Promiscuous 2-aminothiazoles (PrATs): A frequent hitting scaffold

Article abstract of DOI:10.1021/jm501402x

We have identified a class of molecules, known as 2-aminothiazoles (2-ATs), as frequent-hitting fragments in biophysical binding assays. This was exemplified by 4-phenylthiazol-2-amine being identified as a hit in 14/14 screens against a diverse range of

Full text of DOI:10.1021/jm501402x

Subscriber access provided by SELCUK UNIV
Article
Promiscuous 2-Aminothiazoles (PrATs): A Frequent Hitting Scaffold
Shane M. Devine, Mark D Mulcair, Cael O Debono, Eleanor W.W. Leung, Willem Nissink,
San Sui Lim, Indu R. Chandrashekaran, Mansha Vazirani, Biswaranjan Mohanty, Jamie Scott
Simpson, Jonathan B. Baell, Peter J. Scammells, Raymond S Norton, and Martin J Scanlon
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/jm501402x • Publication Date (Web): 05 Jan 2015
Downloaded from http://pubs.acs.org on January 16, 2015
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry is published by the American Chemical Society. 1155
Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 29
Journal of Medicinal Chemistry
1
2
3
Promiscuous 2-Aminothiazoles (PrATs): A Frequent Hitting Scaffold
4
5
6
7
8
9
Shane M. Devine,A Mark D. Mulcair,A Cael O. Debono,A Eleanor W. W. Leung,A J.
Willem M. Nissink,B San Sui Lim,A Indu R. Chandrashekaran,A Mansha Vazirani,A
Biswaranjan Mohanty,A Jamie S. Simpson,A Jonathan B. Baell,A Peter J. Scammells,A
Raymond S. Norton,A Martin J. ScanlonA*
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
A
Medicinal Chemistry, Monash Institute of Pharmaceutical Sciences, Monash University, Parkville, VIC 3052,
Australia
B Oncology iMed, AstraZeneca, Alderley Park, Nether Alderley, Macclesfield SK10 4TG, United Kingdom
* Corresponding author. Email: martin.scanlon@monash.edu; Tel: +61 3 99039540
ABSTRACT
We have identified a class of molecules, known as 2ꢀaminothiazoles (2ꢀATs), as frequentꢀhitting
fragments in biophysical binding assays. This was exemplified by 4ꢀphenylthiazolꢀ2ꢀamine being
identified as a hit in 14/14 screens against a diverse range of protein targets, suggesting that this
scaffold is a poor starting point for fragmentꢀbased drug discovery. This prompted us to analyze this
scaffold in the context of an academic fragment library used for fragmentꢀbased drug discovery
(FBDD) and two larger compound libraries used for highꢀthroughput screening (HTS). This
analysis revealed that such “Promiscuous 2ꢀAminoThiazoles” (PrATs) behaved as frequent hitters
under both FBDD and HTS settings, although the problem was more pronounced in the fragmentꢀ
based studies. As 2ꢀATs are present in known drugs, they cannot necessarily be deemed
undesirable, but the combination of their promiscuity and difficulties associated with optimizing
them into a lead compound, make them, in our opinion, poor scaffolds for fragment libraries.
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 29
1
2
INTRODUCTION
3
4
5
6
7
8
9
Fragmentꢀbased drug discovery (FBDD) is becoming a widelyꢀused technique in drug discovery as
part of a medicinal chemists arsenal.1,2 FBDD strategies utilise small molecules (typically ~200
Da), known as ‘fragments’, that often have relatively low affinities for the target of interest.
Nonetheless, numerous FBDD campaigns have demonstrated that it is feasible to elaborate
fragment hits to achieve clinically useful compounds.3 Furthermore, there are published examples
where FBDD has enabled the development of potent compounds against targets where HTS of a
large library did not yield any useful hits4 and a review of internal projects by scientists at
AstraZeneca revealed that FBDD represents a powerful tool to assess the likelihood of finding
highly potent ligands for any given target.5 This highlights one attraction of FBDD, which is that
fragment libraries contain relatively small numbers of compounds (a few hundred to thousands) but
are very effective at finding hits because they are able to sample chemical space more efficiently
than the larger compounds that are typically found in HTS libraries.6 One way of illustrating this is
to consider that a library of drugꢀlike molecules of 30 heavy atoms might need up to 1060 members
to efficiently cover chemical space, whereas this number is significantly smaller – around 107 – for
a library of fragments with around 12 heavy atoms.7,8 As fragments are smaller, they typically bind
with lower affinity to the relevant target, with KD values ranging from high micromolar to
millimolar. In contrast, drugꢀlike hits from HTS typically have KD values in the high nanomolar to
low micromolar range, although in both cases the binding energy per heavy atom (or “ligand
efficiency”) can be comparable. Fragments can also be developed into lead compounds that are
smaller and less lipophilic than those generated from HTS.9 A highlight of FBDD to date was the
development of Vemurafenib, a BRAF kinase inhibitor used for the treatment of lateꢀstage
melanoma, which, via medicinal chemistry optimisation, became the first FBDDꢀderived compound
to reach the clinic.10,11 Since FBDD libraries are usually small, it is essential to ensure that the
library is populated with high quality fragments. Some of the key considerations in the design of
fragment libraries have been discussed previously.12,13
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
ACS Paragon Plus Environment

Page 3 of 29
Journal of Medicinal Chemistry
1
2
3
4
5
PAINS
6
7
8
9
A matter of growing concern with screening libraries, for FBDD or otherwise, is the inclusion of
compounds that may act promiscuously and display activity both across different target classes and
via a number of different assays or biological readouts. These types of compounds, which are
widely referred to as Pan Assay INterference CompoundS (PAINS), were first described by Baell
and Holloway14 and are identified by the presence of substructural features that promote frequentꢀ
hitting behavior. It has been suggested that they should be excluded from screening libraries for that
reason. Whilst in many cases PAINS may appear to give optimizable hits, elaboration often results
in flat or confusing structureꢀactivity relationships (SAR).15 The reasons for their promiscuity are
varied and include the presence of potential Michael acceptors, chelation, redox activity, and strong
chromophoric interference.16 Some PAINS scaffolds have easily identifiable problems associated
with them, but the chemical basis of the observed promiscuity for many PAINS is unknown.
Confounding the complexity of PAINS identification is the fact that there are a small number of
cases where seemingly “unprogressable” PAINS have in fact been developed into potent and
selective molecules. One example is the PI3Kγ inhibitor (ASꢀ604580) which is based on an
alkylidene rhodanine hit containing this known PAINS motif.17 However, the fact that certain
PAINS or PAINSꢀlike motifs are present in some potent and selective molecules does not imply
that the PAIN is a viable starting point. In general, the chances that a PAIN will be progressed to a
useful lead compound are overwhelmingly smaller than the chances it will not.16
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
There are fourteen subꢀclasses of 2ꢀATs that have been categorized as PAINS.14,16 There are a
number of possible reasons for their reported promiscuity, such as their potential photoreactivity or
the presence of impurities generated by their chemical precursors, for example where 2ꢀATs are
prepared from bromomethylketones.14 2ꢀATs have also been identified as potentially thiolꢀreactive,
which is another mechanism that produces problematic screening hits. For example a subclass of 2ꢀ
3
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 29
1
2
3
4
ATs was identified in the ALARMꢀNMR assay as being thiolꢀreactive,18,19 although in this case it is
possible that their reactive precursors were responsible.20
5
6
7
8
9
That 2ꢀATs can be progressed to generate useful compounds is testified by a number of marketed
drugs, including antibiotics such as carumonam, cefcapene, cefdinir, cefditoren, cefepime,
cefetamet, cefoselis, cefotaxime, cefotiam, cefpodoxime, cefpirome, ceftazidime, ceftibuten,
ceftriaxone; talipexole and pramipexole, dopamine agonists for the treatment of Parkinson’s
disease; mirabegron, a β3ꢀadrenoceptor agonist used to treat overactive bladder; and riluzole, a 2ꢀ
aminobenzothiazole used to treat acute myeloid leukaemia (AML). Conversely, 2ꢀATs have
displayed cytotoxicity and metabolic instability as antimycobacterial and antiplasmodial agents.21
Thus, the value of 2ꢀATs in screening collections is currently unclear. Herein, we describe our
efforts to determine if 2ꢀATs are promiscuous binders by both FBDD and HTS techniques.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
RESULTS
FBDD Screening
FBDD screens were undertaken at the Monash Institute of Pharmaceutical Sciences (MIPS) using a
library of 1137 fragments comprising molecules that pass both biophysical and chemical filters, are
chemically and structurally diverse, are soluble at 1 mM in aqueous buffer, and can be chemically
elaborated from readily accessible precursors.12,22 Chemical filters that were applied in assembling
the fragment library include the removal of PAINS,14 unwanted functionality13 and reactive
groups.23 A review of 14 fragment screening campaigns with this library, in which the primary
readout of fragment binding was saturation transfer difference nuclear magnetic resonance
spectroscopy (STDꢀNMR),24 revealed that at least one 2ꢀAT from the library had been identified as
a hit in every case.12 This led to our investigation of their role as potential promiscuous binders
using an orthogonal biophysical technique. We report here the results of our study and analysis of
all 2ꢀATs in our fragment library (Figure 1) undertaken using surface plasmon resonance (SPR)
4
ACS Paragon Plus Environment

Page 5 of 29
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
against six different protein targets. Target proteins were the Plasmodium falciparum apical
membrane antigen AMA1,22 the E3 ubiquitin ligase adapter protein SPSB2,25 two DsbA
oxidoreductases from different bacterial species (oxidoreductase 1 and 2), carbonic anhydrase II,
and a kinase. These proteins were selected as they exhibit diversity in function and have little
structural similarity across their binding sites. All except oxidoreductase 2 had known positive
controls that were used in the SPR assays. In addition, the fragment screen contained a number of
internal controls. Four fragments used in this study were identified previously as binders to SPSB2
by SPR (1, 3, 15, 17; our unpublished data), carbonic anhydrase II was expected to show a very
strong preference for compounds containing a terminal sulfonamide and a prior fragmentꢀscreen of
oxidoreductase 2 (our unpublished data) had revealed only very weak binding for any fragment.
Thus, oxidoreductase 2 served as a negative control.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. 2ꢀAminothiazole (2ꢀAT) and 2ꢀATꢀlike compounds in the MIPS fragment library12
5
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 29
1
2
3
4
5
6
7
8
Doseꢀresponse determinations were undertaken for positive controls (available for all proteins
except oxidoreductase 2) and gave KD values that were within the expected ranges for each target
protein (Supplementary Figures S1 and S2), confirming that all proteins were stable and active on
the sensor chip under the conditions used. Binding of 2ꢀATs was carried out at a single
concentration (200 µM) in duplicate, and the average raw response was converted to percentage
occupancy of binding, corrected for the molecular weight of each compound (Figure 2). The
sensorgrams for selected 2ꢀATs can be found in Supplementary Figures S3–5. Injections of buffer
blank were included between each compound, both to provide for double referencing and to
minimise carryꢀover problems from poorlyꢀbehaved compounds.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
200
175
150
125
100
75
50
25
0
1
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
AMA1
SPSB2
CAII oxidoreductase 1 kinase oxidoreductase 2
Fragment
Figure 2. Binding of 2ꢀATs to AMA1, SPSB2, CAII, oxidoreductase 1, a kinase and
oxidoreductase 2. Compounds were tested at a concentration of 200 µM, and the raw response
converted to percentage occupancy.
Figure 2 reveals a consistency in the pattern of protein binding. Only two 2ꢀATs (1 and 15)
demonstrated any binding to oxidoreductase 2, which is consistent with previous observations that
this protein has very limited capacity to bind small molecules. For the other five proteins, there
were two cases (fragment 8 and 15) where a large range in the fractional occupancy was observed,
6
ACS Paragon Plus Environment

Page 7 of 29
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
ranging from ~0% to > 150%. Otherwise, most fragments showed little discrimination in binding,
with fragments typically binding to either all five proteins or none at all. While there were some
exceptions, there was also a general tendency for fragments to bind each of the five proteins at
similar occupancy, suggesting approximately the same affinity and thus providing very little clear
SAR.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
To examine the SAR in a more systematic manner, 2ꢀATs from the MIPS library were clustered
based on linear fingerprints and key chemical features, and the screening results examined for each
grouping. Compounds 1 – 7 feature 4ꢀaryl substitutions with a free 2ꢀamino group, with or without
a substituent at the 5ꢀposition. Fragments 8 – 12 maintain the free 2ꢀamino group, with 8 and 9
containing fused aliphatic rings, whereas 10 – 12 have aliphatic substituents at the 4ꢀposition.
Fragments 13 – 15 integrate benzothiazoles with their 2ꢀamino group intact, while fragments 16 and
17 have morpholino or piperazino attachments through their 2ꢀamino group, respectively.
Fragments 18 – 24 are 2ꢀamido containing thiazoles and 27 and 28 are thiazolo[3,2ꢀa]pyrimidines,
in which the 2ꢀamino group is incorporated in the heterocycle. The triazolothione (25) and
thiazolium (26) complete the selection of 2ꢀATs from the MIPS library. These molecules contain
diverse substituents around the 2ꢀAT core. Thus, despite the chemical diversity in the 28 2ꢀATs in
the library, evaluation of their binding to the different targets provided little clear SAR.
As an example, 4ꢀphenylthiazolꢀ2ꢀamine (1) highlights the problem with this class of molecule,
showing binding to all six proteins examined, all with similar occupancy levels (~25%). As this
fragment contains 12 heavy atoms this level of binding corresponds to a favorable ligand efficiency
(i.e. binding energy per heavy atom) in each case.26 Thus, if this fragment were tested in an isolated
setting against one particular protein, the likelihood of follow up by medicinal chemists would be
high, demonstrating the insidious behavior associated with this class. To rule out an impurity in the
commercial preparation of fragment 1, it was resynthesised and purified. This gave similar
7
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 29
1
2
3
4
5
6
7
8
occupancy levels to the commercial product across all protein targets, indicating that the problem is
inherent to the molecule itself, as opposed to reactive precursors or sideꢀproducts potentially present
in the commercial source. Furthermore, the initial STDꢀNMR screening data demonstrated binding
of 1 rather than precursors or sideꢀproducts and also revealed that the binding was nonꢀcovalent
(Supplementary Figure S6). This rules out several of the potential mechanisms for promiscuity,
including photoreactivity, thiol reactivity and the presence of reactive precursors.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Fragments 2 – 6 show binding to all of the proteins examined except oxidoreductase 2. The addition
of the 5ꢀmethyl group appears to have little effect, as does the nature of the substituents on the 4ꢀ
aryl ring, which include hydroxy, methoxy and fluoro moieties. The addition of carboxylic or amido
containing chains (7, 10, 11) at the 4ꢀ or 5ꢀposition, coupled with saturation of the 4ꢀaryl ring (12),
seems to abrogate promiscuity, although, of these fragments, only 12 is observed to bind to any of
the proteins at an occupancy > 10% and 12 binds only to oxidoreductase 1. Of the benzothiazole
fragments with a 2ꢀamino group, 13 does not bind CAII or oxidoreductase 2, whereas 14 does not
bind to oxidoreductase 1 or 2. Fragment 8 binds to three proteins, however occupancy levels >
100% of the theoretical Rmax are observed for two proteins (AMA1 and SPSB2). Fragment 15 is the
only sulfonamide in the series and demonstrates affinity to CAII, as expected for this protein since
it specifically binds terminal sulfonamides, but several other fragments (for example 3, 8, 13) also
demonstrate binding to CAII.27 The 2ꢀsubstituted aminothiazoles are an interesting test case;
morpholino (16) shows little to no binding to any protein tested, whereas piperazino (17) binds to
five of the six proteins. The 2ꢀamido and heterocyclic thiazoles (18 – 24, 27 and 28) show little or
no binding to any of the proteins and do not appear to be inherently promiscuous. Triazolo thione
25 and thiazolium 26 showed no binding to five and four of the targets, respectively. In summary,
2ꢀATs appear to be generally promiscuous at fragment screening concentrations whereas the
corresponding amides are not. Within the set of 2ꢀATs, there are no clear characteristics that
8
ACS Paragon Plus Environment

Page 9 of 29
Journal of Medicinal Chemistry
1
2
3
4
distinguish the more promiscuous binders from their less promiscuous counterparts, and thus there
is no clear SAR.
5
6
7
8
9
In order to extend the SPR a sample of 2ꢀATs were evaluated for binding to oxidoreductase 1 using
2D heteronuclear single quantum correlation (HSQCꢀNMR) studies. Oxidoreductase 1 was chosen
for this analysis as it is relatively small (21 kDa), expressed at high levels in bacteria and amenable
to isotope labeling and analysis by proteinꢀdetected NMR. Four of the fragments (1, 5, 6 and 14)
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
were tested for binding to oxidoreductase 1 by recording Hꢀ15N HSQC NMR in the presence and
absence of each fragment. Binding was analysed by measuring chemical shift perturbations (CSPs)
in the spectrum of the protein upon addition of the fragment (Figure 3).
9
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 29
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3. (AꢀD) HSQC CSPs versus sequence of oxidoreductase 1 at 125 ꢁM with fragment
concentrations at 1 mM for 1, 5, 6 and 1.5 mM for 14. Front (E) and back (F) views showing the
extent of CSPs of 14 with oxidoreductase 1 (dark red to light red).
10
ACS Paragon Plus Environment

Page 11 of 29
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
The SPR data revealed that 1, 5 and 6 showed similar levels of binding to the protein, whereas no
binding was observed by SPR for 14. In contrast, the extent of CSP observed in the NMR data
suggested that whilst 1, 5 and 6 showed similar extents of CSP, fragment 14 showed the largest
CSP in the spectra. This would normally suggest that 14 is the fragment that binds to
oxidoreductase 1 with the highest affinity. However, analysis of the location and extent of CSP on
the structure of oxidoreductase 1 (Figure 3E and 3F) revealed clusters of strong CSP at distinct sites
on the protein, and a concentration dependence of the CSP that was inconsistent with 1:1 binding
stoichiometry (Supplementary Figure S7). Taken together the NMR data suggest that 14 binds to
oxidoreductase 1 at more than one site. For fragments 1, 5 and 6, the smaller CSP observed make it
difficult to determine whether they are binding at more than one site.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Proteinꢀdetected NMR was also used to investigate the binding of 2ꢀATs to a second test protein
SPSB2. In this case, 19FꢀNMR studies on 5ꢀFꢀTrpꢀSPSB2 have proven effective as an analytical
tool for determining binding to the active site of the protein.28 Specific binding of peptides and
ligands to the functionally important binding site is characterised by a substantial downfield shift of
the fluorine resonance corresponding to Trp207 (Figure 4). In contrast, no downfield shift of this
peak was observed for the 2ꢀATs that were observed to bind SPSB2 by SPR, despite their
reasonable binding occupancy at 200 µM. This suggests that these 2ꢀATs do not bind to the active
site of SPSB2, and that binding probably occurs at a secondary site on the protein. The ability to
bind to more than one site on a protein may underpin the promiscuity that we have observed with
some 2ꢀATs.
We have previously reported analysis of binding of a small set of 2ꢀATs to AMA1.22 In the case of
AMA1, 4ꢀphenylthiazolꢀ2ꢀamine (1) was identified as a hit in the primary screen and a series of 4ꢀ
aryl substituted 2ꢀATs was synthesised to generate preliminary SAR. The structures and percentage
occupancy data of an extended series of 4ꢀaryl substituted 2ꢀAT analogues and their binding to
11
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 29
1
2
3
4
5
6
7
AMA1 as determined by SPR are presented in Supplementary Figures S8 and S9. All of the
analogues were found to bind with low affinity to AMA1 and no clear SAR was derived from the
series.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 4. 19FꢀNMR spectra of 5ꢀFꢀTrpꢀSPSB2 alone (bottom), with 500 µM control peptide
(middle), and 3 mM 1 (top). Specific binding of the control peptide to the active site is characterised
by a downfield shift of the peak corresponding to Trp207.25 No such shift is observed for the
thiazole fragment, suggesting a different mode of binding to SPSB2.
HTS Screening – Academic and Corporate Collections
We were interested in assessing the apparent promiscuity of 2ꢀAT derivatives in fragmentꢀbased vs
highꢀthroughput screening. To analyse the latter, we adopted two approaches. The first was to
analyse the relative prevalence of 2ꢀATꢀbased PAINS in a HTS library of 93,000 compounds
housed at the Walter and Eliza Hall Institute (WEHI).29 The second was to analyse the full
AstraZeneca HTS screening deck. With respect to the first approach, a search of the WEHI HTS
library revealed 989 2ꢀATꢀbased compounds and an additional 1012 compounds where the amine
was acylated. Amongst these are 14 subꢀclasses of PAINS, the structural definitions of which are
12
ACS Paragon Plus Environment

Page 13 of 29
Journal of Medicinal Chemistry
1
2
3
4
given in Supplementary Figure S10. This analysis was concordant with data from our fragment
library in that no PAINS class contains 2ꢀATs in which the 2ꢀamino group is acylated.
5
6
7
8
9
The incidence of frequent hitting 2ꢀATs was then investigated with greater statistical power using a
large corporate collection (AstraZeneca, January 2014). These compounds have been tested in a
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
range of HTS campaigns with concentrations typically around 10 µM. To assess the promiscuity of
2ꢀATs we employed a descriptor that has been designed to flag potential frequentꢀhitter behavior
for all compounds with sufficient data.30 This descriptor is calculated by first designating a
compound as active/inactive in each available screen using data from all historical AstraZeneca
HTS campaigns. The body of HTS data that is available varies from compound to compound for a
variety of reasons (age of compound, sample availability, membership of screen sets, manual
collation), so potentially anomalous binding behavior was designated by comparing observed
incidence of activity for a given compound to the expected activity for an ‘average’ compound. In
this way a descriptor (termed the pBSF score) was derived that indicates whether the compound is
more active than expected based on historical observations.30 This knowledgeꢀbased descriptor is
calculated as the negative logarithm of the probability of encountering the observed level of activity
for a compound that behaves normally by chance. A low probability, and hence a high score,
indicates that it is very unlikely that the observed level of activity would be observed for a wellꢀ
behaved compound, which therefore suggests that the compound is a promiscuous hitter. We first
analysed the number of 2ꢀATs that had a pBSF score above a threshold score of 2.0. Secondly, we
divided the 2ꢀATs into classes based on substructures, and determined the proportion of each
substructure that had pBSF scores > 2.0.
A comparison was made of the pBSF score for the 61,040 2ꢀATs within the AstraZeneca collection.
For a visual comparison, a set of the same size was selected at random from the library, as
documented in Figure 5. For the 2ꢀATs there were 6,122 compounds with a score > 2 (incidence of
13
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 29
1
2
3
4
5
6
7
8
10.0%), whereas a randomly picked set of the same size has, on average, 3953 ± 61 compounds
with a score > 2 (incidence of 6.5% ± 0.1%, as determined by a bootstrapping analysis). It is
evident that the incidence of frequent hitters in the random set is significantly lower than for the 2ꢀ
ATs (confidence P>99.9%, as derived from bootstrapping analysis), which suggests that the 2ꢀATs
show a tendency to be promiscuous that is significantly higher than seen in a diverse set of
randomlyꢀselected compounds.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 5. Rankꢀordered frequent hitter scores for the first 6,122 2ꢀATs (red line) from a total set of
61,040 2ꢀATs retrieved by substructure from the AstraZeneca screening collection. Dotted lines
show the numbers of compounds at a threshold value > 2, which is indicative of anomalous binding
behavior.
We then divided the 2ꢀAT structural classes using the substructures shown in Table 1 and counted
the incidence of suspicious compounds using a threshold of > 2 to distinguish potential frequent
hitters from ‘clean’ compounds and those lacking sufficient data.
14
ACS Paragon Plus Environment

Page 15 of 29
Journal of Medicinal Chemistry
1
2
3
4
5
6
Table 1. Nine simplified 2ꢀAT classes were inspected for promiscuous behavior across the
AstraZeneca corporate collection. Counts are shown for subsets of small molecules (MW < 300)
and larger compounds (MW ≥ 300).
Class
Substructurea
7
8
9
N
N with
data
MW
N with
data MW
>=300
N(FH)
MW
<300
N(FH)
MW
>=300
fraction
pBSF>2
LowMW
fraction
pBSF>2
highMW
<300
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
77,826
8,070 52,970
559
17
5,563
144
6.93
10.50
1
aminothiazoles
3,259
284
2,190
4,425
5.99
6.58
2
fused
aminothiazoles
8,678
55,445
13,864
1,781
143
372
53
564
8.03
6.97
5.42
12.75
10.36
10.36
3
4
5
primary
amines
5,336 38,337
3,970
1,068
secondary
amines
978
10,312
tertiary amines
3,517
304
661
2.795
7,495
15
37
108
960
4.93
5.60
3.86
6
7
tertiary
amines,
acyclic
10,307
12.81
tertiary
amines, cyclic
59
21
25
0
0
0.00
3.79
0.00
7.73
8
9
thiazole cyclic
amines
(‘embedded’)
29,816
2,613 21,447
99
1,657
acylated
thiazole
amines
a Structure legend: A, any atom; rn, ring bond; ch, chain bond; sn, substitution count n; rn, number
of ring bonds
The data in Table 1 are consistent with the patterns observed with the MIPS FBDD and WEHI HTS
data, namely that 2ꢀATs show a higher incidence of anomalous binding behavior across a number
15
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 29
1
2
3
4
5
6
7
8
of substructure classes. Typically, structures across the AstraZeneca collection that have on average
≤ 6.0% of the compounds within the class with a pBSF score > 2.0 are not considered to be frequent
hitters. Here, an elevated incidence (approx. 10%) of such behavior can be seen for the overall class
of 2ꢀATs and for subclasses 1, 3, 4, 5 and 7, which is in line with observations for the WEHI HTS
library.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
It is notable that promiscuity can be observed in the HTS data even within sets of lowꢀMW
compounds For example, within class 3, there are 1781 primary amines with MW < 300, of which
8% are classified as frequent hitters. Although these compounds are “fragmentꢀlike” in terms of
their size, they are observed to be promiscuous in the HTS data despite being screened at a typical
HTS concentration of 10 µM, which suggests that their promiscuity in the FBDD data is not due
solely to the higher concentrations used in the fragment screen.
Subclass 9 comprises the acylated 2ꢀATs, which show a lower incidence of frequentꢀhitter
behaviour than the entire class of 2ꢀATs both for the lowꢀMW and highꢀMW set. Nonetheless, the
incidence of frequentꢀhitter behavior in the highꢀMW set is somewhat higher than the expected
“normal” level of 6.0%, although this can be explained by biases that are apparent in targetꢀspecific
analyses of the screening data. Inspection of a targetꢀspecific frequent hitter descriptor (results not
shown) reveals that subclasses 3, 4 and 9 show an increased incidence of kinase activity. As kinases
are typical drug targets pursued by pharma companies, a bias may be observed in the data where
kinaseꢀlike motifs have been screened preferentially in kinaseꢀtargeted screens. This skews the
descriptor results based on such data to some extent, i.e. some of the frequent hitters identified may
actually be kinaseꢀfrequent hitters, and not necessarily problematic. Within the 1,657 potential
frequent hitters of kinaseꢀlike class 9 (MW≥300), a subset of 556 molecules appears to be kinaseꢀ
frequent hitters. Subclass 7 contains a structural motif less likely to hit kinases, yet appears to have
a high frequentꢀhitter incidence and contains 960 suspicious compounds (MW≥300). Of these, a
16
ACS Paragon Plus Environment

Page 17 of 29
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
smaller fraction of only 136 display kinaseꢀrelated frequent hitter behavior, suggesting that the
origin of frequentꢀhitting behavior in the remaining compounds is mostly not related to kinases. We
suggest therefore that the somewhat increased incidence of frequent hitters we observe for the
acylated 2ꢀATs in this data set is an artefact resulting from a kinaseꢀactivity bias, and that the
acylated 2ꢀATs as a whole do not meet the criteria to be categorised as frequent hitters.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Using the AstraZeneca 2ꢀAT set of 61,040 compounds and data derived from the AstraZeneca
corporate collection, we investigated overall correlations of frequent hitter incidence with
lipophilicity (experimental logD, clogP), experimental solubility (pSol, calculated as log[Sol/uM] ),
calculated polar and nonꢀpolar surface area (PSA and NPSA, in Å2 and %PSA in %), and molecular
volume.31,32 Correlations are observed with ion class, PSA, %PSA, as well as donor and acceptor
counts (Supplementary Figure S11), suggesting that there could be a relation with polarity (as each
of these properties reflects polarity and they are interrelated). A weak trend with logD
(experimental octanolꢀwater partitioning coefficient) was observed, with lowꢀlogD compounds
somewhat more likely to be frequent hitters. No relation is observed with experimental aqueous
solubility, but a trend is evident with increasing clogP. Although there are no categorical reasons for
promiscuity in these compounds, the observation that ionisable groups and high lipophilicity may
increase anomalous behavior in the 2ꢀAT class is in line with recent observations by Tarcsay and
Keserű,33 who observed similar trends.
DISCUSSION
We have demonstrated that certain members of the class of compounds containing a 2ꢀAT
substructure are frequent hitting and promiscuous fragments in the context of FBDD, where
screening is undertaken using biophysical binding assays. We have dubbed these fragments
Promiscuous 2ꢀAminoThiazoles (PrATs). Analysis of binding of the 28 2ꢀATs in the MIPS
fragment library was undertaken by SPR and NMR against six unrelated protein targets. Although
some patterns have emerged, such as the presence of the free 2ꢀamino group seeming to promote
17
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 29
1
2
3
4
5
6
7
promiscuity, a clear mechanism of action has not been identified at this point. Nonetheless, our SPR
analysis produced flat and confusing SAR against several of the targets, which has previously been
noted in the characterisation of PAINS identified in HTS.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
As several approved drugs contain a 2ꢀAT, it is clearly possible to accommodate this structure in a
therapeutically useful compound. However, it is noteworthy that a number of PAINS subclasses
also contain the 2ꢀAT chemotype, which suggests that the 2ꢀAT may carry some risks if selected for
development. For example, in common with many aromatic amines, 2ꢀATs can be Ames positive
dependent on the substitution pattern. Further, aminothiazoles have been associated with liver
toxicity via bioactivation of the double bond, leading to formation of thioureas that are further
bioactivated. Problems relating to reactivity and covalent modification, however, are unlikely to
contribute to the observed promiscuity in biophysical binding assays, as the ligandꢀdetected NMR
assays used in FBDD provide evidence of both compound identity and binding simultaneously,
whilst both NMR and SPR discriminate between covalent and nonꢀcovalent interactions. Therefore
the FBDD data suggest that the binding observed is nonꢀcovalent and can be attributed to the 2ꢀATs
themselves rather than reactive precursors or impurities in the samples.
Analysis of the AstraZeneca HTS data reveals that 2ꢀATs show elevated frequency as screening hits
relative to the compound library as a whole, although the percentage (10%) is significantly lower
than the corresponding value for other known PAINS classes, which can be around 15ꢀ20%.30 The
AstraZeneca data also suggest that acylation of the 2ꢀamino group ameliorates the problem of
frequent hitting, which indicates that not all 2ꢀATs are problematic in the context of HTS.
Similarly, only 3.2% of 2ꢀATꢀcontaining compounds in the WEHI HTS library are defined as
PAINS, which does not provide a strong case for exclusion of all 2ꢀATs from HTS collections.14
Thus, the two HTS analyses are in broad agreement with each other and suggest that the majority of
2ꢀATs do not seem to be promiscuous at HTSꢀrelevant concentrations in the 10 – 25
µM range.
18
ACS Paragon Plus Environment

Page 19 of 29
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
Even at HTSꢀrelevant concentrations, some optimizable hits can be relatively promiscuous and we
have previously identified a 2ꢀaminobenzothiazole derivative16 that hits four out of the six HTS
assays selected for the PAINS analysis,14 yet led to a highly selective and potent compound.34
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Nonetheless, many 2ꢀATs were observed to be problematic in both HTS and FBDD, suggesting that
the promiscuity observed in FBDD is not due solely to the higher concentration used in the
fragment screens. Whilst the higher hit rates that are expected in FBDD dictate that many fragments
are likely to hit more than one target, these hits are only useful if they can be elaborated into more
potent compounds. On this basis, it is possible that certain substructures such as 2ꢀATs, which show
some limited promiscuity in HTS assays, but would not be flagged for exclusion from a fragment
library by analysis of chemical or physical properties, are in fact undesirable as members of a
fragment library based on a retrospective analysis of their behaviour in screening assays. A similar
approach has been described previously in the HTS setting for deprioritising compounds that are
known to be promiscuous.35
CONCLUSION
We have identified 2ꢀATs as a promiscuous substructure in screens of our fragment library. Each of
the 28 fragments containing a 2ꢀAT substructure passed all of the biophysical and chemical filters
that were used in designing the library and were demonstrated to have appropriate purity and
aqueous solubility. However, based on our findings reported here and our unsuccessful attempts to
optimize these fragments against different targets, we have removed 2ꢀATs from the fragment
library at MIPS, on the basis that we judge the prospects for such compounds to be progressable as
much smaller than the chances that they will not.
19
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 29
1
2
EXPERIMENTAL SECTION
3
4
5
6
7
8
9
General procedures. Biacore sensor chips, N-hydroxysuccinimide (NHS), NꢀethylꢀN′ꢀ(3ꢀ
dimethylaminopropyl) carbodiimide (EDC), ethanolamine HCl, HBSꢀP running buffer, and Pꢀ20
surfactant were obtained from GE Healthcare. Carbonic anhydrase II and 4ꢀcarboxybenzyl
sulfonamide were obtained from SigmaꢀAldrich. Isabelle Lucet (WEHI) provided the kinase protein
and positive control H4. Doubly Hisꢀtagged oxidoreductases 1 and 2 were provided from Martin
Williams (MIPS). All 2ꢀAT fragments were obtained from Chembridge or Life. The aqueous
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
solubility, identity and purity (>95%) of the purchased compounds was verified by analysis of Hꢀ
1D NMR spectra as described previously.12
Surface Plasmon Resonance (SPR) conditions
Expression and purification of hexahistidineꢀtagged AMA1,36 and GSTꢀtagged SPSB2,25 were as
described previously. Binding of 2ꢀATs to the target proteins was measured on a Biacore T200
instrument (GE Healthcare). AMA1, SPSB2, and CAII were immobilized onto a CM5 sensor chip
(Biacore) by standard amine coupling chemistry using sodium acetate at pH 4.5, in running buffer
A. DoublyꢀHis6ꢀtagged oxidoreductase proteins were immobilised onto an NTA chip (Biacore)
charged with Ni2+ according to manufacturer’s instructions, in running buffer B. SinglyꢀHis6ꢀtagged
Kinase was immobilized to an NTA sensor chip charged with Ni2+ using the CaptureꢀCouple
method,37 in running buffer C. Immobilisation levels were typically 9000 RU for AMA1, 6000 RU
for SPSB2 and CAII, and 2000 – 3000 RU for both oxidoreductases, and the kinase.
Running buffer A: 25 mM HEPES, 150 mM NaCl, 5% DMSO, 0.005% Pꢀ20, pH 7.4.
Running buffer B: 25 mM HEPES, 200 mM NaCl, 5% DMSO, 0.005% Tweenꢀ20, pH 8.0
Running buffer C: 25 mM HEPES, 150 mM NaCl, 4 mM MgCl2, 2 mM TCEP, 3% DMSO, 0.005%
Tweenꢀ20, pH 7.5.
SPR Screening of 2-ATs
Binding assays were carried out at 25°C using a flowꢀrate of 100 µL/min in running buffer A
(AMA1, SPSB2, CAII), running buffer B (oxidoreductase 1 and 2), or running buffer C (kinase).
20
ACS Paragon Plus Environment

Page 21 of 29
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
200 mM fragment stocks were diluted in the appropriate running buffer to obtain 200 µM working
concentrations. The association and dissociation phases of binding were each followed for 30 s. An
identical injection of running buffer was carried out between each fragment injection, and the
average of adjacent blanks was subtracted from the raw fragment response. Raw sensorgram data
were reduced, solventꢀcorrected, and doubleꢀreferenced using BIAEvaluation Software (GE
Healthcare). Raw responses were converted to percentage occupancy by the method of Giannetti38
using an Rmax value based on the response measured with a saturating concentration of the
appropriate control, and the molecular weight of each fragment. For oxidoreductase 2, there is no
positive control available, and binding is reported as percentage of the theoretical Rmax.
SPR Positive Controls
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The R1 peptide39 was used as a control for AMA1. A peptide corresponding to residues within the
Nꢀterminus of inducible nitric oxide synthase was used as a control for SPSB2.25 4ꢀCarboxybenzyl
sulfonamide was used as a control for CAII. An elaborated fragment “H4” was used as a control for
the kinase. An elaborated fragment “LA010” (our unpublished data) was used as a control for
oxidoreductase 1. As oxidoreductase 2 has demonstrated a weak ability to bind fragments
(unpublished data), no positive control was available for this protein. Doseꢀresponse determinations
were carried out for all positive controls to confirm protein activity. A concentration series of 2ꢀfold
dilutions in running buffer was used, with a top concentration of 1 µM (R1), 5 µM (iNOS peptide),
10 µM (4ꢀCBS), 20 µM (H4) or 200 µM (LA010) (see Supplementary Figure S1).
Screening by STD-NMR
Screening was performed by recording STDꢀNMR experiments on cocktails of up to six fragments
in the presence of the target protein (1ꢀ10 ꢁM). Screens were undertaken at 600 MHz or 800 MHz
on spectrometers equipped with a cryogenically cooled probes. Where possible, screens were
2
carried out in phosphate buffer prepared in > 90% H2O, to minimise spectral interference from
1
either buffer signals or the H2O resonance, and at a pH close to 7.0 and temperature of 10 °C.
Other buffer conditions were used where necessary as dictated by the stability of the target protein.
21
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 29
1
2
3
4
5
6
7
8
The resulting data were processed in Topspin (Bruker Biospin) and analysed manually. STD signal
intensities in the spectra were qualitatively classified as strong, medium or weak based on the
relative intensity of signals observed in the lowꢀfield region (> 5.0 ppm). Relative intensities were
based on the most intense STD signal (Imax) identified across all the STD spectra for a particular
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
target. A positive STD signal was categorised as strong where the intensity was > 50% Imax
,
medium where the intensity was > 25% Imax and < 50% Imax or weak where the intensity was < 25%
1
Imax. If the fragment contained no resonances in the lowꢀfield region of the HꢀNMR spectrum, the
aliphatic region (< 4.5 ppm) was analysed and such fragments were considered hits if any positive
STD signal was observed. No attempt was made to categorise aliphatic STD signals due to the
potential for either direct excitation by the onꢀresonance saturating pulse in the STD experiment
and/or interference from overlapping protein resonances (See Supplementary Figure S6).
Analysis of binding by HSQC-NMR
1
To analyze the location of the fragment binding site on oxidoreductase 1, 2D Hꢀ15N HSQC NMR
spectra were recorded in the absence and presence of fragments 1, 5, 6, and 14. A reference 1Hꢀ15N
HSQC of the protein (125 µM [Uꢀ15N]ꢀlabeled protein; 2% DMSOꢀd6; 50 mM HEPES; 50 mM
NaCl; pH 6.8) was acquired and compared with a spectrum acquired under the same conditions for
a sample containing the oxidized oxidoreductase 1 in the presence of each fragment (125 µM [Uꢀ
15N]ꢀlabeled protein; 1ꢀ1.5 mM fragment; 1ꢀ2% DMSOꢀd6; 50 mM HEPES; 50 mM NaCl; pH 6.8).
Subsequently, the protein was titrated with 14 (0.09ꢀ1.5 mM). 3 mm thick walled tubes of sample
volume ~ 160 µL were used for NMR data collection. All data were acquired on a Bruker 600 MHz
spectrometer equipped with autoꢀsampler and CryoProbe at 300K. Standard acquisition and
processing parameters were used throughout (See Supplementary Figure S7).
19F NMR Studies of 5-Trp-SPSB2
19
Construction of the 5ꢀFꢀTrpꢀSPSB2 protein, and F NMR determination of thiazole binding, were
carried out as described previously.28 19F NMR studies were carried out at 30 °C in 50 mM
22
ACS Paragon Plus Environment

Page 23 of 29
Journal of Medicinal Chemistry
1
2
Phosphate, 50 mM NaCl (pH 7.4), with a 5ꢀFꢀTrpꢀSPSB2 concentration of 100 µM, with or without
3
4
5
500 µM control peptide or 3 mM 1, as indicated.
6
7
8
9
ACKNOWLEDGMENTS
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The authors would like to thank Martin Williams for the oxidoreductases 1 and 2, Luke Adams for
the oxidoreductase 1 positive control, Hiromasa Yagi for assistance on the 19F NMR work and
Isabelle Lucet for the kinase protein sample and positive control. This work was supported in part
by grants from the Australian National Health and Medical Research Council (NHMRC) (1045995,
1009785, 1022693, 1025150). J.B.B. and R.S.N. acknowledge fellowship support from the
NHMRC.
ABBREVIATIONS USED
2ꢀAT, 2ꢀaminothiazole; AMA1, apical membrane antigen 1; CAII, carbonic anhydrase II; CSP,
chemical shift perturbation; FBDD, fragmentꢀbased drug discovery; HSQC, heteronuclear single
quantum correlation; HTS, highꢀthroughput screening; MIPS, Monash Institute of Pharmaceutical
Sciences; NMR, nuclear magnetic resonance; PAINS, Pan Assay INterference compoundS; PrATs,
Promiscuous 2ꢀAminoThiazoles; SPSB2, SPRY domainꢀcontaining SOCS box protein 2; SAR,
structureꢀactivity relationships; STD, saturation transfer difference; WEHI, Walter and Eliza Hall
Institute.
ASSOCIATED CONTENT
Supplementary Material
Supplementary material is available, which contains SPR sensorgrams of positive controls and
selected compounds, STD, HSQCꢀNMR and SPRꢀderived binding data for the different targets, 2ꢀ
AT PAINS subclasses and frequent hitter incidence. This material is free of charge via the Internet
at http://pubs.acs.org.
23
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 29
1
2
REFERENCES
3
4
5
6
7
8
9
(1)
(2)
(3)
Shuker, S. B.; Hajduk, P. J.; Meadows, R. P.; Fesik, S. W. Discovering HighꢀAffinity
Ligands for Proteins: SAR by NMR. Science 1996, 274, 1531–1534.
Erlanson, D. A.; McDowell, R. S.; O'Brien, T. FragmentꢀBased Drug Discovery. J. Med.
Chem. 2004, 47, 3463–3482.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Tsai, J.; Lee, J. T.; Wang, W.; Zhang, J.; Cho, H.; Mamo, S.; Bremer, R.; Gillette, S.;
Kong, J.; Haass, N. K.; Sproesser, K.; Li, L.; Smalley, K. S. M.; Fong, D.; Zhu, Y.ꢀL.;
Marimuthu, A.; Nguyen, H.; Lam, B.; Liu, J.; Cheung, I.; Rice, J.; Suzuki, Y.; Luu, C.;
Settachatgul, C.; Shellooe, R.; Cantwell, J.; Kim, S.ꢀH.; Schlessinger, J.; Zhang, K. Y. J.;
West, B. L.; Ben Powell; Habets, G.; Zhang, C.; Ibrahim, P. N.; Hirth, P.; Artis, D. R.;
Herlyn, M.; Bollag, G. Discovery of a Selective Inhibitor of Oncogenic BꢀRaf Kinase with
Potent Antimelanoma Activity. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 3041–3046.
LaPlante, S. R.; Padyana, A. K.; Abeywardane, A.; Bonneau, P.; Cartier, M.; Coulombe,
R.; Jakalian, A.; WildesonꢀJones, J.; Li, X.; Liang, S.; McKercher, G.; White, P.; Zhang,
Q.; Taylor, S. J. Integrated Strategies for Identifying Leads That Target the NS3 Helicase
of the Hepatitis C Virus. J. Med. Chem. 2014, 57, 2074–2090.
(4)
(5)
Edfeldt, F. N. B.; Folmer, R. H. A.; Breeze, A. L. Fragment Screening to Predict
Druggability (Ligandability) and Lead Discovery Success. Drug Discov. Today 2011, 16,
284–287.
(6)
(7)
(8)
Fink, T.; Bruggesser, H.; Reymond, J.ꢀL. Virtual Exploration of the SmallꢀMolecule
Chemical Universe Below 160 Daltons. Angew. Chem. Int. Ed. Engl. 2005, 44, 1504–1508.
Congreve, M.; Chessari, G.; Tisi, D.; Woodhead, A. J. Recent Developments in Fragmentꢀ
Based Drug Discovery. J. Med. Chem. 2008, 51, 3661–3680.
Fink, T.; Reymond, J.ꢀL. Virtual Exploration of the Chemical Universe Up to 11 Atoms of
C, N, O, F: Assembly of 26.4 Million Structures (110.9 Million Stereoisomers) and
Analysis for New Ring Systems, Stereochemistry, Physicochemical Properties, Compound
24
ACS Paragon Plus Environment

Page 25 of 29
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
Classes, and Drug Discovery. J. Chem. Inf. Model. 2007, 47, 342–353.
Murray, C. W.; Carr, M. G.; Callaghan, O.; Chessari, G.; Congreve, M.; Cowan, S.; Coyle,
J. E.; Downham, R.; Figueroa, E.; Frederickson, M.; Graham, B.; McMenamin, R.;
O'Brien, M. A.; Patel, S.; Phillips, T. R.; Williams, G.; Woodhead, A. J.; Woolford, A. J.ꢀ
A. FragmentꢀBased Drug Discovery Applied to Hsp90. Discovery of Two Lead Series with
High Ligand Efficiency. J. Med. Chem. 2010, 53, 5942–5955.
(9)
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(10)
Bollag, G.; Hirth, P.; Tsai, J.; Zhang, J.; Ibrahim, P. N.; Cho, H.; Spevak, W.; Zhang, C.;
Zhang, Y.; Habets, G.; Burton, E. A.; Wong, B.; Tsang, G.; West, B. L.; Powell, B.;
Shellooe, R.; Marimuthu, A.; Nguyen, H.; Zhang, K. Y. J.; Artis, D. R.; Schlessinger, J.;
Su, F.; Higgins, B.; Iyer, R.; D'Andrea, K.; Koehler, A.; Stumm, M.; Lin, P. S.; Lee, R. J.;
Grippo, J.; Puzanov, I.; Kim, K. B.; Ribas, A.; McArthur, G. A.; Sosman, J. A.; Chapman,
P. B.; Flaherty, K. T.; Xu, X.; Nathanson, K. L.; Nolop, K. Clinical Efficacy of a RAF
Inhibitor Needs Broad Target Blockade in BRAFꢀMutant Melanoma. Nature 2010, 467,
596–599.
(11)
Bollag, G.; Tsai, J.; Zhang, J.; Zhang, C.; Ibrahim, P.; Nolop, K.; Hirth, P. Vemurafenib:
the First Drug Approved for BRAFꢀMutant Cancer. Nat. Rev. Drug Discov. 2012, 11, 873–
886.
(12)
(13)
Doak, B. C.; Morton, C. J.; Simpson, J. S.; Scanlon, M. J. Design and Evaluation of the
Performance of an NMR Screening Fragment Library. Aust. J. Chem. 2014, 66, 1465–1472.
Baurin, N.; AboulꢀEla, F.; Barril, X.; Ben Davis; Drysdale, M.; Dymock, B.; Finch, H.;
Fromont, C.; Richardson, C.; Simmonite, H.; Hubbard, R. E. Design and Characterization
of Libraries of Molecular Fragments for Use in NMR Screening Against Protein Targets. J.
Chem. Inf. Comput. Sci. 2004, 44, 2157–2166.
(14)
Baell, J. B.; Holloway, G. A. New Substructure Filters for Removal of Pan Assay
Interference Compounds (PAINS) From Screening Libraries and for Their Exclusion in
Bioassays. J. Med. Chem. 2010, 53, 2719–2740.
25
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 29
1
2
3
4
5
6
7
8
(15)
(16)
Hansch, C.; Fujita, T. PꢀΣꢀΠ Analysis. a Method for the Correlation of Biological Activity
and Chemical Structure. J. Am. Chem. Soc. 1964, 86, 1616–1626.
Baell, J. B.; Ferrins, L.; Falk, H.; Nikolakopoulos, G. PAINS: Relevance to Tool
Compound Discovery and FragmentꢀBased Screening. Aust. J. Chem. 2013, 66, 1483–
1494.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(17)
Camps, M.; Rückle, T.; Ji, H.; Ardissone, V.; Rintelen, F.; Shaw, J.; Ferrandi, C.; Chabert,
C.; Gillieron, C.; Françon, B.; Martin, T.; Gretener, D.; Perrin, D.; Leroy, D.; Vitte, P.ꢀA.;
Hirsch, E.; Wymann, M. P.; Cirillo, R.; Schwarz, M. K.; Rommel, C. Blockade of
PI3Kgamma Suppresses Joint Inflammation and Damage in Mouse Models of Rheumatoid
Arthritis. Nat. Med. 2005, 11, 936–943.
(18)
(19)
Huth, J. R.; Mendoza, R.; Olejniczak, E. T.; Johnson, R. W.; Cothron, D. A.; Liu, Y.;
Lerner, C. G.; Chen, J.; Hajduk, P. J. ALARM NMR: a Rapid and Robust Experimental
Method to Detect Reactive False Positives in Biochemical Screens. J. Am. Chem. Soc.
2005, 127, 217–224.
Huth, J. R.; Song, D.; Mendoza, R. R.; BlackꢀSchaefer, C. L.; Mack, J. C.; Dorwin, S. A.;
Ladror, U. S.; Severin, J. M.; Walter, K. A.; Bartley, D. M.; Hajduk, P. J. Toxicological
Evaluation of ThiolꢀReactive Compounds Identified Using a La Assay to Detect Reactive
Molecules by Nuclear Magnetic Resonance. Chem. Res. Toxicol. 2007, 20, 1752–1759.
Metz, J. T.; Huth, J. R.; Hajduk, P. J. Enhancement of Chemical Rules for Predicting
Compound Reactivity Towards Protein Thiol Groups. J. Comput. Aided Mol. Des. 2007,
21, 139–144.
(20)
(21)
Mjambili, F.; Njoroge, M.; Naran, K.; De Kock, C.; Smith, P. J.; Mizrahi, V.; Warner, D.;
Chibale, K. Synthesis and Biological Evaluation of 2ꢀAminothiazole Derivatives as
Antimycobacterial and Antiplasmodial Agents. Bioorg. Med. Chem. Lett. 2014, 24, 560–
564.
(22)
Lim, S. S.; Debono, C. O.; MacRaild, C. A.; Chandrashekaran, I. R.; Dolezal, O.; Anders,
26
ACS Paragon Plus Environment

Page 27 of 29
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
R. F.; Simpson, J. S.; Scanlon, M. J.; Devine, S. M.; Scammells, P. J.; Norton, R. S.
Development of Inhibitors of Plasmodium Falciparum Apical Membrane Antigen 1 Based
on Fragment Screening. Aust. J. Chem. 2013, 66, 1530–1536.
8
9
(23)
Chuprina, A.; Lukin, O.; Demoiseaux, R.; Buzko, A.; Shivanyuk, A. Drugꢀ and Leadꢀ
Likeness, Target Class, and Molecular Diversity Analysis of 7.9 Million Commercially
Available Organic Compounds Provided by 29 Suppliers. J. Chem. Inf. Model. 2010, 50,
470–479.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(24)
(25)
Mayer, M.; Meyer, B. Characterization of Ligand Binding by Saturation Transfer
Difference NMR Spectroscopy. Angew. Chem. Int. Ed. Engl. 1999, 38, 1784–1788.
Kuang, Z.; Lewis, R. S.; Curtis, J. M.; Zhan, Y.; Saunders, B. M.; Babon, J. J.; Kolesnik, T.
B.; Low, A.; Masters, S. L.; Willson, T. A.; Kedzierski, L.; Yao, S.; Handman, E.; Norton,
R. S.; Nicholson, S. E. The SPRY DomainꢀContaining SOCS Box Protein SPSB2 Targets
iNOS for Proteasomal Degradation. J. Cell Biol. 2010, 190, 129–141.
Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand Efficiency: a Useful Metric for Lead
Selection. Drug Discov. Today 2004, 9, 430–431.
(26)
(27)
(28)
Supuran, C. T.; Scozzafava, A. Carbonic Anhydrase Inhibitors and Their Therapeutic
Potential. Exp. Opin. Ther. Patents 2000, 10, 575–600.
Leung, E. W. W.; Yagi, H.; Harjani, J. R.; Mulcair, M. D.; Scanlon, M. J.; Baell, J. B.;
Norton, R. S. (19) F NMR as a Probe of Ligand Interactions with the iNOS Binding Site of
SPRY DomainꢀContaining SOCS Box Protein 2. Chem. Biol. Drug Des. 2014, 84, 616–
625.
(29)
(30)
(31)
Baell, J. B. Broad Coverage of Commercially Available LeadꢀLike Screening Space with
Fewer Than 350,000 Compounds. J. Chem. Inf. Model. 2013, 53, 39–55.
M Nissink, J. W.; Blackburn, S. Quantification of FrequentꢀHitter Behavior Based on
Historical HighꢀThroughput Screening Data. Future Med. Chem. 2014, 6, 1113–1126.
Bruneau, P. Search for Predictive Generic Model of Aqueous Solubility Using Bayesian
27
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 29
1
2
Neural Nets. J. Chem. Inf. Model. 2001, 41, 1605–1616.
3
4
5
6
7
8
9
(32)
Bruneau, P.; McElroy, N. R. logD7.4 Modeling Using Bayesian Regularized Neural
Networks. Assessment and Correction of the Errors of Prediction. J. Chem. Inf. Model.
2006, 46, 1379–1387.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(33)
(34)
Tarcsay, Á.; Keserü, G. M. Contributions of Molecular Properties to Drug Promiscuity. J.
Med. Chem. 2013, 56, 1789–1795.
Lessene, G.; Czabotar, P. E.; Sleebs, B. E.; Zobel, K.; Lowes, K. N.; Adams, J. M.; Baell, J.
B.; Colman, P. M.; Deshayes, K.; Fairbrother, W. J.; Flygare, J. A.; Gibbons, P.; Kersten,
W. J. A.; Kulasegaram, S.; Moss, R. M.; Parisot, J. P.; Smith, B. J.; Street, I. P.; Yang, H.;
Huang, D. C. S.; Watson, K. G. StructureꢀGuided Design of a Selective BCLꢀX(L)
Inhibitor. Nat. Chem. Biol. 2013, 9, 390–397.
(35)
Nissink, J. W. M.; Schmitt, S.; Blackburn, S.; Peters, S. Stratified HighꢀThroughput
Screening Sets Enable Flexible Screening Strategies From a Single Plated Collection. J.
Biomol. Screen. 2014, 19, 369–378.
(36)
(37)
(38)
Hodder, A. N.; Crewther, P. E.; Anders, R. F. Specificity of the Protective Antibody
Response to Apical Membrane Antigen 1. Infect. Immun. 2001, 69, 3286–3294.
Rich, R. L.; Errey, J.; Marshall, F.; Myszka, D. G. Biacore Analysis with Stabilized Gꢀ
ProteinꢀCoupled Receptors. Anal. Biochem. 2011, 409, 267–272.
Giannetti, A. M. From Experimental Design to Validated Hits a Comprehensive Walkꢀ
Through of Fragment Lead Identification Using Surface Plasmon Resonance. Meth.
Enzymol. 2011, 493, 169–218.
(39)
Harris, K. S.; Casey, J. L.; Coley, A. M.; Masciantonio, R.; Sabo, J. K.; Keizer, D. W.; Lee,
E. F.; McMahon, A.; Norton, R. S.; Anders, R. F.; Foley, M. Binding Hot Spot for Invasion
Inhibitory Molecules on Plasmodium Falciparum Apical Membrane Antigen 1. Infect.
Immun. 2005, 73, 6981–6989.
28
ACS Paragon Plus Environment

Page 29 of 29
Journal of Medicinal Chemistry
1
2
Table of Contents Graphic
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
29
ACS Paragon Plus Environment