.61 which can be considered as indicative of a
relatively high chemical diversity considering the limited number
of building blocks.
Results
Docking of BioPrint and the Kinase-Focused Library.
Figure 3 shows the percentage of compounds from each
collection predicted to bind at least one, two, or the three kinases.

Docking Strategy for Kinase-Targeted Libraries
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 11 3129

3130 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 11
Gozalbes et al.
in the CDK2 assay. Nevertheless, we found to be very significant
the fact that several active-predicted compounds showed activity
to other targets of the CGMC family to which CDK2 belongs.
This result is in perfect accordance with our initial objective,
which was to implement a model able to identify some kinase-
family related features not specific to one particular kinase.
Seventeen compounds of the predicted active group inhibited
more than one kinase, compared to only one compound in the
predicted inactive group. The case of this compound, (3-[4-
(benzothiazol-6-ylamino)-pyrimidin-2-ylamino]-N-[(2-tetrahy-
drofuryl)-methyl]-benzamide (CER0262730), responsible for 20
of the 22 false negative results, was further investigated. While
checking the six individual docking scores for each of the three
kinases, it appeared that the molecule actually passed 15 out of
the 18 filters (six for ABL, four for EGFR, and five for CDK2).
Its experimental activity appears less surprising and reveals the
stringent requirements of our model and the limits of a rapid
docking strategy, since no deep analysis of the docking poses
is done. Obviously the filters can be used in a less drastic manner
(for example, a docking score superior to thresholds values for
more than 9 out of the 18 scoring functions).
Figure 4. Experimental validation: hit rates by kinase family when
comparing the compounds predicted as active (white) by the docking-
scoring approach and those predicted as inactive (black).
one kinase of the panel (more than 40% inhibition at 10 µM),
compared to three compounds (10%) for the inactive group.
Our docking model was then able to generate a 6.7-fold
enrichment, even when applied to a collection of closely
structurally related compounds such as the diaminopyridine
library. We found at least one hit for 28 kinases of the panel
(68%) and for 21 kinases of the CTK, RTK, and CGMC families
(85%). Interestingly, though our docking approach was based
on these three families, some hits were also found for kinases
belonging to other families (AGC and CaMK; Figure 4).
Our kinase-targeted library does not include purine structures
similar to those of selective inhibitors of CDK2 like olomucin
or roscovitine (purvalanol-type compounds). In consequence,
it is not surprising that no activity was seen for any compound
Our docking strategy allowed the identification of several
promiscuous inhibitors interacting with more than 10 kinases
from the panel but also of relatively selective molecules with
significant inhibition for 3 kinases or less, and it was observed
that small structural modifications around a common core lead
to considerable differences in selectivity (Figure 5). However,
the screening concentration (10 µM) was adapted to hit seeking
Figure 5. Promiscuous (more than 40% inhibition at 10 µM on more than 10 kinases) and selective (more than 40% inhibition at 10 µM on less
than 3 kinases) inhibitors.

Docking Strategy for Kinase-Targeted Libraries
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 11 3131
Expert ReV. Anticancer Ther. 2002, 2, 161–168. (c) Smith, J. Erlotinib:
small-molecule targeted therapy in the treatment of non-small-cell lung
cancer. Clin. Ther. 2005, 27, 101513–101534.
and not to specific lead finding so that promiscuity is certainly
overestimated.
(7) (a) Lowrie, J. F.; Delisle, R. K.; Hobbs, D. W.; Diller, D. J. The
different strategies for designing GPCR and kinase targeted libraries.
Comb. Chem. High Throughput Screening 2004, 7, 495–510. (b)
Schnur, D.; Beno, B. R.; Good, A.; Tebben, A. Approaches to target
class combinatorial library design. Methods Mol. Biol. 2004, 275, 355–
378.
(8) (a) Trumpp-Kallmeyer, S.; Rubin, J. R.; Humblet, C.; Hamby, J. M.;
Hollis Showalter, H. D. Development of a binding model to protein
tyrosine kinases for substituted pyrido[2,3-d]pyrimidine inhibitors.
J. Med. Chem. 1998, 41, 1752–1763. (b) Diller, D. J.; Li, R. Kinases,
homology models, and high throughput docking. J. Med. Chem. 2003,
46, 4638–4647. (c) Naumann, T; Matter, H. Structural classification
of protein kinases using 3D molecular interaction field analysis of
their ligand binding sites: target family landscapes. J. Med. Chem.
2002, 45, 2366–2378.
(9) (a) Zheng, J.; Knighton, D. R.; Ten Eyck, L. F.; Karlsson, R.; Xuong,
N.; Taylor, S. S.; Sowadski, J. M. Crystal structure of the catalytic
subunit of cAMP-dependent protein kinase complexed with MgATP
and peptide inhibitor. Biochemistry 1993, 32, 2154–2161. (b) Knighton,
D. R.; Zheng, J. H.; Ten Eyck, L. F.; Ashford, V. A.; Xuong, N. H.;
Taylor, S. S.; Sowadski, J. M. Crystal structure of the catalytic subunit
of cyclic adenosine monophosphate-dependent protein kinase. Science
1991, 253, 407–414.
(10) Cheek, S.; Ginalski, K.; Zhang, H.; Grishin, N. V. A comprehensive
update of the sequence and structure classification of kinases. BMC
Struct. Biol. 2005, 5, 1–19.
(11) Honma, T.; Yoshizumi, T.; Hashimoto, N.; Hayashi, K.; Kawanishi,
N.; Fukasawa, K.; Takaki, T.; Ikeura, C.; Ikuta, M.; Suzuki-Takahashi,
I.; Hayama, T.; Nishimura, S.; Morishima, H. A novel approach for
the development of selective Cdk4 inhibitors: library design based on
locations of Cdk4 specific amino acid residues. J. Med. Chem. 2001,
44, 4628–4640.
(12) Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T. N.;
Weissig, H.; Shindyalov, I. N.; Bourne, P. E. The Protein Data Bank.
Nucleic Acids Res. 2000, 28, 235–242.
(13) Schindler, T.; Bornmann, W.; Pellicena, P.; Miller, W. T.; Clarkson,
B.; Kuriyan, J. Structural mechanism for STI-571 inhibition of Abelson
Tyrosine Kinase. Science 2000, 289, 1938–1942.
Conclusion
Our results confirm the feasibility of designing kinase targeted
libraries by selecting privileged heteroaromatic central cores and
taking advantage of the diversity generated by combinatorial
chemistry to explore the hydrophobic pocket and fine-tune
selectivity and potency. The method we have developed for
kinase binding prediction is based on an original and rapid
docking and scoring strategy using three different kinases and
is particularly adapted for library design and virtual screening.
This approach has shown its ability to identify promiscuous
kinase targeted cores which, when subjected to combinatorial
variations, can generate more or less selective hits for a large
variety of kinases. It has been validated by an experimental HTS
assay and could be applied to future libraries with a central core
being a privileged structure for interacting with the kinase ATP
site. Thanks to the rapidity and simplicity of the method,
thousands of virtual compounds can be filtered, allowing an
optimized selection of central heteroaromatic cores and of
hydrophobic partners which will increase the hit rate of kinase-
targeted libraries.
This efficient docking and scoring approach may easily be
modified or extended by selecting other crystallized kinases for
docking, and may be adapted to less drastic selections by, for
example, picking up compounds passing only some of the filters.
Supporting Information Available: Set of 123 kinase inhibitors
described in publications and patents which served to determine
the best scoring function values; NMR spectroscopy of the 27
building blocks that were selected and synthesized for the design
of the kinase targeted library; and structures of the 60 compounds
from the diaminopyrimidine library that were assayed against 41
kinases. This material is available free of charge via the Internet at
http://pubs.acs.org.
(14) Stamos, J.; Sliwkowski, M. X.; Eigenbrot, C. Structure of the
Epidermal Growth Factor Receptor kinase domain alone and in
complex with a 4-anilinoquinazoline inhibitor. J. Biol. Chem. 2002,
277, 46265–46272.
(15) Gray, N. S.; Wodicka, L.; Thunnissen, A. W. H.; Norman, T. C.;
Kwon, S.; Hernan Espinoza, F.; Morgan, D. O.; Barnes, G.; LeClerc,
S.; Meijer, L.; Kim, S-H.; Lockhart, D. J.; Schultz, P. G. Exploiting
chemical libraries, structure, and genomics in the search for kinase
inhibitors. Science 1998, 281, 533–538.
References
(1) Burnett, G.; Kennedy, E. P. The enzymatic phosphorylation of proteins.
J. Biol. Chem. 1954, 211, 969–980.
(2) (a) Levitzki, A.; Gazit, A. Tyrosine kinase inhibition: an approach to
drug development. Science 1995, 267, 1782–1788. (b) Traxler, P.
Tyrosine kinases as targets in cancer therapy - successes and failures.
Expert. Opin. Ther. Targets. 2003, 2, 215–34.
(16) Manning, G; Whyte, D. B.; Martinez, R.; Hunter, T.; Sudarsanam, S.
The protein kinase complement of the human genome. Science 2002,
298, 1912–1934.
(17) (a) Krejsa, C. M.; Horvath, D.; Rogalski, S. L.; Penzotti, J. E.; Mao,
B.; Barbosa, F.; Migeon, J. C. Predicting ADME properties and side
effects: The BioPrint approach. Curr. Opin. Drug. DiscoV. DeVel. 2003,
6, 470–480. (b) Froloff, N.; Hamon, V.; Dupuis, P.; Otto-Bruc, A.;
Mao, B.; Merrick, S.; Migeon, J. In: Jacoby, E. (Ed.), Chemogenomics
Knowledge-based Approaches to Drug DiscoVery, Imperial College
Press, London, 2006, pp. 175-206. (c) http://www.cerep.fr/cerep/users/
pages/ProductsServices/bioprintservices.asp.
(18) (a) Blumenkopf, T.; Mueller, E.; Roskamp, E. WO 0140215 (Pfizer,
2001). (b) Kath, J. C.; Luzzio, M. J. WO 04056786 (Pfizer, 2004) (c)
Thomas, A. P. WO 9515952 (Zeneca, 1995). (d) Feurer, A.; Bennabi,
S.; Heckroth, H.; Ergueden, J.; Schenke, T.; Bauser, M.; Kast, R.;
Stasch, J. P.; Stahl, E.; Muenter, K.; Lang, D.; Ehmke, H. WO
03106450 (Bayer, 2003). (e) Luk, K. C.; Rossman, P. L.; Scheiblich,
S.; So, S. S. U.S. Patent 7129351 (Hoffmann-La Roche Inc., 2004).
(19) (a) Haesslein, J. L.; Jullian, N. Recent Advances in Cyclin-Dependent
Kinase Inhibition. Purine-Based Derivatives as Anti-Cancer Agents.
Roles and Perspectives for the Future. Curr. Top. Med. Chem. 2002,
1, 1037–1050. (b) Breault, G. A.; Ellston, R. P. A.; Green, S.; James,
S. R.; Jewsbury, P. J.; Midgley, C. J.; Pauptit, R. A.; Minshull, C. A.;
Tucker, J. A.; Pease, J. E. Cyclin-dependent kinase 4 inhibitors as a
treatment for cancer. Part 2: identification and optimisation of
substituted 2,4-bis anilino pyrimidines. Bioorg. Med. Chem. Lett. 2003,
13, 2961–2966.
(3) (a) Gold, M. G.; Barford, D.; Komander, D. Lining the pockets of
kinases and phosphatases. Curr. Opin. Struct. Biol. 2006, 6, 693–
701. (b) Martin, B. T.; Strebhardt, K. Polo-like kinase 1: target and
regulator of transcriptional control. Cell Cycle 2006, 24, 2881–2885.
(c) Cowan-Jacob, S. W. Structural biology of protein tyrosine kinases.
Cell. Mol. Life Sci. 2006, 22, 2608–2625.
(4) (a) Druker, B. J.; Lydon, N. B. Lessons learned from the development
of an Abl tyrosine kinase inhibitor for chronic myelogenous leukaemia.
J. Clin. InVest. 2000, 105, 3–7. (b) Drews, J. Drug discovery: a
historical perspective. Science 2000, 287, 1960–1964. (c) Haluska,
P.; Adjei, A. A. Receptor tyrosine kinase inhibitors. Curr. Opin. InVest.
Drugs. 2001, 2, 280–286. (d) Fabbro, D.; Ruetz, S.; Buchdunger, E.;
Cowan-Jacob, S. W.; Fendrich, G.; Liebetanz, J.; Mestan, J.; O’Reilly,
T.; Traxler, P.; Chaudhuri, B.; Fretz, H.; Zimmermann, J.; Meyer, T.;
Caravatti, G.; Furet, P.; Manley, P. W. Protein kinases as targets for
anticancer agents: from inhibitors to useful drugs. Pharmacol. Ther.
2002, 93, 79–98.
(5) Fabian, M. A.; Biggs, W. H, 3rd; Treiber, D. K.; Atteridge, C. E.;
Azimioara, M. D.; Benedetti, M. G.; Carter, T. A.; Ciceri, P.; Edeen,
P. T.; Floyd, M.; Ford, J. M.; Galvin, M.; Gerlach, J. L.; Grotzfeld,
R. M.; Herrgard, S.; Insko, D. E.; Insko, M. A.; Lai, A. G.; Lelias,
J. M.; Mehta, S. A.; Milanov, Z. V.; Velasco, A. M.; Wodicka, L. M.;
Patel, H. K.; Zarrinkar, P. P.; Lockhart, D. J. A small molecule-kinase
interaction map for clinical kinase inhibitors. Nat. Biotechnol. 2005,
23, 29–36.
(6) (a) Capdeville, R.; Buchdunger, E.; Zimmermann, J.; Matter, A. Glivec
(STI571, imatinib), a rationally developed, targeted anticancer drug.
Nat. ReV. Drug DiscoVery 2002, 1, 493–502. (b) Ranson, M.; Mansoor,
W.; Jayson, G. ZD1839 (IRESSA): a selective EGFR-TK inhibitor.
(20) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J.
Experimental and computational approaches to estimate solubility and
permeability in drug discovery and development settings. AdV. Drug
DeliVery ReV. 2001, 46, 3–26.
(21) Cerius2, version 4.10; Accelrys Software Inc, San Diego, CA, USA,
2004.

3132 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 11
Gozalbes et al.
(22) (a) Beattie, J. F.; Breault, G. A.; Ellston, R. P. A.; Green, S.; Jewsbury,
P. J.; Midgley, C. J.; Naven, R. T.; Minshull, C. A.; Pauptit, R. A.;
Tucker, J. A.; Pease, J. E. Cyclin-dependant kinase 4 inhibitors as a
treatment for cancer. Part 1: Identification and optimisation of
substituted 4,6-bis anilino pyrimidines. Bioorg. Med. Chem. 2003, 13,
2955–2960. (b) Breault, G. A.; Ellston, R. P. A.; Green, S.; James,
S. R.; Jewsbury, P. J.; Midgley, C. J.; Pauptit, R. A.; Minshull, C. A.;
Tucker, J. A.; Pease, J. E. Cyclin-dependant kinase 4 inhibitors as a
treatment for cancer. Part 2: Identification and optimisation of
substituted 2,4-bis anilino pyrimidines. Bioorg. Med. Chem. 2003, 13,
2955–2960. (c) Aoyagi, T.; Yanada, R.; Bessho, K.; Yoneda, F.
Synthesis of 2-amino-2-deoxy-5-deazaflavins and related compounds.
J. Heterocyclic Chem. 1991, 28, 1537–1539.
5,6-polymethylenepyrimidines. Eur. J. Med. Chem. Chim. Ther. 1982,
17, 75–80.
(24) Maes, B. U. W.; Lemie`re, G. L. F.; Dommisse, R.; Augustyns, K.;
Haemers, A. A new approach towards the synthesis of 3-amino-6-
(hetero)arylpyridazines based on palladium catalyzed cross-coupling
reactions. Tetrahedron 2000, 56, 1777–1781.
(25) Bernhart, C. A.; Condamines, C.; Demarne, H.; Roncucci, R.; Gagnol,
J. P.; Gautier, P. J.; Serre, M. Synthesis and antiarrhythmic activity
of new [(dialkylamino)alkyl]pyridylacetamides. J. Med. Chem. 1983,
26, 451–455.
(26) Flower, D. R. J. On the properties of bit string-based measures of
chemical similarity. J. Chem. Inf. Comput. Sci. 1998, 38, 379–386.
(27) MDL ISIS/HOST; MDL Information Systems, Inc., San Ramo´n, CA,
USA, 1991.
(23) Sekiya, T.; Hiranuma, H.; Kanayama, T.; Hata, S. Pyrimidine
derivatives III (1): synthesis of hypoglycemic 4-alkoxy-2-piperazino-
JM701367R