Off-rate screening (ORS) by surface plasmon resonance. An efficient method to kinetically sample hit to lead chemical space from unpurified reaction products

Article abstract of DOI:10.1021/jm401848a

The dissociation rate constant k_d (off-rate) is the component of ligand-protein binding with the most significant potential to enhance compound potency. Here we provide theoretical and empirical data to show that this parameter can be determined accurately from unpurified reaction products containing designed test compounds. This screening protocol is amenable to parallel chemistry, provides efficiencies of time and materials, and complements existing methodologies for the hit-to-lead phase in fragment-based drug discovery.

Full text of DOI:10.1021/jm401848a

Featured Article
pubs.acs.org/jmc
Off-Rate Screening (ORS) By Surface Plasmon Resonance. An Efficient
Method to Kinetically Sample Hit to Lead Chemical Space from
Unpurified Reaction Products
James B. Murray,* Stephen D. Roughley, Natalia Matassova, and Paul A. Brough*
Vernalis (R&D) Ltd., Granta Park, Great Abington, Cambridge CB21 6GB, U.K.
S
* Supporting Information
ABSTRACT: The dissociation rate constant kd (off-rate) is the component of ligand−protein binding with the most significant
potential to enhance compound potency. Here we provide theoretical and empirical data to show that this parameter can be
determined accurately from unpurified reaction products containing designed test compounds. This screening protocol is
amenable to parallel chemistry, provides efficiencies of time and materials, and complements existing methodologies for the hit-
to-lead phase in fragment-based drug discovery.
simple ratio of dissociation/association rate constants as
described in eq 1:
INTRODUCTION
■
The pressure on the pharmaceutical industry to develop new
drugs has led to a search for efficiency gains within research
organizations globally. Since the widespread adoption of
parallel chemistry techniques, the purification of synthesized
products has become a significant bottleneck within medicinal
chemistry. The use of automated HPLC systems has improved
this situation. However, such systems are expensive to operate,
use large volumes of solvents, and suffer from poor material
recovery. Alternative approaches, such as solid-supported
scavenger reagents or ion-exchange chromatography, are only
applicable to certain reaction types with limited substrate scope.
These methods still require individual weighing and dilution
prior to biological screening.
[PF][CF]
[PC]
kd
ka
KD =
=
(1)
where P is target, C is compound, F is free, and kd and ka are the
dissociation and association rate constants, respectively.
Measuring affinity kinetically throughout a drug discovery
program is more informative. For example, a compound with a
10-fold slower on- and off-rate would not be recognized as
different if evaluated by equilibrium measures of affinity. Such
an observation may indicate a novel protein conformation or
significant internal strain in the bound ligand. Either way, the
bound conformation is a productive one that may yield to the
significant progress toward the drug candidate if exploited
further. This is exemplified with the concept of kinetic
efficiency.3
A number of papers have been devoted to the analysis and
consequences of varying residence times for drug candidates.
The dissociation rate can vary from the immeasurably fast to
the immeasurably slow (i.e., stable covalent complex). Thus,
the off-rate (defined by Copeland et al)4 is the component of
Fragment-based drug discovery (FBDD)1,2 has become
prevalent within the pharmaceutical industry as an effective
method to identify hits for existing and novel targets. However,
fragments hits are typically of a lower affinity than HTS hits and
often generate many tens, if not hundreds, of potential starting
points. The problem can perhaps be best described as not “Do I
have anything to work on?” but more “What should I work
on?”
The affinity of a compound for the target is described by the
steady state affinity equilibrium dissociation constant KD (or its
surrogates IC50/EC50). Affinity (KD) can determined from the
bound and unbound concentrations of ligand and target or as a
Received: November 29, 2013
Published: February 12, 2014
© 2014 American Chemical Society
2845
dx.doi.org/10.1021/jm401848a | J. Med. Chem. 2014, 57, 2845−2850

Journal of Medicinal Chemistry
Featured Article
binding that has the greatest potential to improve potency.
Early papers in this field were devoted to biological
consequences of prolonged target inhibition, leading to lower
toxicity due to more rapid off-target dissociation.4,5 For in vivo
systems, the maximal contribution of the dissociation rate to
efficacy will be limited by the rate of target resynthesis. More
recently, there has been an effort to understand the molecular
determinants leading to increased residence times.6,7 While
progress is being made, this remains a largely unsolved
problem. We reason here that one could exploit the
dissociation rate empirically and remove significant bottlenecks
in drug discovery, in particular within fragment-to-lead efforts.
We provide both the theoretical and experimental support
for this approach. We demonstrate this by using surface
plasmon resonance (SPR) assessment of the dissociation rate
constants from a library of crude (unpurified) reaction mixtures
which contain designed target compounds.
a minimum, and upon completion the reactions were
evaporated and dissolved in DMSO to nominally match the
stock concentration of the pure compound (assuming 100%
yield). These samples were diluted (1:99) with the running
buffer and injected over HSP90 immobilized on the SPR chip.
Injection of blank reaction sample served as the reference
control.
Dissociation phase of the resulting double referenced
sensorgrams were evaluated using BIAevaluation 1.1 (BIAcore
GE Healthcare Bio-SciencesCorp) or Scrubber2 (BioLogic)
software by fitting with a one-phase exponential decay model,
eq 2:
R = R0·e−k ·(t−t ) + R∞
d
0
(2)
where the R0 and t0 are the response and the time at the start of
the dissociation phase, respectively, and R∞ is a residual
response after complete dissociation.
Initial SPR experiments were carried out using a BIAcore
T100. We have stored all the samples at −18 °C and retested
these 3.5 years later using a BIAcore T200. This permits the
evaluation of the crude library stability over a considerable time
and the responses of differing equipment (T200 cf T100). The
dissociation rate constants determined for the pure samples and
those generated from the ORS library are shown in Table 1.
The T200 data shows and average difference in the kds between
crude and pure samples of 19%, similarly, the older T100 data
had a 15% difference. Satisfyingly, we observe that the kds vary
by an average of only 30% when compared across instruments
and across time. This demonstrates that carryover contami-
nation, long-term storage, and differing equipment has a
modest effect on the observed kds. Indeed, these small
deviations in the kds closely reflect the differences observed
in multilaboratory studies where variability observed has been
reported to be from 14% to 40% depending on the system.9
Potter et al. have previously described a series of inhibitors of
the prolyl-isomerase PIN1 based on the β-(benzimidazol-2-
yl)alanine scaffold (R)-7.10 We resynthesized a set of these
compounds according to Scheme 2 and analyzed the crude
reaction products by ORS. Furthermore, we investigated
whether the more readily available racemates (rac-7a−g)
RESULTS
■
We have compared the kd observed for crude reaction mixtures
with the kd of the pure compound. First, this was done by
recapitulating early medicinal chemistry efforts targeted at
HSP90 that led to the preclinical candidate NVP-BEP800.8 A
small library of 13 compounds was generated for off-rate
screening (ORS) experiments using the same Suzuki chemistry
employed in the original work, Scheme 1. Workup was kept to
a
Scheme 1
a
Reagents and conditions: (a) ArB(OH)2, NaHCO3, PdCl2(PPh3)2,
DMF, H2O, 100 °C microwave, 10 min.
Table 1. Results of the Hsp90 Screen: Dissociation Rate Constants Determined by ORS for the Crude Reaction Mixtures and
for the Respective Pure Compounds
T200 kd (s−1
)
T100 kd (s−1
)
compd
R1
R2
R3
R4
pure
crude
pure
crude
a
1
fast
fast
fast
0.10
0.09
nt
2a
2b
2c
2d
2e
2f
H
H
H
H
Me
H
H
H
H
H
H
F
H
H
H
H
H
1.05
0.16
0.26
nt
1.08
0.19
0.29
fast
fast
0.10
0.09
fast
Me
H
H
Cl
H
Me
Me
Me
H
H
Me
0.15
0.03
0.16
0.02
0.03
nt
0.09
0.03
0.20
0.02
0.03
0.02
nt
0.12
0.02
0.12
0.01
0.03
nt
0.10
0.02
0.15
0.02
0.02
0.02
fast
Me
CN
Cl
H
H
2g
2h
2i
H
Cl
H
Me
Me
H
CO2Me
2j
CN
CN
H
H
H
H
H
H
2k
2l
0.63
fast
fast
nt
fast
fast
fast
nt
Et
H
H
H
fast
fast
2m
2n
F
F
0.78
0.31
fast
OMe
H
0.37
a
Fast = kd > 1.2 s−1. nt is not tested
2846
dx.doi.org/10.1021/jm401848a | J. Med. Chem. 2014, 57, 2845−2850

Journal of Medicinal Chemistry
Featured Article
a
identified (Table 2). Closer inspection showed that, as above,
the kd determined for the crude reactions was in very good
Scheme 2
Table 2. Measured Dissociation Rate Constants of 7a−7g
Binding to PIN1
a
Ratio of starting material (5) to products as defined in Scheme 2.
*Fast refers to kd > 2 s−1.
a
Reagents and conditions: (a) SOCl2, MeOH, 80%; (b) RCO2H,
COMU, Et3N, DMF; (c) (1) LiOH (5 equiv), MeOH-H2O, (2)
AcOH (5 equiv).
agreement with those for the pure compound. We can see that
the dissociation rates differ by less than 15% on average for
either the racemic or enantiomerically pure reagents. This
clearly demonstrates that, even with complex crude mixtures,
we are able to recapitulate the observed kds of pure samples and
to readily identify the most active final compound with minimal
effort.
Finally, to further explore if any of the most commonly
deployed chemistries within the drug discovery industry11
possess any inherent carryover liability, we conducted a series
of Faux reactions by combining simple example reagents and
catalysts to generate simple compounds (see Supporting
Information) that were unlikely to bind to the target but able
to replicate reagent carryover and related degradation products.
The resultant mixtures were then subjected to the normal
heating/cooling or irradiation procedures (See Supporting
Information) and worked up by evaporation. A 20 mM stock
solution of pure compound (2i for HSP90; 7g for PIN1) was
added to the dried residue and mixed with gentle heating for 30
min. The final samples were evaluated as per the standard ORS
protocol (Table 3).
would also demonstrate the expected SAR from ORS.
Additionally, we chose to conduct the repeat syntheses in
disposable plastic tubes with rubber septa. The tubes were
tested for solvent compatibility by loading with deuterated
solvents and assessed for solvent loss (by weighing), tube
condition (by visual inspection), and leaching of contaminants
1
by the solvent (by H NMR analysis).
At ambient temperature (18−22 °C), only tubes containing
DCM and CDCl3 showed appreciable solvent loss. At 55 °C,
CDCl3 also showed significant losses (Figure 1). In all cases,
It can be seen that the dissociation rate constant varied on
average by 5.8% and 7.8%, with maximal variations of 12% and
18% for HSP90 and PIN1, respectively. This clearly
demonstrates that generic carryover from more than a half of
the most commonly utilized drug discovery chemistries has no
significant effect on the observed off-rates.
Figure 1. Solvent loss from 96-well Matrix polypropylene tubes: (A) at
ambient temperature 18−22 °C, (B) at 55 °C.
there was leaching of unidentified aliphatic hydrocarbon
contaminants into the solvents, especially for acetone and
chlorinated solvents. If this material interferes (unlikely) with
the SPR binding signal, it is compensated for by inclusion of an
appropriate “blank” control. However, the tubes maintained
their integrity throughout the experiment. Because of
evaporation, the more volatile solvents such as DCM or
CDCl3 were unsuitable for high temperature or long reaction
times.
DISCUSSION
■
Our approach relies on two independent components. The first
one is the fractional ligand binding at equilibrium. It relates to
the extent to which a compound (“product” compound or
starting material (SM)) in the crude reaction mixture is bound
to the protein immobilized on the surface of the SPR chip. The
second one is the rate of decay of the chip-bound complexes. It
relates to the dissociation rates of each individual complex
present in the mixture.
Within the set of PIN1 compounds, all of the reaction
mixtures containing the most active compounds were easily
2847
dx.doi.org/10.1021/jm401848a | J. Med. Chem. 2014, 57, 2845−2850

Journal of Medicinal Chemistry
Featured Article
Table 3. Measured Dissociation Rate Constants of 2i
Binding to HSP90 in the Faux Reaction Mixtures
Table 4. Initial Response As Function of LE/HA Addition
with 5% Yield
a
HSP90 kd
(s−1
HSP90
PIN1 kd
(s−1
PIN1
product LE
Faux reaction
)
variation (%)
)
variation (%)
added HA
0.4
0.35
0.3
amide formation
Boc deprotection
0.042
0.042
0.043
4.0
3.7
0.2
0.41
0.42
0.37
13.6
18.4
4.4
1
3
5
7
9
0.098
0.317
0.661
0.891
0.971
0.024
0.080
0.236
0.521
0.792
0.005
0.016
0.046
0.127
0.301
N-alkylation with
R-X
N-arylation
(SNAr)
0.046
0.042
0.042
0.041
0.040
0.040
5.6
3.8
4.6
5.1
8.8
8.6
0.38
0.37
0.37
0.40
0.36
0.40
6.0
4.0
O-alkylation
(Mitsunobu)
explored, the higher the initial product signal is likely to be,
even with poor group efficiencies.
O-alkylation with
R-X
3.3
Protein−ligand dissociation: once sample is no longer
applied to the protein surface, the bound components
dissociate in a zero-order process. The sensorgram signal is
proportional to the relative fraction bound at the end of the
injection and the kd of each bound component. For a simple
single-step mechanism with two binding components, the
dissociation sensorgram can be described by eq 4.
reductive
amination
12.6
1.7
Sonogashira
coupling
sulphonamide
formation
11.5
Suzuki reaction
urea formation
control
0.042
0.038
0.044
3.5
0.37
0.38
0.36
4.4
6.2
11.7
1
2
RU = RUeq e−k (t−t ) + RUeq e−k (t−t )
1
2
d
0
d
0
obs
(4)
a
Percentage variation is calculated relative to the Blank well (reagent
2i).
where RU1eq and RUe2q are the weighted responses at the start of
the dissociation phase, see eq 3. t0 is time at the start of
dissociation.
Starting materials of low affinity (KD > 10−5 M) and
molecular weight (app 190) will typically have a fast
dissociation rate. The exemplified compounds will typically
be of greater mass in average of higher than 270. If the extra
mass contributes to affinity in an efficient way, the newly
synthesized compound should be correspondingly 20−50-fold
more potent, much of which will be exhibited through the
slower off-rate. Figure 2 demonstrates the effect on the
observed dissociation rate for such a mixture.
Active site occupancy: in a two-component system, relative
site occupancy at equilibrium is simply a ratio of the potency of
the two components and the concentration at which they are
present in the solution. The relative response at equilibrium for
the components can be given by eq 3:
KD2
KD1
F/(1 − F)
RU1eq
=
× Mw ×
1
1
1
KD2
KD1
F1
1 +
×
1 − F1
(
)
(3)
where RU1eq is the relative response at equilibrium and is
assumed to be directly proportional to the mass. KD is the
equilibrium affinity constant of the relevant component. F is the
fractional purity of the relevant component.
Thus, when two components with masses of 230 and 270 Da
are present in equal concentrations, and one is 10-fold more
potent than another, 93% of the binding signal will be due to
the more potent component. Indeed, even at concentration
ratio of 19:1, 40% of the binding signal will be due to the more
potent component. Screening is conducted at high nominal
concentration to ensure that equilibrium on the surface of the
SPR chip is reached rapidly and bound signal is high.
Ligand efficiency12 (LE; kcal/mol/HA) gains will be
apparent even at relatively low yields and few heavy (HA)
atom additions. For example, at 5% yield with the addition of
three heavy atoms (17 to 20 HA) and maintenance of a LE of
0.4 kcal/mol, 31% of the occupancy (initial response units in
SPR) will be due to the product. Interestingly, for the greater
heavy atom additions, significant losses in LE are more
tolerated. For example, if we have added seven heavy atoms
and the product LE was reduced to 0.35 kcal/mol (17 to 24
HA), over 50% of the initial response will be due to the
product, despite the poor group efficiency13 of 0.22 kcal/mol.
This is exemplified more fully in Table 4 for a 5% reaction
yield.
Figure 2. Theoretical dissociations phase for sensorgrams with varying
reaction yield using eq 4. The SM (mw 190) with KD 1E−5 M and kd 1
s
−1 and the putative product (mw 270) with KD of 1E−6 M and kd 0.1
s−1.
For a 1:1 ratio of starting material to product at 3 s into the
dissociation, more than 99% of the signal observed is due to the
product. Furthermore, at 19:1 ratio and 3 s of dissociation, 92%
of the observed signal is due to the more potent compound.
Thus, even at low yields the more substantial the gain in off-
rate, the more readily the active component will be detected.
Each reaction mixture is designed to generate a single
product (with starting material plus remaining unreacted
reagents and by products). LCMS analysis was used to provide
Here the assumptions are that the product LE is maintained
(0.4) or degraded (0.35 and 0.3) compared with the SM LE.
This demonstrates that as more diverse chemical space is
2848
dx.doi.org/10.1021/jm401848a | J. Med. Chem. 2014, 57, 2845−2850

Journal of Medicinal Chemistry
Featured Article
evidence that the desired target compound was present in each
crude reaction mixture.
thus eliminating a significant bottleneck in the early stages of
medicinal chemistry efforts.15 Application of this approach will
shorten fragment to lead times, reduce solvent and reagent use,
leading to “greener” more productive fragment-based lead
generation.16
We have tested this concept in two sets of experiments on
the drug targets HSP90 (classically drugable) and PIN1 (hard
to drug). These tests were to determine: (a) if we could reliably
detect the most active substituents from crude reaction
mixtures and, (b) if the most commonly employed reactions
were amenable to this approach. As shown in the Results
section and Supporting Information, we could easily identify
the active substituents for both HSP90 and PIN1, with the kd
values correlating to the known SAR derived from biochemical
assays. Gratifyingly, the measured kds are typically within 25%
of the pure kd, which is as expected due to the slowly
dissociating component dominating the dissociation phase.
Using a series of Faux reactions (Table 3), we sought to
determine the effect on the observed off-rate of any carryover
contaminants from 62% of the most commonly employed
reactions.11 We observed a maximal difference of 12% in the
observed kd, clearly indicating that there is no appreciable effect
caused by the reaction components.
An important consideration is the dissociation rate of the
SM, if this is slow, then in low yielding reactions it will be
difficult to identify the product with slower kd. This can be
ameliorated by conducting the screening at higher temperatures
where the kd is increased, thereby contributing less to the
observed dissociation sensorgram. Conversely, for rapidly
dissociating systems, the screening temperature can be lowered,
e.g., PIN1.
Considerable resources are expended to evaluate compounds
in concentration sensitive assays. The likely outcome for any
one compound would be a poor response in the relevant
bioassay, particularly during the early stages of fragment-to-lead
chemistry. These resources are primarily: large scale synthesis
(typically >5 mgs), time-consuming analytical chemistry,
purification (high solvent use), accurate weighing (time-
consuming), storage of both liquid and solid samples (extensive
compound management), and bioassays (multiple dilutions/
data analysis). These all add significantly to the cost of the early
stage drug development. The ORS approach described here
mitigates the use of such expensive and time-consuming
resources, although it is important to note that hits identified
from ORS are generally reprepared and purified for
confirmation and further characterization
EXPERIMENTAL SECTION
Libraries were made and stored as 20 mM solutions in DMSO or
DMSO-d6 at −20 °C.
■
General Procedure A: Preparation of 4-Arylthieno[2,3-
d]pyrimidines 2a−2n by Suzuki Cross-Coupling. To each of 14
microwave vials was added 2-amino-4-chloro-thieno[2,3-d]pyrimidine-
6-carboxylic acid ethyl ester 1 (50 mg, 0.194 mmol),8 DMF (3 mL),
NaHCO3 (1N aq; 0.5 mL), the appropriate boronic acid (0.291 mmol,
1.5 equiv), and bis(triphenylphosphine)palladium(II) dichloride (14
mg, 10 mol %). Each vial was sealed and heated in microwave
synthesizer at 100 °C for 10 min. The reaction mixtures were
transferred to 50 mL boiling tubes and solvents evaporated in vacuo
(Genevac). Each crude product was partitioned between satd NaCl
(aq) solution (3 mL) and EtOAc (3 mL), stirred for 2 min, and the
EtOAc layer pipetted off and filtered through a small plug of
anhydrous Na2SO4 in a SPE cartridge. The filtrates were evaporated in
vacuo to generate the crude products which were analyzed by LCMS
and ca. 1 mg of each submitted for SPR testing.
General Procedure B: Synthesis of Acids 7a−7g for Crude
Screening. To a Matrix tube was added the acid (“RCO2H” in
Scheme 2; 1 M in DMF; 5.0 μL. 5.0 μmol), the amine 5 (1 M in DMF;
5.0 μL, 5.0 μmol), and triethylamine (3.5 uL, 25.0 μmol). A solution of
COMU (1.2 M in DMF; 5.0 μL, 6.0 μmol) was added and the tubes
capped and agitated briefly to ensure mixing. After 21 h, the solvents
were evaporated (Genevac EZ2; medium bp solvent; Tmax 45 °C). The
crude esters were redissolved in MeOH (25 μL) and LiOH (1.0 M aq;
25 μL, 25.0 μmol) added. The tubes were capped, agitated briefly, and
allowed to stand at rt for 4 h. Acetic acid (1.72 μL; 30 μmol) was
added and the mixtures evaporated to dryness (Genevac EZ2;
medium−low bp mixture; Tmax 45 °C). The crude products were
redissolved in DMSO (250 μL) to a nominal concentration of 20 mM.
Libraries were made and stored as 20 mM solutions in DMSO or
DMSO-d6 at −20 °C.
ASSOCIATED CONTENT
* Supporting Information
■
S
Additional experimental for chemical syntheses, SPR data, and
equation derivations. This material is available free of charge via
the Internet at http://pubs.acs.org.
Mild “near-ambient” conditions exist for many of the most
commonly deployed reactions, enabling large libraries to be
generated without special equipment to handle difficult high
throughput chemistries (e.g., benzamide and amide syn-
thesis).14 Furthermore, as poor conversion can be tolerated
in our approach, more challenging chemistries requiring forcing
conditions to obtain good conversion, which have previously
been difficult to apply to library synthesis may also become
accessible using this approach.
AUTHOR INFORMATION
Corresponding Authors
*For J.B.M.: phone, (+44)1223895555; fax, (+44)1223895556;
E-mail, j.murray@vernalis.com.
*For P.A.B.: phone, (+44)1223895555; fax, (+44)1223895556;
E-mail, p.brough@vernalis.com.
■
Notes
The authors declare no competing financial interest.
We anticipate that further efficiency increases can be realized
by the elimination of protecting groups (which account for
around 1 in 5 of all medicinal chemistry transformations) prior
to library synthesis, in situations where derivatization of second
reactive functionalities involved directly in binding will result in
a noninterfering nonbinder, for example, N−H functions
responsible for hinge-binding interactions in kinase ligands.
We have demonstrated that the screening of crude unpurified
reaction mixtures of elaborated fragments allows the rapid
identification of compounds with increased residence times
without need for significant reaction workup and purification,
ABBREVIATIONS USED
■
kd, dissociation rate constant; ka, association rate constant; KD,
dissociation constant; ORS, off-rate screening; SM, starting
material; MeOH, methanol; COMU, 1-[(1-(cyano-2-ethoxy-2-
o x o e t h y l i d e n e a m i n o o x y ) - d i m e t h y l a m i n o -
morpholinomethylene)]methanaminium hexafluorophosphate
REFERENCES
■
(1) Congreve, M.; Murray, C. W.; Carr, R.; Rees, D. C. Fragment-
based lead discovery. Annu. Rep. Med. Chem. 2007, 42, 431−448.
2849
dx.doi.org/10.1021/jm401848a | J. Med. Chem. 2014, 57, 2845−2850

Journal of Medicinal Chemistry
Featured Article
(2) Hubbard, R. E.; Davis, B.; Chen, I.; Drysdale, M. J. The SeeDs
approach: integrating fragments into drug discovery. Curr. Top. Med.
Chem. (Sharjah, United Arab Emirates) 2007, 7, 1568−1581.
(3) Holdgate, G. A.; Gill, A. L. Kinetic efficiency: the missing metric
for enhancing compound quality? Drug Discovery Today 2011, 16,
910−913.
(4) Copeland, R. A.; Pompliano, D. L.; Meek, T. D. Drug-target
residence time and its implications for lead optimization. Nature Rev.
Drug Discovery 2006, 5, 730−739.
(5) Lu, H.; Tonge, P. J. Drug-target residence time: critical
information for lead optimization. Curr. Opin. Chem. Biol. 2010, 14,
467−474.
(6) Pan, A. C.; Borhani, D. W.; Dror, R. O.; Shaw, D. E. Molecular
determinants of drug-receptor binding kinetics. Drug Discovery Today
2013, 18, 667−673.
(7) Schmidtke, P.; Luque, F. J.; Murray, J. B.; Barril, X. Shielded
hydrogen bonds as structural determinants of binding kinetics:
application in drug design. J. Am. Chem. Soc. 2011, 133, 18903−18910.
(8) Brough, P. A.; Barril, X.; Borgognoni, J.; Chene, P.; Davies, N. G.
M.; Davis, B.; Drysdale, M. J.; Dymock, B.; Eccles, S. A.; Garcia-
Echeverria, C.; Fromont, C.; Hayes, A.; Hubbard, R. E.; Jordan, A. M.;
Jensen, M. R.; Massey, A.; Merrett, A.; Padfield, A.; Parsons, R.;
Radimerski, T.; Raynaud, F. I.; Robertson, A.; Roughley, S. D.;
Schoepfer, J.; Simmonite, H.; Sharp, S. Y.; Surgenor, A.; Valenti, M.;
Walls, S.; Webb, P.; Wood, M.; Workman, P.; Wright, L. Combining
Hit Identification Strategies: Fragment-Based and in Silico Approaches
to Orally Active 2-Aminothieno[2,3-d]pyrimidine Inhibitors of the
Hsp90 Molecular Chaperone. J. Med. Chem. 2009, 52, 4794−4809.
(9) Katsamba, P. S.; Navratilova, I.; Calderon-Cacia, M.; Fan, L.;
Thornton, K.; Zhu, M.; Bos, T. V.; Forte, C.; Friend, D.; Laird-
Offringa, I.; Tavares, G.; Whatley, J.; Shi, E.; Widom, A.; Lindquist, K.
C.; Klakamp, S.; Drake, A.; Bohmann, D.; Roell, M.; Rose, L.;
Dorocke, J.; Roth, B.; Luginbuhl, B.; Myszka, D. G. Kinetic analysis of
a high-affinity antibody/antigen interaction performed by multiple
Biacore users. Anal. Biochem. 2006, 352, 208−221.
(10) Potter, A. J.; Ray, S.; Gueritz, L.; Nunns, C. L.; Bryant, C. J.;
Scrace, S. F.; Matassova, N.; Baker, L.; Dokurno, P.; Robinson, D. A.;
Surgenor, A. E.; Davis, B.; Murray, J. B.; Richardson, C. M.; Moore, J.
D. Structure-guided design of alpha-amino acid-derived Pin1
inhibitors. Bioorg. Med. Chem. Lett. 2010, 20, 586.
(11) Roughley, S. D.; Jordan, A. M. The medicinal chemist’s toolbox:
an analysis of reactions used in the pursuit of drug candidates. J. Med.
Chem. 2011, 54, 3451−3479.
(12) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a
useful metric for lead selection. Drug Discovery Today 2004, 9, 430−
431.
(13) Verdonk, M.; Rees, D. Group Efficiency: A Guideline for Hits-
to-Leads Chemistry. ChemMedChem 2008, 3, 1179−1180.
(14) Thomas, J. B.; Fall, M. J.; Cooper, J. B.; Burgess, J. P.; Carroll, F.
I. Rapid in-plate generation of benzimidazole libraries and amide
formation using EEDQ. Tetrahedron Lett. 1997, 38, 5099−5102.
(15) Weller, H. N.; Nirschl, D. S.; Paulson, J. L.; Hoffman, S. L.;
Bullock, W. H. Addressing the Medicinal Chemistry Bottleneck: A
Lean Approach to Centralized Purification. ACS Comb. Sci. 2012, 14,
520−526.
(16) Bryan, M. C.; Dillon, B.; Hamann, L. G.; Hughes, G. J.; Kopach,
M. E.; Peterson, E. A.; Pourashraf, M.; Raheem, I.; Richardson, P.;
Richter, D.; Sneddon, H. F. Sustainable Practices in Medicinal
Chemistry: Current State and Future Directions. J. Med. Chem. 2013,
56, 6007−6021.
2850
dx.doi.org/10.1021/jm401848a | J. Med. Chem. 2014, 57, 2845−2850