Lead generation: Sowing the seeds for future success

Article abstract of DOI:10.2533/000942904777677542

Lead generation and the associated hit-to-lead process are key strategic elements in modern pharmaceutical research, and most companies have implemented this concept. Efficient lead generation is one of the main attempts to reduce the high attrition rates observed along the drug discovery process by focussing on the early developmental phases. The level of integration of the lead generation activities within the discovery organization, the flexibility in assessing and implementing new chemistries and new technologies, the high-quality standards set for the identification of the best possible chemical lead series will ultimately determine the future success in discovering new medicines.

Full text of DOI:10.2533/000942904777677542

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
588
CHIMIA 2004, 58, No. 9
Chimia 58 (2004) 588–600
© Schweizerische Chemische Gesellschaft
ISSN 0009–4293
Lead Generation: Sowing the Seeds for
Future Success
Konrad H. Bleicher, Matthias Nettekoven, Jens-Uwe Peters, and René Wyler*
Abstract: Lead generation and the associated hit-to-lead process are key strategic elements in modern pharma-
ceutical research, and most companies have implemented this concept. Efficient lead generation is one of the main
attempts to reduce the high attrition rates observed along the drug discovery process by focussing on the early
developmental phases. The level of integration of the lead generation activities within the discovery organization,
the flexibility in assessing and implementing new chemistries and new technologies, the high-quality standards set
for the identification of the best possible chemical lead series will ultimately determine the future success in dis-
covering new medicines.
Keywords: Lead generation · Lead series · Parallel synthesis · Privileged structures · Targeted libraries
1. Introduction
2. Lead Generation at Roche Basel
(Fig. 2) which handles all tasks associated
with early phase medicinal chemistry. A
project in lead generation is initiated upon
the identification of ‘Hits’ [5] either by
Lead Generation [1] as a Way to Re- 2.1. The Position of Lead Genera-
duce Late Phase Attrition Rates
tion within Discovery Chemistry
All pharmaceutical companies have ex-
Besides the two medicinal chemistry high-throughput screening (HTS) [6], ra-
perienced the painful loss of NCEs (new sections dealing with the lead optimization tional design, or inspired by literature in-
chemical entities) in late stage pre-clinical phases in the central nervous system (CNS) formation. After an intensive period of hit-
or clinical development phases. Such late- area and the cardio-vascular/metabolic dis- validation and hit-evaluation, the identifica-
stage failures are costly [2], both in terms of eases (CV/met) area respectively, we have tion of more than one lead series is our goal.
money and time [3]. A detailed analysis of created a lead generation chemistry section On average this lead series identification
the combined numbers available from both
in-house and external sources indicates that
roughly one third of the projects fail to en-
ter the lead optimization phase [4], but of
the ones that successfully pass this devel-
opmental stage, 95% never end up as drugs
on the market due the above-mentioned
late-stage failures (Fig. 1).
Efforts and measures to minimize this
attrition rates are taken at every stage of the
development. The goal of our lead genera-
tion efforts is to increase the chances for a
program transitioning into the LO-phase
(lead optimization phase) to proceed into
clinical development and ultimately to
reach ‘the market’.
*Correspondence: Dr. R. Wyler
F. Hoffmann-La Roche Ltd
Pharmaceuticals Division
Lead Generation
CH–4070 Basel
Tel.: +41 61 688 72 24
Fax: +41 61 688 64 59
Fig. 1. Average attrition rates from early discovery to market launch
E-Mail: rene.wyler@roche.com

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
589
CHIMIA 2004, 58, No. 9
successful, the ‘best’molecules are profiled
in a broad panel of enzymes, receptors and
channels to identify potential flags to mon-
itor during the later optimization phases. In
addition, we have included a series of crite-
ria centred on ‘bioavailability’ [7]. Physi-
cochemical parameters like solubility, per-
meability and lipophilicity are also mea-
sured and/or calculated in the highest
possible throughput mode to ensure the best
series with only minimal liabilities are
identified early in the hit-to-lead process.
Only in exceptional cases optimization ef-
forts will be initiated in series with low sol-
ubility (<1 µg/ml at physiological pH), as
we have experienced major difficulties lat-
er in the development of such series. In par-
allel with their physicochemical properties
the drug–drug interaction potential of all
new series is assessed by measuring
CYP450 inhibition for the most relevant
iso-enzymes. Poor or insufficient metabol-
ic stability/DMPK-profile being one of the
main causes for the high attrition rate in
drug development, metabolic stability de-
termination experiments are now routinely
conducted as part of our screening cas-
cades, and microsomal stability (and in
some cases hepatocyte-stability) is used to
guide the optimization of all our lead series.
To complete the picture, a full in vivo PK
profile for representative examples of each
series is determined. The early availability
of a suitable in vivo animal model is essen-
Fig. 2. The drug discovery process in chemistry at Roche Basel
process takes about a year for a team of 2–3 action (nature of binding: competitive, non-
chemists. At this stage the project is enter- covalent, etc.), the activity in a cell-based
ing lead optimization (LO) and within 2–3 assay (agonistic properties, antagonistic
years, a clinical candidate with the potential properties, etc.) and the selectivity towards
to be submitted to human clinical trials related or other relevant targets is a key as-
(EIH) is usually identified.
set in efficiently moving a programme
ahead. If the first optimization rounds are
2.2. The Roche Lead Series
Identified Criteria (LSI-Criteria)
Table 1. The Roche lead series identified criteria (LSI-Criteria)
To ensure a smooth transition from LI
into LO and to define this milestone, we
have implemented a set of stringent quality
criteria (Table 1). The indication and/or dis-
ease specific criteria are agreed upon at the
initiation of the project between the lead
generation chemist and the lead optimiza-
tion chemist who, at this point in time al-
ready is part of the project team, will take
over the project at the end of the lead gen-
eration phase. This has to be agreed upon
with the responsible biologists in charge of
the pharmacology. The achievement of the
LSI milestone based on the defined criteria
is peer-reviewed for maximal independence
in the assessment of the progress made
since hit-finding. However, the criteria
agreed upon at the initiation of the project
remain dynamic in nature and will, if nec-
essary and beneficial, be adapted as soon as
more knowledge becomes available. The
emphasis here lies on the delivery of high-
quality lead series which display an en-
hanced likelihood to be further progressed
into LO and beyond.
Starting from the original hits, it is usu-
ally relatively simple to obtain ‘optimized’
molecules even if the desired level of po-
tency may not be reached in all cases. Be-
ing able to determine accurately and reli-
ably, and as early as possible the mode of

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
590
CHIMIA 2004, 58, No. 9
tial for the identification of high-quality timization need to be available early on in useful entities to inhibit subtypes of the
lead series, and our LSI-criteria specify the the discovery process. A tailored screening adenosine receptor family. In another ex-
need for such a model to be established. cascade is then implemented to address the ample novel inhibitors of dipeptidyl pepti-
Safety aspects are also considered during compound specific issues with the goal to dase IV (DPP-IV) as potential new medi-
the hit-to-lead process, by identifying po- gradually transform the original hits into cines for the treatment of type 2 diabetes
tential liabilities and in silico alerts and by high-quality lead series with an overall pro- were sought. With the aid of parallel chem-
measuring selected parameters (hERG- file matching the desired target product pro- istry, high-throughput in vitro assays and
potassium channel activity, mutagenicity, file (TPP). This multi-dimensional opti- computational tools a weakly active screen-
etc.) in all series. Last but not least, and cer- mization strategy (MDO) takes advantage ing hit was rapidly developed into a lead se-
tainly earlier on, all our efforts need to fo- of the simultaneous assessment of key pa- ries of DPP-IV inhibitors that combine high
cus on patentable series. Extensive litera- rameters at the same time. This enables the activities with favourable molecular prop-
ture searches are often necessary to deter- project team to arrive at decisions whether erties.
mine the patentability of a novel lead series, to pursue or abandon certain molecules or
and this is becoming increasingly difficult. hit series more rapidly based on profound
The fierce competition among pharmaceu- data sets (Fig. 3).
3. Case Studies Within Lead
tical companies, academic institutions and
This approach of evolving the original Generation
biotechs leaves less and less space to patent hits into high-quality lead series is not al-
proprietary compounds. Navigating safely ways successful and it is preferable to aban- 3.1. Identification of Novel NK-1
around large claims from existing patents or don a series or even a project if the desired Receptor Ligands via Privileged
other publications can be a real challenge criteria cannot be met already in this early Structure-Based Library Design
and needs a high level of creativity both phase, thus allowing for the redistribution
High-throughput screening has devel-
from the chemistry and from the patenting of the resources to other series or projects, oped into a well-established methodology
side. Some of our earlier lead series had to ultimately resulting in an increased overall within the early drug discovery phase of
be abandoned due to the impossibility to se- success rate. We have also learned that the both big pharmaceutical companies as well
cure a safe intellectual property position.
identification of more than one lead series as specialized small biotechs. In fact most
Achieving all of the described require- either in parallel or in a sequential manner of the research projects where chemistry is
ments within a lead series is not a trivial en- to generate back-up series has proven bene- initiated deliver the first starting points
deavour, but the challenge to reach an LSI ficial. A switch from the first series to a from massive random screening of large
milestone has also had a beneficial impact back-up series can be triggered either compound inventories. The luxury of hit
on the way projects are selected and initiat- by unexpected difficulties in the lead lists and compound clusters identified via
ed. By raising the awareness of all scientists optimization phase or by the discovery of a HTS are highly appreciated in the chem-
involved in the process, more impact is giv- superior class of compounds in terms of istry community. Nevertheless tedious fol-
en to the ‘chemical feasibility’ and to the overall profile.
low-up work is necessary such as hit con-
‘drugability’of the target(s) of interest. The
The following section deals with recent firmation by re-ordering/synthesis of com-
resources are thus spent on chemical series examples of successful hit-to-lead chem- pounds, their quality assessment as well as
that have a higher chance of successfully istry in the area of CNS- and cardiovascu- structural confirmation to name but a few.
completing later optimization rounds.
lar-metabolic diseases chemistry. The first Lead generation chemists are dealing with
section will exemplify a strategy that was these issues on a routine basis to generate as
followed using privileged scaffolds to gen- much information as possible from an HTS
erate both hits and leads and its application campaign before follow-up chemistry is
2.3. The Optimization Process
in Lead Generation
The rapid and most complete possible in the field of G-protein coupled receptors started. High-throughput screening is par-
characterization of the hits is a key element (GPCRs). Subsequently, the endeavour to ticularly interesting for targets where no
in designing a proper optimization strategy. arrive at conclusive structure–activity rela- biostructural- or patent data is available and
The properties to be improved need to be tionships within a certain heterocyclic scaf- for such projects where the ‘unexpected’ is
identified and the corresponding assays fold is exemplified with various triazolopy- intended to add benefit on the development
with a sufficient throughput for parallel op- ridine derivatives which proved potentially of a research program. Nevertheless screen-
ing capacity is not unlimited. Although the
actual testing phase might be very short
(100 000 cpds/day) and the costs per data
point fairly low (0.1–1 US$) protein prepa-
ration, assay development, compound pur-
chasing and logistics etc. all add to the HTS
overhead. In addition, increasing com-
pound inventories both in quality and quan-
tity, constantly upcoming novel targets and
requested selectivity screening campaigns
are setting clear limits for this approach. It
is obvious that there is a need for comple-
mentary technologies which allow the initi-
ation of chemistry programs either in addi-
tion to the HTS route or without any ran-
dom screening method.
One such alternative source to generate
novel hits besides the random screening ap-
proach of huge compound collections is the
biased screening of so-called targeted li-
Fig. 3. Optimization of several parameters in parallel (MDO)
braries. The design of compound collec-

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
591
CHIMIA 2004, 58, No. 9
been discussed, a clear understanding on
the binding event of privileged structures to
their proteins is still lacking.
A number of privileged structures are
discussed as depicted in Fig. 4. Exemplified
are six different chemotypes representing
ligands for GPCRs.
One very prominent chemotype of
such privileged structures are the
spiropiperidines, which have been widely
reported in the literature. Some examples
are shown below (Fig. 5).
In contrast to the ‘privileged structure’,
the terminus ‘needle’ was described in the
literature as a fragment of an active mole-
cule showing very specific interactions
with one particular biological target. One
very well-known example for such a needle
within the GPCR area is the ortho-substi-
tuted biphenyl tetrazole (or bioisosteres),
an element present in most angiotensin-1
antagonists currently on the market [11].
In our efforts towards identifying novel
small-molecule ligands targeting the neu-
rokinin receptor (NK-1) we initiated a fo-
cused library synthesis based on the design
strategy to combine the two concepts of
‘privileged structures’ and ‘needles’. The
neurokinin receptors belong to the target
family of 7-transmembrane G-protein cou-
pled receptors [12][13]. Due to the size and
the high lipophilicity of such membrane-
bound receptors the isolation and crystalli-
sation of such target proteins is extremely
difficult. Therefore biostructure-based li-
brary design is currently limited, leading us
to employ a ligand-based library design ap-
proach. For the neurokinin receptors three
subtypes have been identified so far (NK-1,
NK-2, and NK-3). They are expressed in
the periphery (mainly NK-2) as well as in
the central nervous system (NK-1 and NK-
3). Hence their therapeutic utility ranges
from CNS indications to the potential treat-
ment of respiratory and gastric diseases.
The endogenous ligands for the neurokinin
receptors are the tachykinins, a group of
peptides that all share a common C-termi-
nal amino acid sequence Phe-X-Gly-Leu-
Met-NH2 where X is either Phe or Val. The
most prominent member of this peptide
family is the undecapeptide ‘Substance P’
(X = Phe) which shows highest affinity for
the NK-1 receptor, whereas NKA and NKB
(X = Val) are both decapeptides that bind
preferentially to the NK-2 and NK-3 recep-
tor, respectively. Although GPCRs with
such large peptide ligands as natural sub-
strates are supposed to be rather difficult to
be modulated by small molecules, several
drug-like NK-1 receptor modulators have
been reported in the literature where some
of the representatives are depicted in Fig. 6.
Besides the fact that these molecules
again represent some of the well-known
privileged structures (e.g. methylene
biphenyl- or arylpiperidines) many of the
Fig. 4. Privileged GPCR structures (highlighted in red) as promiscuous elements of protein ligands
Fig. 5. Various spiropiperidines as privileged structures for reported GPCR ligands
tions is essential in this regard to result in plied currently. Especially in the area of G-
compound arrays showing higher hit rates protein coupled receptors (GPCRs), where
for particular protein families compared to the lack of biostructural information drives
others. Where for random screening cam- alternative approaches, several successful
paigns an average hit rate of ca. 0.1–1% can projects have been disclosed [8]. Neverthe-
be observed (this is very much dependent less it is still unclear which molecular fea-
on the target screened and the threshold set) tures are required for a chemotype to result
targeted libraries must deliver much higher in a privileged structure and how such
hit rates to be competitive with the HTS ap- chemotypes are recognized by the target
proach. Various design methodologies have proteins. Although speculations about com-
been reported where the privileged struc- mon binding sites [9] and ‘first contact’mo-
tures approach seems to be most widely ap- tives at the surface of the proteins [10] have

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
592
CHIMIA 2004, 58, No. 9
As a second example we started from
the correspondingly orthogonal protected
or resin-bound spiropyrrolo-pyrrole. This
template was identified by us in the or-
phanin FQ (OFQ) area; a target protein that
belongs to the superfamily of peptide class
1 GPCRs. Again the fairly specific NK-1
needle was introduced into the scaffold at
all three possible positions 3a–c and the re-
maining vectors decorated using solution-
and/or solid-phase chemistry [15].
As already observed for the spirohydan-
toins, independent of the position of the
NK-1 needle within the spiropyrrolo-pyr-
role template various low nanomolar bind-
ing molecules were obtained. From an array
of 132 compounds submitted for testing, 91
ligands were identified with a binding affin-
ity of pKi (hNK-1) >5, resulting in a hit rate
of 69%. Fifteen representatives are depict-
ed in Table 3 (4a–o).
The two examples discussed above ex-
emplify the application of privileged struc-
ture-based library design for the generation
of small-molecule GPCR ligands. As pow-
erful as this technique can be, it is obvious
that the search for novel privileged struc-
tures and/or needles is essential. De novo
design tools such as Skelgen [16] or TOPAS
[17] have been described that allow so-
called scaffold hopping, a method to move
into novel and therefore patent-free chemo-
types. Although affinity is usually the first
indication of a valuable compound class,
other properties such as selectivity, func-
tion, solubility and many more are equally
important.
Fig. 6. Reported NK-1 receptor ligands
ligands were identified that showed decent
reported NK-1 ligands show the 3,5-bistri-
fluoromethyl phenyl motif (highlighted in
red) which is meanwhile well-recognized
as an ‘NK-1 needle’. With that information
in hand we considered generating small
compound arrays based on several
spiropiperidine scaffolds (privileged GPCR
structures) and adding the 3,5-bistrifluo-
romethyl phenyl (NK-1 needle) to further
decorate the remaining exit vectors ran-
domly. Instead of extensively exploiting the
chemotype by analoging each exit vector in
the templates we decided to either concen-
trate on one additional modification or em-
ploy only relatively small building blocks
to keep the molecular weight and the re-
sulting lipophilicity in an appropriate
range.
affinities (pKi >5) in a radioligand dis-
placement assay. The highest affinity lig-
ands generated are disclosed below (Table
2) : 2a–o. From a set of 136 compounds, 97
ligands showed a binding affinity of
pKi(hNK-1) >5 which corresponds to a hit
rate of 71%.
Table 2. Spirohydantoin as NK-1 receptor ligands
The first compound arrays were based
on the spirohydantoin template. These
compounds can be rapidly generated from
the corresponding amino acid. Starting
from commercially available 4-amino-
piperidone the orthogonally protected 4-
aminopiperidine-carboxylic acid was there-
fore generated. The NK-1 needle was either
introduced via the benzyl chloride at the α-
amino acid nitrogen 1a, the corresponding
benzoic acid at the piperidine nitrogen 1b,
or the aniline or -benzylamine at the imide
nitrogen 1c, respectively. The final com-
pounds were generated using both parallel
solution- and Merrifield-resin based solid-
phase chemistry as reported elsewhere
(Table 2) [14].
For all three possibilities where the NK-
1 needle directs to either the northern,
southern or western part of the template,

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
593
CHIMIA 2004, 58, No. 9
Table 3. Spiropyrrolo-pyrrols as novel NK-1 receptor ligands.
93% yield [21]. Conversion of the ester
functionalities to yield the bis-hydrazide 7
was achieved by treatment with hydrazine
hydrate in 90% yield in pure form after
crystallisation. Subsequently, diazotisation
of 7 was smoothly accomplished by treat-
ment with NaNO2 under acidic conditions
in 90% yield. The double Curtius re-
arrangement, was triggered by heating the
intermediate in t-butanol under reflux and
subsequent elaboration of the isocyanates
to bis N-t-butoxycarbonyl pyridine, which
was not isolated, and after removal of ex-
cess t-butanol was treated with excess TFA
in DCM to liberate the desired 2,6-diamino-
4-bromopyridine (8) in 74% yield after
crystallisation. The subsequent amination
of the pyridine-nitrogen was successfully
achieved by employing O-mesitylenesul-
fonylhydroxylamine (9) as an aminating
source. The reaction with various substitut-
ed benzaldehydes, substituted furan-car-
boxaldehydes, substituted thiophene-car-
boxaldehydes, substituted pyridine-carbox-
aldehydes as well as cycloaliphatic
aldehydes proceeded smoothly and upon
addition of aqueous KOH and exposure of
the reaction vessel to ambient air, an effi-
cient cyclisation and oxidation took place to
obtain the triazolopyridines 10 in yields up
to 79%. The final Suzuki reaction was per-
formed under the preferred method em-
degenerative diseases (e.g. Parkinson,
.
3.2. Establishment of SAR with
ploying Pd(dppf)Cl2 DCM as catalyst and
Alzheimer, depression) and therefore allow
for a new entry into the regulation of be-
havioural disorders [18]. The previously
described [19] synthesis of 7-amino-tria-
zolopyridines afforded the desired com-
pounds only in very moderate overall yields
due to the harsh reaction conditions em-
Triazolopyridine Derivatives as
aqueous sodium bicarbonate in boiling
dioxane. Boronic acids and boronic esters
worked equally well in the cross-coupling
reaction and the desired products 11 were
isolated in yields up to 49% (Scheme 1).
A library of 260 triazolopyridine deriv-
atives 11 were synthesised using the late
step Suzuki reaction as described. All prod-
ucts were tested in an in vitro radioligand
displacement assay against the human A2a
and A1 receptor. Based on these results,
partially shown in Table 4, a preliminary
structure–activity relationship was deduced
and the influence of Ar and R could be ex-
amined. Derivatives like 11a (R = phenyl)
display a moderate activity towards A2a
and the selectivity towards A1 was also in-
sufficient. In comparison derivatives with R
= furyl (11b–c) show a much higher affini-
ty towards A2a with IC50s recorded in the
low nanomolar range. The selectivity to-
wards the A1 receptor reached 2 log units,
however was found to be strongly depend-
ent on the nature of Ar. In order to circum-
vent the potential metabolic instability of
the unsubstituted furyl-derivatives poten-
tially more stable compounds like 11d–f (R
= 2-methyl-furyl) were characterised. In
general the affinity towardsA2a was dimin-
ished but the selectivity was retained. Oth-
er heterocyclic surrogates like thiophenyl
11g, pyridinyl 11h and methoxy-thiophenyl
11i–j proved to be less potent. Partially sat-
urated heterocyclic compounds like 11k–l
Potential Inhibitors of Receptors
of the Adenosine Family
In a recent medicinal chemistry project,
triazolopyridine derivatives attracted con-
siderable attention since such molecules
were identified from HTS as potential in-
hibitors of adenosine receptor subtypes.
The main focus was centered on the identi-
fication of potent inhibitors of the adeno-
sine 2a (A2a) receptor which were required
to be selective against the adenosine 1 (A1)
receptor under the assumption that they
possess the potential to modulate neuro-
(two steps)
ployed and a more generally applicable re-
action sequence was sought [20]. Cheli-
damic acid (5) can be conveniently bromi-
nated in the 4-position by phosphorous
pentabromide and upon quenching of the
intermediate with ethanol; the desired 4-
brominated di-ethyl ester 6 was isolated in
Scheme 1. Synthetic access to 7-amino-triazolopyridine derivatives 11

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
594
CHIMIA 2004, 58, No. 9
Table 4. Binding affinities towards A2a and selectivities for A1 for selected triazolopyridines 11
displayed no advantage in comparison to
the previously described compounds.
A complementary synthetic approach
allowed access to novel triazolopyridine de-
rivatives with a substitution pattern differ-
ent to the one described for the compounds
already examined [21]. To the best of our
knowledge, synthetic access to 5-aryl-8-
methoxy-[1,2,4]triazolo[1,5-a]pyridines 18
has not been described, however structural-
ly related compounds have been identified
as potential herbicidial agents [22]. There-
fore, 2-amino-3-methoxy pyridine (12) was
reacted with ethoxycarbonyl isothiocyanate
to yield almost quantitatively the thiourea
derivative 13 which was subjected to a cy-
clisation procedure, transitioning via the
proposed intermediate 14, employing hy-
droxylamine and DMAP in a protic solvent
to afford 8-methoxy-triazolopyridine 15 in
78% yield. Regioselective iodination of 15
with KIO3 in sulfuric acid yielded iodo-
triazolopyridine 16, in 49% yield, which
conveniently underwent Suzuki coupling
reactions with a set of 34 boronic acids or
esters. Reaction conditions employing
.
Pd(dppf)Cl2 DCM in dioxane with sodium
carbonate as base at elevated temperatures
furnished the 5-aryl-8-methoxy-triazolopy-
ridine derivative 17 in yields up to 82%.
The amidation of 17 with acid chlorides un-
der prolonged reaction times concluded this
five-step synthetic sequence giving access
to a range of desired triazolopyridines 18
with yields up to 61% (Scheme 2).
The obtained compounds were tested in
in vitro radioligand displacement assays
against the humanA2a andA1 receptor. Un-
fortunately, all the compounds isolated dis-
played an insufficient potency and selectiv-
ity profile. The most potent compound 18k
showed an IC50 of 110 nM with a selectiv-
ity of 36. Preliminary structure–activity re-
lationship of the new triazolopyridine deriv-
atives with such substitution pattern led to
the termination of this series (Table 5).
Scheme 2. Synthetic access to 8-methoxy-[1,2,4]triazolo[1,5-a]pyridine derivatives 18
Table 5. Binding affinities towards A2a and selectivities for A1 for selected 2-amido-5-aryl-8-
methoxy-triazolopyridine derivatives 18
Although we established
a novel,
straightforward route to aminotriazolopy-
ridines we concentrated our efforts towards
the identification of triazolopyridines with
improved binding properties. Based on ini-
tial results regarding the initially studied 7-
amino-triazolopyridines derivatives an ex-
tension of this substitution pattern was
sought. Instead of an aryl moiety in the 5-
position a carboxamide moiety was deemed
interesting instead [23]. A straightforward
reaction sequence was devised starting
from 2,6-diamino-pyridine carboxylic acid
methyl ester (19) [24]. The adduct resulting
from N-amination of 19 with O-mesity-
lene-sulfonylhydroxylamine was used di-
rectly without further purification in a one-
pot reaction for the condensation with
aldehydes. At elevated temperature the
ring-closure took place upon addition of
potassium hydroxide in methanol followed

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
595
CHIMIA 2004, 58, No. 9
physicochemical properties predictions of
the end products [25]. A virtual screening
procedure based on a topological pharma-
cophore similarity metric and self-organis-
ing maps (SOM) was developed and ap-
plied to optimising combinatorial tria-
zolopyridines 21. The starting point for this
was a set of 153 combinatorial products de-
rived from scaffold structure 21 (Fig. 3)
with known Ki values for the two adenosine
receptor subtypes A2a and A1. With this in-
formation a preliminary structure–activity
relationship (SAR) model was built, which
could be used to virtually synthesize novel
combinatorial products with improved ac-
tivity and selectivity for the A2a receptor.
For the identification of ‘best-suited’ build-
ing-blocks a novel virtual screening proce-
dure based on artificial neural networks was
followed (self-organizing maps, SOM [26]
and topological pharmacophore similarity
(CATS method) [27]):
Scheme 3. Synthetic access to 5-amino-2-aryl-[1,2,4]triazolo[1,5-a] pyridine-7-carboxylic acid amide
derivatives 21
Table 6. Binding affinities towards A2a and selectivities for A1 for selected 5-amino-2-aryl-[1,2,4]tri-
azolo[1,5-a]pyridine-7-carboxylic acid amide derivatives 21
1. Encoding the set of tested compounds
by the CATS method
2. Training of a SOM for feature mapping
(SAR model) (Fig. 7)
3. Identification of ‘seed’ compounds
from the SOM
4. Variation of the seed by virtual library
enumeration
5. Projection of the virtual library onto the
SOM obtained by step 2
6. Selection of candidates for synthesis
and testing (focused library design)
7. Chemical synthesis and determination
of in vitro activity
by the final aromatisation initialised by
opening of the reaction vessel to ambient
air to yield the triazolo-pyridine ester 20. In
total 21 substituted benzaldehydes, fur-
furals, pyridine carboxaldehydes and thio-
phene carboxaldehydes were utilised to
affect the generation of 1-aryl-substituted
triazolo-pyridine methyl esters 20 in satis-
factory yields up to 51% for the three-
step/one-pot reaction sequence. The proto-
col worked reliably for all selected aldehy-
des, however, the isolated yields of 20 were
dependant on steric and electronic features
of the aldehydes which determined their re-
activity. The conversion of the ester func-
tionality in the triazolo-pyridine derivatives
20 to form the desired amide derivatives 21
was straightforward. The most promising
protocol for a clean transformation em-
ployed conditions where methylaluminox-
ane pre-mixed with amines was reacted
with the respective esters in dioxane at 90
°C for a prolonged period of time (48–96 h)
giving access to the desired triazolopyri-
dine carboxamides 21 in yields up to 70%
(Scheme 3).
vantage of parallel solution-phase chem-
istry thus allowing for maximal flexibility
in chemistry and maximal efficiency in in
vitro biological activity optimisation. A
representative selection is shown in Table 6.
The synthesised compounds showed
varying activity/selectivity profiles and sev-
eral conclusions could be drawn regarding
the structure–activity relationship. Com-
pounds with 6-membered aromatic R moi-
eties (21a–c) were only moderately potent
and overall showed poor selectivity towards
the A1 receptor. Furyl-substituted tria-
zolopyridines like 21d–h started to yield
very active compounds with moderate to
acceptable selectivities. The influence of
steric and electronic properties of various
amines in combination with variously
substituted furyl-triazolopyridines 21d–h
could be studied and depending on the pat-
tern, compounds with low nanomolar
affinities and acceptable selectivity profiles
could be identified. Thiophenyl-triazolopy-
ridines, in general, were less active and se-
lective than their furyl-counterparts 21i–k.
However, some activity and selectivity
could be regained with thiazolyl-tria-
zolopyridine derivatives like 21l.
8. Go to step 1, or terminate
With the exception of step 7 of this
scheme, optimisation takes place in silico.
A single synthetic optimisation round
based on two seed structures which popu-
lated mainly neuron (4/3) and neuron (5/3)
(Scheme 4) was performed and a small fo-
cused library of 17 selected combinatorial
products was synthesised and tested.
On average, the synthesised compounds
based on this design cycle yielded a three-
fold smaller binding constant (~33 vs. ~100
nM) and 3.5 fold higher selectivity (50 vs.
14) than the initial library. In general, bro-
mofuryl-triazolopyridine
derivatives
21m–r show binding affinities in the low
nanomolar range with a decent to good se-
lectivity. Thiazolyl-triazolopyridine deriva-
tives 21s–v were found to be less potent, but
still moderately selective. The most selec-
tive compound 21p obtained revealed a
121-fold relative selectivity forA2a with Ki
(A2a) = 2.4 nM, and Ki (A1) = 292 nM.
Parts of the results are shown in Table 7.
This result demonstrates that it was pos-
sible to design a small, activity-enriched fo-
cused library with an improved property
profile using the SOM virtual screening ap-
proach. This strategy will be further pur-
sued and it might prove particularly useful
in projects where structure-based design
A
library of 500 5-amino-2-aryl-
[1,2,4]triazolo[1,5-a]pyridine-7-carboxylic
acid amide derivatives 21 was synthesised
in a combinatorial iterative fashion (typical
size of 24–48 members/array) taking ad-
The amines to be employed in the final
synthetic arrays were chosen according
to diversity considerations and desired

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
596
CHIMIA 2004, 58, No. 9
cannot be applied due to a lack of receptor
structure information, e.g. in the many proj-
ects aimed at finding new GPCR modula-
tors.
The synthesis of isomeric triazolopyri-
dine derivatives offered another good possi-
bility to receive a deeper insight into the un-
derlying principles which affect binding
and selectivity [28]. In this context it ap-
peared particularly interesting to observe
any changes in activity/selectivity towards
the adenosine receptors for triazolopyridine
derivatives 21 in comparison to 22 as all
main vectors extend from a structurally
very similar main scaffold (Fig. 8).
Fig. 7. Self-organising maps showing the distribution of selectivity values, [Ki (A1)/Ki (A2a)], of (a) the
initial 153-member library, and (b) the 192 virtually generated combinatorial products
By comparison of binding affinities/se-
lectivities from a set of triazolopyridines 21
versus the isomeric structures 22 electronic
properties of the main scaffold should
mainly contribute towards any adenosine
antagonism. The synthesis of the desired
triazolopyridine derivatives 22 followed
the well-established general synthetic se-
quence previously described (Scheme 5).
5,6-Diamino-nicotinic acid methyl ester
23 [29] can be reacted in a three-step one-
pot procedure with eight different aldehy-
des to furnish a small array of triazolopyri-
dine carboxylic acid ester derivatives 24 in
yields up to 52%. The desired 8-amino-
2-aryl-[1,2,4]triazolo[1,5-a]pyridine-6-car-
boxylic acid amides 22 were conveniently
accessed by reacting the respective amine
with AlMe3 for 1 h and subsequently with
the ester 24 in dioxane at elevated tempera-
tures for a prolonged period of time. In total
118 final products were obtained in yields
up to 44%. A total of twenty 8-amino-2-
aryl-[1,2,4]triazolo[1,5-a]pyridine-6-car-
boxylic acid amides derivatives 22 were
obtained which had an isomeric 5-amino-2-
aryl-[1,2,4]triazolo[1,5-a]pyridine-7-car-
boxylic acid amide derivatives counterpart
21. All of the compounds were tested
against the human A2a and A1 receptor.
Part of the results of these isomeric pairs are
displayed in Table 7 and reveal that in gen-
eral triazolopyridine derivatives 21 have a
higher binding affinity towards the A2a re-
ceptor and overall have an improved selec-
tivity profile versus A1 (Table 8).
Scheme 4. Seed struc-
tures from neurons (4/3)
and (5/3)
Table 7. Binding affinities towards A2a and selectivities for A1 for selected in silico optimised 5-ami-
no-2-aryl-[1,2,4]triazolo[1,5-a]pyridine-7-carboxylic acid amide derivatives 21
Both scaffolds are very similar with re-
spect to topology, shape and H-bond
donor/acceptor regions, but subtle differ-
ences in the H-bond donor strength of the
amino group influence the affinity towards
the A2a receptor. The molecular modelling
software package Moloc provides quantita-
tive descriptors of hydrogen bond donor
and acceptors strengths for selected atoms
[30]. When the donor and acceptor
strengths of the three nitrogen atoms with
hydrogen bond donor/acceptor capabilities
in the two scaffold series were compared
(relative to water: 1.748) it became appar-
ent that donor and acceptor strengths were
markedly changed for the amino group and
Fig. 8. Isomeric tria-
zolopyridine derivatives
21 and 22
Scheme 5. Reaction sequence towards triazolopyridine derivatives 22 and 24

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
597
CHIMIA 2004, 58, No. 9
Table 8. Binding affinities towards A2a and selectivities for A1 for selected isomeric triazolopyridine
derivatives 21 and 22
synergistic effects since it was possible to
arrive at more active and selective tria-
zolopyridine derivatives and on the other
hand allowed for further refinement of un-
derlying molecular modelling programs
through a more thorough understanding of
molecular events.
3.3. Aminomethylpyrimidines as
DPP-IV Inhibitors
In another project we sought novel in-
hibitors of dipeptidyl peptidase IV (DPP-
IV) as potential new medicines for the
treatment of type 2 diabetes. DPP-IV
cleaves and inactivates glucagon-like pep-
tide 1 (GLP-1) [33], which is an important
stimulator of insulin secretion [34]. DPP-
IV inhibition prevents the degradation of
GLP-1 and so indirectly improves the im-
paired insulin secretion in type 2 diabetic
patients [35]. In an HTS campaign of our
compound collection, we discovered
aminomethylpyrimidine 25 as a weak in-
hibitor of DPP-IV.
Fig. 9. Exemplified H-bond donor- and H-bond acceptor strengths for se-
lected atoms of 21w compared to 22a
the triazolo nitrogen atom, which is located structures, similar bioactivities’) [31]
on the same side as the amino moiety, how- should be more precisely formulated as
ever, all following the same trend. This can ‘similar recognition motifs, similar activi-
be explicitly exemplified with the two iso- ties’.
We decided to prepare a number of ana-
logues of this single hit in a parallel chem-
istry setup to quickly evaluate the potential
of this structural class for further optimisa-
tion (Scheme 6). Benzylamidine (26) was
reacted with a number of arylidene-
malononitriles 27 under basic conditions
with a subsequent oxidative step [36]. This
gave cyanopyrimidines 28 as intermediates,
which were then reduced to the desired
aminomethylpyrimidines 29. The working
procedures were amenable to parallel solu-
tion phase chemistry utilising previously
described workflows [32].
mers 21w and 22a (Fig. 9).
These lead generation activities within a
The H-bond acceptor strength of N(1) medicinal chemistry program focused sole-
of the 8-amino derivative 22a was relative- ly towards the optimisation of triazolopyri-
ly higher (approx. 0.35 units) in compari- dine derivatives in respect to their ability to
son to the N(3) H-bond acceptor strength of potently bind to theA2a receptor with a sig-
the 5-amino derivative 21w. The H-bond nificant selectivity against the A1 receptor.
donor strength of the amino group in 21w In the course of these endeavours many
was markedly higher (approx. 0.5 units) in avenues towards the synthesis of new
comparison to the H-bond donor strength of triazolopyridine derivatives have been
the amino group in 22a. From the correla- explored, established and subsequently
tion of hydrogen bond donor and acceptor utilised with the aid of parallel iterative so-
strengths of the nitrogen atoms of all iso- lution phase chemistry [32]. This approach
mers in both scaffold series it can be con- allowed for the optimal balance between
cluded that the generally more pronounced adequate number of compounds prepared
donor strength of the amino group in the 5- and the knowledge content inherent in each
amino series 21 seemed the main determi- new class. The main advantage in the ade-
nant for increased affinity and A2a/A1 se- quate utilisation of automated chemistry
lectivity in comparison to the 8-amino se- devices and procedures can be seen in the
ries 22. Although both scaffold series are shortening of compound synthesis cycle
structurally highly similar, subtle changes times which in turn allow for a more pro-
in the electron density greatly impact the found decision event whether to pursue or
molecular recognition ‘event’ and cause a terminate one compound class over the oth-
difference in biological activity. Thus, the er. The seamless integration of various mo-
molecular similarity principle (‘similar lecular modelling activities also yielded
The obtained compounds 29 were eval-
uated for DPP-IV inhibition in a fluoro-
genic in vitro assay (Table 9). We observed
that ortho- and para substitution at the 6-
phenyl ring generally increased the in-
hibitory activity, whereas meta substituents
had no effect or even decreased the in-
hibitory activity. An attempt to combine the
favourable ortho- and para substitution in
one compound resulted in the ortho-para-
dichloro derivative 30 (Table 10), which
had a 1000-fold improved activity com-
pared to the original screening hit 25 [37].

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
598
CHIMIA 2004, 58, No. 9
Scheme 6. Synthesis of 2-phenylpyrimidines 29
Table 9. DPP-IV - IC50 [µM] data for aminome-
4. Conclusion and Outlook
11), in which we replaced the 2-phenyl sub-
thylpyrimidine derivatives 29
stituent by a variety of amino substituents.
These compounds were predicted by com-
putational tools to be less lipophilic
(KOW_ClogP) [45] and less amphiphilic
(CAFCA) [46] than 30 (amphiphilicity de-
scribes the tendency of a molecule to orient
itself in an environment such as a phospho-
lipid bilayer). Since lipophilicity is a recog-
nition criterion for CYP3A4 [47], and
amphiphilicity is predictive for phospho-
lipidosis [48], we speculated that such com-
pounds would be less likely to show
CYP3A4 inhibition and phospholipidosis.
We obtained these compounds in a parallel
fashion from a common building block 31,
which was reacted with a choice of suitable
amines to give cyanopyrimidines 32
[49], which were then reduced to
aminomethylpyrimidines 33a–g (Scheme
7).
The level of integration of the lead gen-
eration chemistry unit within the rest of the
organization and its ability to actively share
information with all the partners in the lead
identification phase are crucial aspects in
discovering the best leads. Rapid feed-back
loops within the project, the ability to col-
lect, understand, integrate and interpret the
enormous amount of data generated in par-
allel since the early days of the project con-
stantly need to be optimized. At all levels in
the chain ‘small’ but essential improve-
ments have tremendously helped in this re-
spect: The implementation of a corporate
database capturing all of the above-men-
tioned data, the installation of a ‘liquid mas-
ter store’ to handle and process orders of up
to one thousand ‘cherry-picked’compounds
from the HTS-group or the lead generation
chemists to probe early SAR, the ‘geo-
graphical concentration’ of all disciplines
within one site to allow also for personal
contacts between the scientists on top of
sharing data in the electronic format, the
constant efforts to miniaturize all the meas-
urements to reduce the consumption of the
precious molecules of interest, the steady in-
crease in the throughput of most of the es-
sential tests to give just a few examples. As
demonstrated in this review, the clear crite-
ria for the identification of an LSI also con-
tribute to streamline the whole process and
to limit the time spent on ‘less promising’
series as well as to guide the target selection
towards more tractable or ‘drugable’project
even before initiation. Chemistry is still the
main driving force behind any lead identifi-
cation process and constant efforts need to
be made to continuously increase the under-
standing of medicinal chemistry aspects on
top of the general chemical and synthetic
expertise. The quality of the equipment
Table 10. Properties of 30
An evaluation of these compounds in
our fluorogenic DPP-IV inhibition assay
revealed even sub-nanomolar inhibitors
(33b, 33f, Table 11) [50]. However, we
found 33g to have the most appealing bal-
ance between inhibitory activity and calcu-
lated lipophilicity and amphiphilicity, and
profiled this compound in the above-men-
tioned in vitro assays. As documented in
Table 12, 33g is comparable to 30 with re-
gard to inhibitory activity, solubility, per-
meability, and microsomal stability. 33g
has a much reduced lipophilicity (logD)
[51], shows a sufficiently reduced CYP3A4
inhibition, and does not induce phospho-
lipidosis in cultured fibroblasts up to the
highest test concentration of 20 µM. Thus,
parallel chemistry, high-throughput in vitro
assays and computational tools were used
to rapidly evolve a weakly active screening
hit, 25, into a lead series of DPP-IV in-
hibitors, 33, that combine high activities
with favourable molecular properties.
Having shown that high inhibitory ac-
tivity can be achieved, we next turned our
attention to the molecular properties of this
structural class. Several representatives of
this series were profiled in a number of
high-throughput in vitro assays. As exem-
plified by the properties of 30 (Table 10),
the compounds had generally high solubil-
ity [38], medium to high stability in rat and
human liver microsome preparations [39],
and showed good permeation through an ar-
tificial membrane (PAMPA assay) [40].
These results indicated a potential for good
pharmacokinetic properties of this class.
However, in subsequently performed in vit-
ro assays, 30 was found to inhibit CYP3A4
[41] and to cause phospholipidosis (a phos-
pholipid storage disorder) in cultured fi-
broblasts [42].
These latter findings raised safety con-
cerns because CYP3A4 inhibition is a ma-
jor cause of drug–drug interactions [43],
and phospholipidosis can potentially lead
to a variety of adverse effects [44]. There-
fore we decided to address these issues with
a second series of compounds 33a–g (Table Scheme 7. Synthesis of 2-aminopyrimidines 33

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
599
CHIMIA 2004, 58, No. 9
Table 11. Inhibitory activity and in-silico values of aminomethylpyrimidines 33 (in comparison to 30).
Lipophilicity (KOW_ClogP) was regarded as a contributing factor to CYP3A4 interaction, and am-
phiphilicity (∆∆GAM) as a measure of phospholipidosis potential.
names appear in the references and whose con-
tributions made the described work possible
and so enjoyable. The authors also would like
to thank Dr. A.W. Thomas for helpful discus-
sions during the preparation of the manuscript.
Prof. Dr. K. Müller and Dr. G. Adam are
thanked for the donation of schemes.
Received: July 1, 2004
[1] a) A. Alanine, M. Nettekoven, A.W.
Thomas, E. Roberts, Comb. Chem. High
Throughput Screen. 2003, 6, 51; b) W.F.
Michne, Pharma News 1996, 3, 19; c) S.
Campbell, E. Differding, Chim. Nouv.
2001, 19, 3347; d) A. Tropsha, C.H.
Reynolds, J. Mol. Graph. Mod. 2002, 20,
427; e) A.D. Baxter, P.M. Locky, Drug
Discovery World 2001, 2, 9; f) D. Hunter,
J. Cell. Biochem. 2001, Suppl. 37, 22; g)
M.H. Hann, T.I. Oprea, Curr. Opinion
Chem. Biol. 2004, 8, 255; h) M.J. Sofia,
Drug Disc. Today 2004, 9, 293-295.
[2] a) J.A. Dimasi, Clin. Pharmacol. Therap.
2001, 69, 297; b) N.D. Shah, L.C. Ver-
meulen, J.P. Santell, R.J. Hunkler, K.
Hontz, Am. J. Health-System Pharm.
2002, 59, 131.
[3] a) M. Zall, Modern Drug Discovery 2001,
4, 36; b) M. Zall, Modern Drug Discovery
2001, 4, 41; c) E. Ratti, D. Trist, Il Farma-
co 2001, 56, 13; d) A.E.P. Adang, P.H.H.
Hermkens, Curr. Med. Chem. 2001, 8,
985.
[4] Parexel’s Pharmaceutical R&D Statistical
Sourcebook, 2002, 102.
[5] a) J. Knowles, G. Gromo, Nature Rev.
Drug Discov. 2003, 2, 63; b) B.R. Roberts,
Targets 2003, 2, 14.
Table 12. Properties of 33g
come vital for the success and an even more
[6] a) T. Langer, R.D. Hoffmann, Curr.
Pharm. Design 2001, 7, 509; b) A.C.
Good, S.R. Krystek, J.S. Mason, Drug
Discovery Today 2000, 5, S61; c) J. Bur-
baum, G.M. Tobal, Curr. Opin. Chem. Bi-
ol. 2002, 6, 427; d) R.M. Eglen, G. Schnei-
der, H.-J. Boehm, Methods Princ. Med.
Chem. 2000, 10, 1; e) H.-J. Boehm, G.
Schneider, Methods Princ. Med. Chem.
2000, 10, 307; f) K. Bleicher, Curr. Med.
Chem. 2002, 9, 2077.
streamlined and harmonised working
process in the future not only to access all
data that can be accessed but to get hold of
data which need to be accessed; meaningful
data which will allow for the
projection/prediction of future success of
hit series and lead series. This will enhance
the probability of ‘non’-late stage failure.
As early as in the hit/lead generation phase
it will be highly useful to encounter the
overall process and already then think about
or even address potential pitfalls in later
stages. Issues that will have to be addressed
very early on, obviously depending on the
nature of the target and the task, embrace
topics like toxicology, formulation and the
set-up of in vivo/in vitro tests which are pre-
[7] a) J. Sadowski, H. Kubinyi, J. Med.
Chem. 1998, 41, 3325; b) D.F. Veber,
S.R. Johnson, H.-Y. Cheng, B.R. Smith,
K.W. Ward, K.D. Kopple, J. Med. Chem.
2002, 45, 2615; c) P.J. Sinko, Curr. Opin.
Drug Disc. Devel. 1999, 2, 42; d) C.A.
Lipinski, J. Pharm. Tox. Meth. 2001, 44,
235; e) I. Kariv, R.A. Rourik, D.B. Kas-
sel, T.D.Y. Chung, Comb. Chem. High
Throughput Screen. 2002, 5, 459; f)
C.M. Masimirembwa, R. Thompson,
T.B. Andersen, Comb. Chem. High
Throughput Screen. 2001, 4, 245; g) J.A.
DiMasi, E. Caglarcan, M. Wood-Ar-
many, Pharmacogenomics 2001, 19,
753; h) P.J. Eddershaw, A.P. Beresford,
M.K. Bayliss, Drug Discovery Today
2000, 5, 409; i) D.B. Kassel, Curr. Opin-
ion Chem. Biol. 2004, 8, 339.
available to the chemical community, the
dictive for the clinical, human in vivo situa-
level of automation and the possibility to ac-
tion, because this is where we head.
cess novel laboratory and in silico technolo-
gies has also steadily improved recently, as
Our mind is capable of passing beyond
major investments were made (and continue
the dividing line we have drawn for it. Be-
to be made) in these fields.
yond the pairs of opposites of which the
Lead generation has matured over the
years and has become an integral part of the
drug discovery process. Since hit-to-lead
activities are unavoidably connected with
the collection, validation and interpretation
of considerably large data sets it will be-
world consists, other, new insights begin.
Hermann Hesse (1877–1962 )
Acknowledgements
It is with real pride and pleasure that we
wish to thank all our collaborators whose

PHARMACEUTICAL INDUSTRY IN SWITZERLAND
600
CHIMIA 2004, 58, No. 9
[8] K. Bleicher, L. Green, R. Martin, M. [24] E.R. v. Beck, H. Meyer, Monatshefte Chem. [39] Measurements were performed according
Rogers-Evans, Curr. Opin. Chem. Biol.
2004, 8, 287.
verw. Teile and. Wiss. 1915, 36, 731.
[25] M. Nettekoven, G. Schneider, J. Comb.
Chem. 2003, 5, 233.
to R.S. Obach, J.G. Baxter, T.E. Liston,
B.M. Silber, B.C. Jones, F. MacIntyre, D.
Rance, P. Wastall, J. Pharm. Exp. Ther.
1997, 283, 46.
[9] K. Bondensgaard, M.Ankernsen, H. Thor-
gensen, B. Hansen, B. Wulff, R. Bywater, [26] a) T. Kohonen, ‘Self-Organization andAs-
J. Med Chem. 2004, 47, 888.
[10] J. Perlman, A.-O. Colson, R. Jain, B.
Czyzewski, C. Cohen, R. Osman, M. Ger-
shengorn, Biochemistry 1997, 36, 15670.
[11] M. Burnier, H. Brunner, Lancet 2000, 355,
637.
sociative Memory’, Springer, Heidelberg, [40] Measurements were performed according
1984; b) G. Schneider, P. Wrede, Prog.
Biophys. Mol. Biol. 1998, 70, 175.
to M. Kansy, F. Senner, K. Gubernator, J.
Med. Chem. 1998, 41, 1007.
[27] G. Schneider, W. Neidhart, T. Giller, G. [41] Measurements were performed according
Schmid, Angew. Chem. Int. Ed. 1999, 38,
to D.M. Stresser, S.D. Turner, A.P. Blan-
chard, V.P. Miller, C.L. Crespi, Drug
Metab. Dispos. 2002, 30, 845 (BFC was
used as a substrate).
2894.
[12] S. Watson, S. Arkinstall, ‘The G-Protein [28] W. Guba, M. Nettekoven, B. Püllmann, C.
Linked Receptors Facts Book’, Academic
Riemer, S. Schmitt, Bioorg. Med. Chem.
Lett. 2004, 14, 3307.
Press, New York, 1994.
[42] a) K. Handrock, R. Lüllmann-Rauch, R.D.
Vogt, Toxicology 1993, 85, 199; b) R. Lüll-
mann-Rauch, R. Pods, B. von Witzen-
dorff, B. Toxicology 1996, 110, 27; c) R.J.
Mason, S.R. Walker, B.A. Shields. J.E.
Henson, M.C. Williams, Amer. Rev.
Respir. Dis. 1985, 131, 786.
[13] D.R. Flower, Biochem. Biophys. Acta. [29] K. Amin, M. Dahlstrom, P. Nordberg, I.
1999, 1422, 207.
Starke,
PCT
Int.
Appl.
1999,
[14] K. Bleicher, Y. Wüthrich, M. Deboni, S.
WO9955706A1. CAN 131:322617.
Kolczewski, T. Hoffmann, A. Sleight, [30] a)
Bioorg. Med. Chem. Lett. 2002, 12, 2519.
Gerber
Molecular
Design;
http://www.moloc.ch; b) P.R. Gerber, J.
Comp.-Aid. Mol. Design 1998, 12, 37.
[15] K. Bleicher, Y. Wüthrich, G. Adam, T.
Hoffmann, A. Sleight, Bioorg. Med. [31] M.A. Johnson, G.M. Maggiora, Eds. [43] G.K. Dresser, J.D. Spence, D.G. Bailey,
Chem. Lett. 2002, 12, 3073.
‘Concepts and Applications of Molecular
Similarity’, Wiley, New York, 1990.
Clin. Pharmacokinet. 2000, 38, 41.
[44] M.J. Reasor, S. Kacew, Exp. Biol. Med.
2001, 226, 825.
[16] D. Lloyd, C. Bünemann, N. Todorov, D.
Manalack, P. Dean, J. Med. Chem. 2004, [32] For a more detailed description of work-
47, 493.
flow procedures, workflow concepts and [45] W.M. Meylan, P.H. Howard, J. Pharm.
[17] G. Schneider, O. Clément-Chomienne, O.
Hilfiger, P. Schneider, S. Kirsch, H.-J.
Böhm, W. Neidhart, Angew. Chem. Int. Ed.
2000, 39, 4130.
automation devices utilised in the prepara-
Sci. 1995, 84, 83.
tion of compound arrays please refer to: [46] H. Fischer, M. Kansy, D. Bur, Chimia
M. Nettekoven, A. W. Thomas, Curr. Med.
Chem. 2002, 9, 2179.
2000, 54, 640.
[47] a) R.J. Riley, A.J. Parker, S. Trigg, C.N.
Manners, Pharm. Res. 2001, 18, 652; b)
A.D. Smith, H. van de Waterbeemd, D.K.
Walker, in ‘Pharmacokinetics and Metab-
olism in Drug Design’, Eds. R. Mannhold,
H. Kubinyi, H. Timmermann, Wiley-
VCH, Weinheim, 2001.
[18] a) P. Svenningsson, C. Le Moine, G. [33] a) R. Mentlein, B. Gallwitz, W.E. Schmidt,
Fisone, B.B. Fredholm, Progress in Neu-
robiology 1999, 59, 355; b) J.-L. Moreau,
G. Huber, Brain Res. Rev. 1999, 31, 65; c)
S.M. Kaiser, R.J. Quinn, Drug Discovery
Today 1999, 4, 542; d) K.C. Agarwal,
Platelet Physiol. Pharmacol. 1999, 102; e)
S.-A. Poulsen, R. J. Quinn, Bioorg. Med.
Chem. 1998, 6, 619.
Eur. J. Biochem. 1993, 214, 829; b) T.J.
Kieffer, C.H.S. McIntosh, R.A. Pederson,
Endocrinology 1995, 136, 3585; c) H.
Lene, C.F. Deacon, C. Orskov, J.J. Holst,
Endocrinology 1999, 140, 5356; d) For a
review, see J.S. Rosenblum, J.W. [48] H. Fischer, M. Kansy, M. Potthast, M.
Kozarich, Current Opinion in Chemical
Biology 2003, 7, 496.
Csato, Proceedings of the 13th European
Symposium on Quantitative Structure-
Activity Relationships, Duesseldorf, Ger-
many, Aug 27–Sept 1, 2000; Prous Sci-
ence: Barcelona, Spain, 2001.
[19] a) R. Jakob-Roetne, C. Riemer, M. Net- [34] a) J.J. Holst, Gastroenterology 1994, 107,
tekoven, W. Hunkeler, G. Huber, G. Kilpa-
trik, WO 0117999, 2001; b) R. Jakob-
Roetne, C. Riemer, M. Nettekoven, W.
1048; b) C. Orsakov, Diabetologia 1992,
35, 701; c) D.J. Drucker, Diabetes 1998,
47, 159.
[49] T. Masquelin, D. Sprenger, R. Baer, F.
Gerber, Y. Mercadal, Helv. Chim. Acta
1998, 81, 646.
Hunkeler, G. Huber, G. Kilpatrik, Chem. [35] For a review, see D.J. Drucker, Expert
Abstr. 2001, 134, 237479.
Opinion on Investigational Drugs 2003,
12(1), 87.
[20] M. Nettekoven, Synlett 2001, 12, 1917.
[50] J.-U. Peters, D. Hunziker, H. Fischer, M.
Kansy, S. Weber, S. Kritter, A. Müller, A.
Wallier, F. Ricklin, M. Boehringer, S. M.
Poli, M. Csato, B.-M. Loeffler, Bioorg.
Med. Chem. Lett. 2004, 14, 3575.
[21] M. Nettekoven, B. Püllmann, S. Schmitt, [36] A.M. Abd-Elfattah, S.M. Hussain, A.M.
Synthesis 2003, 11, 1649.
El-Reedy, N.M.Yousif, Tetrahedron 1983,
39, 3197.
[22] a) K.E. Arndt, T.C. Johnson, D.G. Ouse,
PCT Int. Appl. 2002, WO0238572A1; b) [37] J.-U. Peters, S. Weber, S. Kritter, P. Weiss,
[51] Measurements were performed according
to an in-house developed miniaturized
method to measure the shake-flask oc-
tanol-water partition coefficient at pH 7.4
(logD).
T.C. Johnson, R.J. Ehr, R.D. Johnston,
W.A. Kleschick, A. William, T.P. Martin,
M.A. Pobanz, J.C. van Heertum, R.K.
A. Wallier, M. Boehringer, M. Hennig, B.
Kuhn, B.-M. Loeffler, Bioorg. Med.
Chem. Lett. 2004, 14, 1491.
Mann, US PatentAppl. 1999, US5858924; [38] Measurements were performed according
c) J.C. van Heertum, W.A. Kleschick, K.E.
Arndt, M.J. Costales, R.J. Ehr, K.B.
Bradley, W. Reifschneider, Z.L. Benko,
M.L. Ash, J.J. Jachetta, PCT Int. Appl.
1996, WO9601826A1.
to an in-house developed method to assess
solubility from a 10 mM DMSO stock solu-
tion. This method is similar to the classical
thermodynamic shake-flask solubility with
the only difference that DMSO is removed
in the beginning by an additional lyophiliza-
tion step. Therefore this assay is called
LYophilisated Solubility Assay (LYSA).
[23] B. Brodbeck, B. Püllmann, S. Schmitt, M.
Nettekoven, Tetrahedron Lett. 2003, 44,
1675.