Multidimensional de novo design reveals 5-HT2B receptor-selective ligands

Article abstract of DOI:10.1002/anie.201410201

We report a multi-objective de novo design study driven by synthetic tractability and aimed at the prioritization of computer-generated 5-HT_2B receptor ligands with accurately predicted target-binding affinities. Relying on quantitative bioactivity models we designed and synthesized structurally novel, selective, nanomolar, and ligand-efficient 5-HT_2B modulators with sustained cell-based effects. Our results suggest that seamless amalgamation of computational activity prediction and molecular design with microfluidics-assisted synthesis enables the swift generation of small molecules with the desired polypharmacology.

Full text of DOI:10.1002/anie.201410201

Angewandte
Chemie
DOI: 10.1002/anie.201410201
Drug discovery
Multidimensional De Novo Design Reveals 5-HT2B Receptor-Selective
Ligands**
Tiago Rodrigues, Nadine Hauser, Daniel Reker, Michael Reutlinger, Tiffany Wunderlin,
Jacques Hamon, Guido Koch, and Gisbert Schneider*
Abstract: We report a multi-objective de novo design study
driven by synthetic tractability and aimed at the prioritization
of computer-generated 5-HT2B receptor ligands with accurately
predicted target-binding affinities. Relying on quantitative
bioactivity models we designed and synthesized structurally
novel, selective, nanomolar, and ligand-efficient 5-HT2B mod-
ulators with sustained cell-based effects. Our results suggest
that seamless amalgamation of computational activity predic-
tion and molecular design with microfluidics-assisted synthesis
enables the swift generation of small molecules with the desired
polypharmacology.
mation of such computational tools with the automated
synthesis of designer molecules may be suitable for rapidly
obtaining NCEs with the desired target engagement.
We focused on the 5-HT2B receptor as a challenge for this
proof-of-concept study. Antagonistic small-molecular effec-
tors may serve as probes to study the pathophysiology of
migraine, pulmonary hypertension, and chronic heart fail-
ure.[5] On the other hand, engagement of 5-HT2B receptors by
small molecule agonists results in cardiac valvulopathy,
a severe adverse drug reaction that leads to chemotype de-
prioritization.[6] Consequently, reports of selective ligands
that may yield therapeutic benefits are rare, and clinical trials
with 5-HT2B receptor antagonist PGN-1164 were stopped due
to unsuitable pharmacokinetics.[5d] Moreover, the rational
design of 5-HT2B selectivity remains arduous.[2c] Building on
our previous work using Gaussian process (GP) models for
designing NCEs[7] and de-orphaning a focused combinatorial
library[8] we envisaged that such a computational approach,
coupled to microfluidics-assisted syntheses, could efficiently
afford 5-HT2B ligands.
Computational molecular design methods offer an oppor-
tunity to promptly combat the perceived dearth of innovation
in drug discovery.[1] In particular, de novo design and
prediction of polypharmacology networks are maturing
technologies for productive lead identification in chemical
biology and medicinal chemistry.[1a,2] However, de novo
design has only been applied on a limited scale.[3] Here, we
present a pioneering large-scale de novo design study with
swift design–synthesis–bioassay cycles, aimed at the discovery
of new chemical entities (NCEs) engaging the human
serotonin 2B (5-HT2B) receptor. By combining the computa-
tional design and quantitative affinity prediction with micro-
fluidics synthesis,[4] the multi-objective NCE prioritization
afforded structurally novel, nanomolar-potent, and ligand-
efficient compounds (LE > 0.34) with G protein-coupled
receptor (GPCR) selectivity in functional cell-based assays.
Our findings not only corroborate de novo design and
machine-learning methods as prime utensils for exploring
uncharted chemical space, but also suggest that the amalga-
We employed the MAntA molecular design software[7]
implementing the reductive amination as a privileged tool
reaction for adaptive building block prioritization to obtain 5-
HT2B-selective ligands. Using both CATS2[9] pharmacophores
and Morgan substructure fingerprints,[10] the machine-learn-
ing algorithm generated 5774 candidate amines, from which
we selected 1–4 for synthesis and biochemical profiling
[Table 1, Figure 1A, Supporting Information (SI)], based on
the predicted 5-HT2B selectivity over 5-HT2A/C. To enable
rapid design cycles, our efforts initially focused on developing
[*] Dr. T. Rodrigues,[+] N. Hauser,[+] D. Reker, Dr. M. Reutlinger,
Prof. Dr. G. Schneider
Department of Chemistry and Applied Biosciences
Swiss Federal Institute of Technology (ETH)
Vladimir-Prelog-Weg 4, 8093 Zurich (Switzerland)
E-mail: gisbert.schneider@pharma.ethz.ch
T. Wunderlin, Dr. J. Hamon, Dr. G. Koch
Novartis Institutes for BioMedical Research (NIBR)
Novartis AG, 4056 Basel (Switzerland)
Figure 1. Microfluidics-assisted synthesis of amines.
[+] These authors contributed equally to this work.
a scalable and robust reductive amination process in flow.[11]
We used an automated Cetoni neMESYS microfluidic system
equipped with low-pressure, pulsation-free syringe pumps.
The platform included a 3-2-way solenoid valve, a 5 mL
borosilicate DeanFlow microreactor chip as primary reactor,
and a coil reactor (2 mL) as secondary reactor (Figure 1).
Stock solutions of all reaction components were prepared in
dimethylacetamide (DMA) and the reaction was quenched
[**] N.H. and T.R. performed organic syntheses. D.R. and M.R.
contributed computational tools and performed in silico experi-
ments together with N.H. and T.R. T.W., J.H., and G.K. contributed
binding assays. All authors analyzed and discussed data. T.R. and
G.S. designed the study and wrote the manuscript with feedback
from the remaining authors.
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201410201.
Angew. Chem. Int. Ed. 2015, 54, 1551 –1555
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
1551

.
Angewandte
Communications
Table 1: MAntA de novo designed molecules 1–4 and nearest neighbors from the ChEMBL[14] training data.
MAntA designs
Nearest neighbors from training data
Cpd.
Structure
IC50 ÆSD [nm][a]
ID[b]
Structure
Structural similarity[c]
pKi
1
886Æ314
196514
0.32
9.0[15]
2
3
3570Æ40
1214961
209324
0.57
0.37
8.0[16]
7.3[17]
7109Æ1722
4
1051Æ61
196514
0.31
9.0[15]
[a] Functional cell-based assay (IP1 quantification by HTRF detection, see SI); [b] “CHEMBL” prefix omitted from ID; [c] Tanimoto similarity index
calculated with Morgan substructure fingerprints (radius=4, 2048 bit). SD: standard deviation (n=3).
with NaHCO3 following a residence time of ca. 10 min (SI).
Whereas 1–3 could be acquired in flow, 4 was obtained in
batch due to low solubility of the building blocks in DMA.
With screening compounds 1–4 in hand, we performed
radioligand displacement assays, confirming the usefulness
of the predicted binding affinities (pAffinity) (5-HT2B
pAffinity ꢀ 7.0, experimental pKi ꢀ 6.0, Figure 2A). The
majority of the pAffinity estimates were in agreement with
the measured data, taking into account the GP modelsꢀ
background error (pKi = pAffinity Æ 0.8; Figure 2A). Fur-
thermore, all compounds showed moderate antagonistic
behavior in a functional cell-based assay (Table 1), and
good ligand efficiency (LE ꢀ 0.35, Figure 2A). Under
dynamic light scattering none of the compounds presented
measurable colloidal aggregation at a concentration of 100 mm
that could equate unspecific binding and artefacts when
probing for antagonistic behavior in GPCRs.[12]
We further screened compounds 1–4 in functional assays
at 10 mm, to assess the scope of the predictive models. As tool
GPCRs, we selected 5-HT2A and related off-targets (adrener-
gic a1A and dopamine D2 receptors;[2c,6,13] Figure 2B). The
potent in vitro effects and high ligand efficiencies of the
de novo designed compounds, the moderate selectivity
Figure 2. A) Predicted variance-corrected affinity (pAffinity) and exper-
imental pKi values obtained for 1–4 through radioligand displacement
assays against 5-HT2A–C. Ligand efficiency LE=(À1.4 logKi)/number
(Table 1 and Figure 2B,C) and general scaffold innovation
(SciFinder, www.scifinder.org) compared to the nearest
heavy atoms. B) Cell-based functional effects (%antagonism and
agonism) of 1–4 against selected off-targets (10 mm, n=2). Note: Full
details on controls are provided in the SI. C) LiSARD[18] landscapes
(CATS2 pharmacophore descriptors) for 5-HT2A–C receptor binding and
5-HT2B selectivity over 5-HT2A/C. Landscapes were computed from
available ChEMBL version 16 affinity data and 1–4.: 3269 (5-HT2A),
1719 (5-HT2B), and 3307 (5-HT2C) small molecules, totaling 4137
unique entities.
neighbors in ChEMBL (substructure-based Tanimoto sim-
ilarity ~ 0.4), suggest that 1–3 may serve as 5-HT2B-selective
leads for future development. The LiSARD[18] software
generates interactive graphics by computing structure–activ-
ity relationship landscapes that may serve as roadmaps for
molecular design and compound prioritization. For this study,
the overlapping preferred design areas for 5-HT2A–C receptors
certified the difficulty of simultaneously obtaining potent and
selective 5-HT2B-targeting NCEs (Figure 2C). Notwithstand-
ing these promising preliminary results obtained with
MAntA, we recognized that engagement of the dopamine D2
receptor by 1–4 would constitute a significant liability (Fig-
ure 2B). As a consequence, we envisaged that increasing the
1552
www.angewandte.org
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 1551 –1555

Angewandte
Chemie
scaffold diversity of the computationally designed chemo-
types might help us find molecules with improved 5-HT2B
selectivity. Therefore, we employed the de novo design
software DOGS[19] coupled to the GP prediction models to
generate NCEs, taking 239 GPCR-targeting and FDA-
approved small molecules as design templates (GPCR
SARfari, http://www.ebi.ac.uk). We aimed at increasing the
framework diversity of de novo-designed molecules by
expanding the search space with up to 58 tractable reaction
types for stepwise fragment growing.[19] Furthermore, we
employed a large template set (SARfari), as target promis-
cuity is common among GPCR ligands.[2c] From the DOGS
runs we obtained 111854 unique, synthetically accessible and
GPCR-ligand-like designs. A total of 25363 Murcko scaffolds
with profound dissimilarity to the design templates and
annotated high-affinity 5-HT2B ligands in ChEMBL (average
structural Tanimoto similarity = 0.13 Æ 0.07 and 0.17 Æ 0.04,
respectively) highlight the scaffold-hopping capabilities and
intrinsic explorative nature of the DOGS algorithm.
We subjected all designed NCEs to the PAINS[20] and
ADMET[21] filters to eliminate potentially undesirable small
molecules and scored the remaining 78468 designs with the
GP target prediction models. In line with previous findings in
which ca. 30% of the Novartis corporate compound library
binds to 5-HT2B at 10 mm,[13] the computationally designed
molecules presented an average pAffinity value of 5.0 Æ 0.6.
For the training data we computed an average pAffinity value
of 5.1 Æ 0.5, suggesting no bias in the design and scoring
algorithms towards 5-HT2B ligands over other GPCRs. We
selected compounds 5–8 for subsequent synthesis according
to either pAffinity and/or GPCR-panel selectivity criteria.
Although no 5-HT2B selectivity was forecast for 5 (Fig-
ure 3A), we chose it because of its
Consequently, we prioritized compounds 6–8 based on GPCR
panel selectivity, including the 5-HT2A and D2 receptors
(Figure 3A and SI). Compounds 6 and 7 were designed from
the low-affinity 5-HT2B ligands rizatriptan and metaraminol,
respectively. Molecule 6 had already been disclosed as a 5-
HT2B receptor agonist,[25] but its selectivity over the dopa-
mine D2 receptor remained unknown. Although limited
utility as a drug lead may be foreseen for 6, due to 5-HT2B
agonism, the herein experimentally confirmed potent func-
tional effect (EC50 = 18 Æ 3 nm; Figure 2B) and GPCR panel
selectivity (Figure 3C, SI) may still warrant high value as
a chemical probe. A two-step synthetic route afforded 7.
Despite its excellent overall selectivity (Figure 3C), the
moderate 5-HT2B binding affinity (Ki = 1700 Æ 25 nm; Fig-
ure 3A) was translated into an undesired partial agonistic
effect (EC50 = 2365 Æ 243 nm; Figure 3B). Irrespective of the
functional activity of the profiled ligands, the computer-
assisted design method (DOGS) fulfilled our primary goal of
designing GPCR-tailored ligands with accurately predicted
selectivity profiles. Compound retrieval by library enumera-
tion and similarity searching would probably have failed to
identify 6–7 as screening candidates due to their low structural
similarity to the design templates as well as the ChEMBL
training data (Table 2). Our final focus was on the piperazine
derivative 8, which we acquired in flow. Binding assessment
against the 5-HT2 receptor panel revealed selectivity toward
5-HT2B (Ki = 251 Æ 0.02 nm, LE = 0.42; 5-HT2A: Ki = 3383 Æ
1100 nm; 5-HT2C: Ki = 18100 Æ 1570 nm), a potent antagonis-
tic effect in vitro (IC50 = 611 Æ 240 nm, Figure 3B) and func-
tional selectivity against a panel of GPCR off-targets and the
hERG cardiac potassium channel (Figure 3C). These results
are important taking into account that 8 presented no
chemical similarity to aripipra-
zole[24] (Tanimoto similarity =
0.51) and its predicted 5-HT2B
binding affinity (pAffinity = 6.8).
Analogously to its template
(Table 2) and as predicted by the
GP models, compound 5 showed
binding affinity for 5-HT2A-B (Fig-
ure 3A). Our results suggest the
importance of structural features in
aripiprazole for potent 5-HT2B
binding and overall 5-HT2 receptor
subtype selectivity. In an orthogo-
nal assay, 5 showed a strong 5-HT2B
antagonistic effect (IC50 = 225 Æ
26 nm; Figure 3B), fully corrobo-
rating receptor binding. The
remaining pooled designs were
analyzed for potential 5-HT2B
selectivity over other GPCRs and
Figure 3. A) Predicted affinity (pAffinity) and experimental pKi values obtained for 5–8 and template
structures through radioligand assays against 5-HT2A–C. Ligand efficiency LE=(À1.4 logKi)/number
heavy atoms. B) Cell-based functional effect (IP1 quantification by HTRF detection, see SI) of 5–8
against 5-HT2B. Compounds 5 and 8 are antagonists of 5-HT2B, whereas 6 and 7 are full and partial
agonists, respectively. C) Cell-based functional effects (%antagonism and agonism) of 6–8 against
selected off-targets (10 mm, n=2), suggesting selectivity for the 5-HT2B receptor. Note: Full details on
hERG. Although no design
revealed clear predicted selectivity
over the 5-HT2A/C and D2 receptors
simultaneously, the engagement of
5-HT2C is apparently uncorrelated
with known adverse drug reactions. controls are provided in the SI; For hERG, %inhibition is given.
Angew. Chem. Int. Ed. 2015, 54, 1551 –1555
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
1553

.
Angewandte
Communications
Table 2: DOGS de novo designed molecules 5–8 and nearest neighbors from the ChEMBL training data.
DOGS designs
Cpd. Structure
Nearest neighbors from training data
Template
Structural
similarity[a]
ID[b]
Structure
Structural
similarity[a]
pKi
5
0.51
1112
0.51
8.9[15]
6
7
0.16
0.16
1255834
402794
1.0
8.6[22]
0.29
<5.0
8
0.16
7257
0.26
7.3[23]
[a] Tanimoto similarity index calculated with Morgan substructure fingerprints (radius=4, 2048 bit); [b] “CHEMBL” prefix omitted from ID.
Keywords: computer-assisted molecular design ·
drug discovery · GPCR · microfluidics · organic synthesis
measurable colloidal aggregation at concentrations at least
200-fold higher than the relevant bioactive concentration.
Furthermore, 8 presents high water solubility (kinetic sol-
ubility > 80 mgmLÀ1) and can be acquired in one synthetic
step. Peculiarly, the design template of 8, chlorpromazine,
revealed complete absence of 5-HT2 receptor panel selectiv-
ity, and the nearest neighbor in the training data
(CHEMBL7257, Table 2) is a 5-HT2B agonist with similar
receptor affinity as 8.[23]
Although there are more potent and 5-HT2B receptor-
selective antagonists described in the literature, for example,
RS-127445[26] and PRX-08066,[27] we consider the design of 8
a successful outcome of this proof-of-concept study aiming at
target-panel selectivity. The compound was readily obtained
without the need for high-throughput screening or lengthy hit
optimization. The discovery platform described herein is
widely applicable to quickly identify starting points for other
GPCRs and drug target families, provided that templates for
de novo design are available. Our GP models currently
encompass several hundred human drug targets. Additional
affinity prediction models should become available with
publicly accessible database updates. The results of this study
suggest a feasible solution for fast fragment-based de novo
design of NCEs with accurately predicted designer polyphar-
macological or selectivity profiles. Together with the poly-
genic nature of several diseases, such platforms may be suited
to economically prototype efficacious tools for personalized
medicine.[28] The results obtained validate the combination of
machine-learning methods with automated chemical synthe-
sis and fast bioassay turnover as a general approach for rapid
hit and lead discovery.
.
[1] a) T. Rodrigues, G. Schneider, Synlett 2014, 25, 170 – 178; b) Y. L.
Bennani, Drug Discovery Today 2011, 16, 779 – 792.
[2] a) A. L. Hopkins, Nat. Chem. Biol. 2008, 4, 682 – 690; b) J.
Besnard, G. F. Ruda, V. Setola, K. Abecassis, R. M. Rodriguiz,
X. P. Huang, S. Norval, M. F. Sassano, A. I. Shin, L. A. Webster,
F. R. Simeons, L. Stojanovski, A. Prat, N. G. Seidah, D. B.
Constam, G. R. Bickerton, K. D. Read, W. C. Wetsel, I. H.
Gilbert, B. L. Roth, A. L. Hopkins, Nature 2012, 492, 215 – 220;
c) H. Lin, M. F. Sassano, B. L. Roth, B. K. Shoichet, Nat.
Methods 2013, 10, 140 – 146; d) E. Lounkine, M. J. Keiser, S.
Whitebread, D. Mikhailov, J. Hamon, J. L. Jenkins, P. Lavan, E.
Weber, A. K. Doak, S. Cote, B. K. Shoichet, L. Urban, Nature
2012, 486, 361 – 367; e) M. J. Keiser, B. L. Roth, B. N. Armbrus-
ter, P. Ernsberger, J. J. Irwin, B. K. Shoichet, Nat. Biotechnol.
2007, 25, 197 – 206; f) D. Reker, T. Rodrigues, P. Schneider, G.
Schneider, Proc. Natl. Acad. Sci. USA 2014, 111, 4067 – 4072.
[3] a) T. Rodrigues, T. Kudoh, F. Roudnicky, Y. F. Lim, Y.-C. Lin,
C. P. Koch, M. Seno, M. Detmar, G. Schneider, Angew. Chem.
Int. Ed. 2013, 52, 10006 – 10009; Angew. Chem. 2013, 125, 10190 –
10193; b) B. Spꢁnkuch, S. Keppner, L. Lange, T. Rodrigues, H.
Zettl, C. P. Koch, M. Reutlinger, M. Hartenfeller, P. Schneider,
G. Schneider, Angew. Chem. Int. Ed. 2013, 52, 4676 – 4681;
Angew. Chem. 2013, 125, 4774 – 4779; c) T. Rodrigues, F.
Roudnicky, C. P. Koch, T. Kudoh, D. Reker, M. Detmar, G.
Schneider, Chem. Sci. 2013, 4, 1229 – 1233.
[4] a) T. Rodrigues, P. Schneider, G. Schneider, Angew. Chem. Int.
Ed. 2014, 53, 5750 – 5758; Angew. Chem. 2014, 126, 5858 – 5866;
b) K. S. Elvira, X. Casadevall i Solvas, R. C. Wootton, A. J.
deMello, Nat. Chem. 2013, 5, 905 – 915.
[5] a) M. Berger, J. A. Gray, B. L. Roth, Annu. Rev. Med. 2009, 60,
355 – 366; b) N. Moss, Y. Choi, D. Cogan, A. Flegg, A. Kahrs, P.
Loke, O. Meyn, R. Nagaraja, S. Napier, A. Parker, J. T. Peterson,
P. Ramsden, C. Sarko, D. Skow, J. Tomlinson, H. Tye, M.
Whitaker, Bioorg. Med. Chem. Lett. 2009, 19, 2206 – 2210; c) Y. J.
Kwon, S. Saubern, J. M. Macdonald, X. P. Huang, V. Setola, B. L.
Received: October 20, 2014
Revised: October 30, 2014
Published online: December 4, 2014
1554
www.angewandte.org
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 1551 –1555

Angewandte
Chemie
Roth, Bioorg. Med. Chem. Lett. 2010, 20, 5488 – 5490; d) Ency-
clopedia of Molecular Pharmacology, 2nd ed. (Eds.: S. Offer-
manns, W. Rosenthal), Springer, Heidelberg, 2008, p. 1125.
[6] J. Hamon, S. Whitebread, V. Techer-Etienne, H. Le Coq, K.
Azzaoui, L. Urban, Future Med. Chem. 2009, 1, 645 – 665.
[7] M. Reutlinger, T. Rodrigues, P. Schneider, G. Schneider, Angew.
Chem. Int. Ed. 2014, 53, 4244 – 4248; Angew. Chem. 2014, 126,
4330 – 4334.
[8] M. Reutlinger, T. Rodrigues, P. Schneider, G. Schneider, Angew.
Chem. Int. Ed. 2014, 53, 582 – 585; Angew. Chem. 2014, 126, 593 –
596.
[9] M. Reutlinger, C. P. Koch, D. Reker, N. Todoroff, P. Schneider, T.
Rodrigues, G. Schneider, Mol. Inf. 2013, 32, 133 – 138.
[10] D. Rogers, M. Hahn, J. Chem. Inf. Model. 2010, 50, 742 – 754.
[11] J. E. Hochlowski, P. A. Searle, N. P. Tu, J. Y. Pan, S. G. Spanton,
S. W. Djuric, J. Flow Chem. 2011, 2, 56 – 61.
R. Nudelman, M. Kauffman, S. Noiman, J. Med. Chem. 2006, 49,
3116 – 3135.
[18] M. Reutlinger, W. Guba, R. E. Martin, A. I. Alanine, T.
Hoffmann, A. Klenner, J. A. Hiss, P. Schneider, G. Schneider,
Angew. Chem. Int. Ed. 2011, 50, 11633 – 11636; Angew. Chem.
2011, 123, 11837 – 11840.
[19] M. Hartenfeller, H. Zettl, M. Walter, M. Rupp, F. Reisen, E.
Proschak, S. Weggen, H. Stark, G. Schneider, PLoS Comput.
Biol. 2012, 8, e1002380.
[20] J. B. Baell, G. A. Holloway, J. Med. Chem. 2010, 53, 2719 – 2740.
[21] D. Lagorce, O. Sperandio, H. Galons, M. A. Miteva, B. O.
Villoutreix, BMC Bioinf. 2008, 9, 396.
[22] S. P. Alexander, A. Mathie, J. A. Peters, Br. J. Pharmacol. 2009,
158, S1 – 254.
[23] J. A. May, A. P. Dantanarayana, P. W. Zinke, M. A. McLaughlin,
N. A. Sharif, J. Med. Chem. 2006, 49, 318 – 328.
[12] M. F. Sassano, A. K. Doak, B. L. Roth, B. K. Shoichet, J. Med.
Chem. 2013, 56, 2406 – 2414.
[24] J. D. Urban, G. A. Vargas, M. von Zastrow, R. B. Mailman,
Neuropsychopharmacology 2007, 32, 67 – 77.
[13] J. Bowes, A. J. Brown, J. Hamon, W. Jarolimek, A. Sridhar, G.
Waldron, S. Whitebread, Nat. Rev. Drug Discovery 2012, 11,
909 – 922.
[25] S. P. Alexander, A. Mathie, J. A. Peters, Br. J. Pharmacol. 2011,
164, S1 – 324.
[26] D. W. Bonhaus, L. A. Flippin, R. J. Greenhouse, S. Jaime, C.
Rocha, M. Dawson, K. Van Natta, L. K. Chang, T. Pulido-Rios,
A. Webber, E. Leung, R. M. Eglen, G. R. Martin, Br. J.
Pharmacol. 1999, 127, 1075 – 1082.
[27] S. L. Porvasnik, S. Germain, J. Embury, K. S. Gannon, V.
Jacques, J. Murray, B. J. Byrne, S. Shacham, R. Al-Mousily, J.
Pharmacol. Exp. Ther. 2010, 334, 364 – 372.
[28] W. E. Evans, J. A. Johnson, Annu. Rev. Genomics Hum. Genet.
2001, 2, 9 – 39.
[14] A. Gaulton, L. J. Bellis, A. P. Bento, J. Chambers, M. Davies, A.
Hersey, Y. Light, S. McGlinchey, D. Michalovich, B. Al-Lazikani,
J. P. Overington, Nucleic Acids Res. 2012, 40, D1100 – D1107.
[15] J. H. Lange, J. H. Reinders, J. T. Tolboom, J. C. Glennon, H. K.
Coolen, C. G. Kruse, J. Med. Chem. 2007, 50, 5103 – 5108.
[16] I. A. Moussa, S. D. Banister, C. Beinat, N. Giboureau, A. J.
Reynolds, M. Kassiou, J. Med. Chem. 2010, 53, 6228 – 6239.
[17] O. M. Becker, D. S. Dhanoa, Y. Marantz, D. Chen, S. Shacham, S.
Cheruku, A. Heifetz, P. Mohanty, M. Fichman, A. Sharadendu,
Angew. Chem. Int. Ed. 2015, 54, 1551 –1555
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
1555