Drug Discovery Today: Technologies
Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021)
TECHNOLOGIES
Division of Medicinal Chemistry, Amsterdam Institute of Molecular and Life Sciences (AIMMS), Faculty of Science, Vrije Universiteit
Amsterdam, De Boelelaan 1108, 1081 HZ Amsterdam, The Netherlands 2
Department of Chemistry, University of York, Heslington, York, YO10 5DD, UK
Fragment-based drug discovery (FBDD) has grown into
a well-established approach in the pursuit of new therapeutics. Key to the success of FBDD is the low molecular complexity of the initial hits and this has resulted in
fragment libraries that mainly contain compounds with
a two-dimensional (2D) shape. In an effort to increase
the chemical diversity and explore the impact of increased molecular complexity on the hit rate of fragment library screening, several academic and industrial
groups have designed and synthesised novel fragments
with a three-dimensional (3D) shape. This review provides an overview of 25 synthetic 3D fragment libraries
from the recent literature. We calculate and compare
physicochemical properties and descriptors that are
typically used to measure molecular three-dimensionality such as fraction sp3 (Fsp3
), plane of best fit (PBF)
scores and principal moment of inertia (PMI) plots.
Although the libraries vary widely in structure and
properties, some key common features can be identi-
fied which may have utility in designing the next
generation of 3D fragment libraries.
Section editor: IJP de Esch – Faculty of Science, Chemistry
and Pharmaceutical Sciences, AIMMS.
Introduction
Fragment-based drug discovery (FBDD) is now a well-established approach to drug discovery [1–3] having produced four
FDA-approved drugs as of 2020 and several clinical candidates [4–10]. Fragment libraries are generally better able to
sample their respective chemical space compared to highthroughput libraries and typical diverse fragment libraries
Drug Discovery Today: Technologies Vol. xxx, No. xx 2019
Editors-in-Chief
Kelvin Lam – Simplex Pharma Advisors, Inc., Boston, MA, USA
Henk Timmerman – Vrije Universiteit, The Netherlands
*Corresponding author: Iwan J.P. de Esch ([email protected])
1740-6749/$ © 2021 Published by Elsevier Ltd. https://doi.org/10.1016/j.ddtec.2021.05.001 1
consist of merely 1500–2000 entities [11]. By design, fragments have a low molecular complexity [12] and, considering
the high percentage of (hetero)aromatic rings in the fragment
libraries, the shape of many screening fragments has been
referred to as 2D [13,14]. Introducing more fragments with
3D character, for example with sp3
-hybridised carbon atoms
embedded in aliphatic ring systems, potentially incorporates
more chiral centres and growth vectors for fragment elaboration. It can be argued that the associated increase in molecular complexity could result in significantly lower hit rates in
fragment screening campaigns [12]. Nevertheless, as demonstrated by the recent synthetic interest in this area, there
remains an interest in more thoroughly investigating the use
of 3D fragments, not only to better explore the impact of the
3D character on the fragment screening hit rates, but also to
allow the identification of novel structures as starting points
for hit optimisation studies [15]. Next to novelty, 3D fragments could bring other advantages to the early hit identifi-
cation stage [16]. Three-dimensionality in structures is
believed to increase water solubility [17]. The higher molecular complexity might actually be an advantage when targeting more difficult targets such as in the (de)stabilisation of
protein-protein interactions (PPIs) [18,19] or obtaining selectivity over closely-related proteins [20]. As was described by
Lovering et al. in their seminal ‘‘Escape from Flatland’’ papers
[14], the proportion of sp3
-hybridised carbon atoms in compounds increases during hit and lead optimisation, suggesting that increased three-dimensionality may offer a reduced
rate of candidate attrition. Although the cited studies on
advantages ofthree-dimensionality do not appertain to FBDD
approaches specifically, it is not unreasonable to propose that
3D fragment hits may well be valuable new starting points for
drug development programmes [21].
In general, a bias toward 2D structures in medicinal chemistry as a whole has been rationalised by an analysis conducted by Walters et al., which showed that over the five
decades prior to 2009 there had indeed been a surge in the
number of compounds containing sp2
–sp2 bonds [22]. The
authors attribute this to developments in synthetic methodology which favour formation of such connections. One
potential challenge associated with 3D fragments is the perceived difficulty of synthesis, which is only exaggerated when
compared to the more commonly used methodology for sp2
–
sp2 couplings such as the robust Suzuki reaction. Several
publications have arisen over the last decade – the last few
years in particular – which address the lack of synthetic
methodology for the generation of 3D fragments and scaffolds [23–27]. Moreover, Rees and co-workers have emphasised the requirement for incorporation of efficient growth
vectors with which to evolve a 3D fragment into a lead
compound, coined ‘fragment sociability’ [25].
The aim of this review is to analyse selected physicochemical properties and the 3D character of recently published 3D
fragment collections that have been assembled by tailored
synthesis. To this end, key physicochemical parameters as
well as the three most commonly used 3D descriptors will be
used in a consistent and comprehensive manner, allowing for
fair comparison between the discussed fragment collections.
The sophisticated strategies for the synthesis of the 3D fragment collections presented herein are highlighted in our
companion article in this issue.
Common 3D descriptors
There are a number of descriptors through which 3D shape
may be assessed. Commonly, the simple metric of fraction sp3
(Fsp3
) as defined by Yan and Gasteiger [17] has been used to
quantify the 3D character of molecules. It is defined as the
number of sp3
-hybridised carbon atoms divided by the total
number of carbon atoms in a given molecule (Fig. 1A). Whilst
a definitive cut-off has not been established for this metric,
Kombo et al. suggest an Fsp3 value 0.42 to be suitable as a
three-dimensionality criterion [28], whilst others have stated
a value of Fsp3 0.45 as appropriate [29]. This descriptor has
the advantage of giving chemists a very rapid, single-number
value with which to compare molecules. However, it does not
take into consideration other atom types such as nitrogen or
oxygen which may constitute a considerable portion of the
molecule, nor whether these non-carbon heavy atoms contribute to three-dimensionality, possibly leading to a misleading value.
Two alternative prevalent measures, which are more
information-rich and have often been used in the publications discussed in this review, are Plane of Best Fit (PBF) [30]
and Principal Moment of Inertia (PMI) [11] (Fig. 1B and C).
PBF analysis is a descriptor proposed by Firth et al. [30], in
which a plane is fitted through a computationally obtained
molecular conformation by using a least-squares method to
minimise the average distance from each heavy atom to the
plane. This best-fitting plane includes the central point of
mass of the molecule and is termed the PBF, and the
average distance (in A˚ ) is referred to as the PBF score.
Typically, more 2D planar compounds lie along this plane
of best fit (for example benzene in Fig. 1B), whilst threedimensional compounds have atoms that deviate more
from the plane (for example cyclohexane in Fig. 1B) –
therefore, 3D molecules have a greater PBF score than 2D
molecules. Firth et al. proposed a cut-off value of PBF 0.60
for a molecule to be deemed 3D. This method gives a single
value, allowing for easy comparison. It should be noted that
the PBF method varies with molecular size and hence only
molecules which are similar in size may be compared
without bias.
PMI analysis was first proposed by Saur and Schwarz [11]
and serves as a measure of molecular shape in terms of rod-,
disc- and sphere-like character. The moment of inertia of a
rigid body is a quantitative measure of the amount of torque
Drug Discovery Today: Technologies | Vol. xxx, No. xx 2019
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
2 www.drugdiscoverytoday.com
that is required to rotate that body around an axis. The PMI
method computationally calculates the moments of inertia of
a particular conformation of a molecule around its three
principal axes, which are then sorted by ascending magnitude (I1 > I2 > I3) and subsequently converted into the
normalised PMI ratios, NPR1 (I1/I3) and NPR2 (I2/I3). Plotting
these two values against each other in a two-dimensional
triangular graph allows for quick visualisation of the molecular shape (Fig. 1C). The top left vertex of the plot (0.0,1.0)
denotes linear, rod-shaped molecules such as acetylene. The
bottom vertex (0.5,0.5) denotes planar, disc-shaped molecules such as benzene. The top right corner (1.0,1.0) denotes
spherical molecules such as adamantane. For a numerical
comparison of three-dimensionality, the two NPRs can be
summed (
PNPR), which yields a value between 1.0 and 2.0. A
higher PNPR value corresponds to a greater deviation from
the rod-disc axis, and therefore a higher three-dimensional
character. It was proposed by Firth et al. [30] that a molecule
may be deemed to be 3D if it lies off the rod-disc axis, with a
PNPR 1.07.
On occasion, the sets of descriptors have been compared in
a systematic fashion. Firth et al. have shown there is no
correlation between Fsp3 and PBF for a wide range of medicinally-relevant compounds [30], whilst the O’Brien group has
shown there is no correlation between Fsp3 and PMI for
various commercially available fragment libraries [31]. Whilst
this may suggest limitations of the Fsp3 descriptor as a measure of three-dimensionality, it remains widely in use. PMI
and PBF tend to correlate considerably, with most libraries
which show such analysis as part of their work demonstrating
this trend.
Approach
A total of 25 recent publications were selected. In addition to
some seminal papers predating 2015, most articles are from
the period 2015–mid-May 2020, all of which fulfil the criteria
of describing organic molecules as fragments (excluding organometallic fragments, e.g. those presented by the Cohen
group [32], and synthesis of intermediates for fragment generation, e.g. substituted piperazines by the Young group [33–
35]) which had been targeted by tailored synthesis. The
publications include the terms ‘sp3
-rich fragment,’ ‘shapediverse fragment’ or ‘3D fragment,’ or a variation thereof. We
found that a few groups prominent in the FBDD field do not
necessarily emphasise these keywords and from these groups
we included publications deemed relevant for this review.
The libraries appeared qualitatively rule of three (RO3)-compliant [36] which is important to ensure the collection
remains (largely) within fragment space. The selected articles,
associated scaffolds and key features are summarised in
Table 1 (scaffolds highlighted in blue are key examples as
there are too many disparate scaffolds in those papers and it is
not possible to be comprehensive) [31,37–60]. For the purpose of comparison, scaffold interpretation and counting
have been conducted in the context of this review. The
following criteria were used to assess scaffolds in the discussed
libraries: (1) unique ring systems with respect to atom type(s)
and ring size (e.g. spirocycles and bicyclics) are classified as
distinct; (2) alteration of endocyclic functional groups (i.e.
constituting the ring system) leads to a distinct scaffold (with
the exception of alkene saturation, to avoid overinflation of
scaffold count); (3) regioisomers are considered distinct; (4)
stereoisomers of the same ring unit are not considered to be
distinct; (5) if replacing substituents with an R-notation
results in the same outcome, the scaffolds are deemed to
be identical unless a new ring system is generated. As can be
seen, the synthetic chemistry efforts of several groups led to a
wide variety of different scaffolds – particularly those for
which diversity-oriented synthesis (DOS)-based approaches
were implemented, such as in the groups of Young, Nelson/
Marsden, Spring and Clausen [37,41,45,51,54–57,60]. In such
cases, diversity is built into the molecule at an early stage in
Vol. xxx, No. xx 2019 Drug Discovery Today: Technologies |
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
(a) (b) (c)
Drug Discovery Today: Technologies
Fig. 1. (A) Fraction sp3 (Fsp3
) equation; (B) Example of Plane of Best Fit (PBF) generation for a cyclohexane molecule (upper) in a chair conformation and a
benzene ring (lower). The plane is shown in semi-transparent light blue (modified from the graphical depiction from Firth et al. [30]); (C) Principal Moment
of Inertia (PMI) plot showing three examples of simple molecules, acetylene, benzene and adamantane, defining the extremities of molecular shape at the
three vertices.
www.drugdiscoverytoday.com 3
Drug Discovery Today: Technologies | Vol. xxx, No. xx 2019
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
Table 1. Representative examples or generalised scaffold structures of 3D fragment libraries analysed in this review.
Entry Research
>
4 www.drugdiscoverytoday.com
Vol. xxx, No. xx 2019 Drug Discovery Today: Technologies |
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
For clarity, the research groups are referred to by the corresponding author(s)/group leader. If multiple publications arise from the same group, they are numbered according to their
publication date and this notation (in brackets) is used in other figures. b General description of the scaffold(s). c
Structures of the disclosed scaffold(s). Where possible, allscaffolds were generalised into one structure (black). In cases where allscaffolds cannot be summarised, the shown structures are
exemplary and coloured blue. IUPAC rules for depicting relative and absolute stereochemistry were followed. d Number of fragments disclosed – for cases in which mixtures of diastereomers are present, both compounds are included. e Number of distinct scaffolds comprising the fragment library, according to the rules described in the main text. f
3D metrics used in the paper. g
Two compounds are omitted from our data set for Willand 2 (entry 7) as conformations could not be generated for compounds 2g and 2h from the publication. h Purchased compounds for Nelson 3 (entry 16) are not included in this review. i
Only compounds which passed quality controls according to the authors are included for Clausen (entry 19).
www.drugdiscoverytoday.com 5
order to maximise structural diversity, rather than diversifying a common intermediate.
It is generally not easy to compare PMI or PBF outputs from
different research groups, as these descriptors require computational generation of molecular conformations, which may
differ amongst different software packages. Thus, to compare
the properties offered by each fragment collection in a consistent fashion, the libraries of fragments described in each
publication were extracted and processed as follows. Structures of all final compounds successfully synthesised were
manually drawn in ChemDraw 19.1 and saved as an SD file or,
if available, obtained from SMILES tables in the supporting
information and saved as CSV files. Common non-decorative
protecting groups (Boc, CBz, Ts) were removed from all
compounds, while other protecting groups that may have
been incorporated for fragment decoration purposes (Ac, Bn,
PMP, PMB) were only removed when specifically mentioned
as protecting groups by the authors. This resulted in the
removal of all PMP and PMB groups with the occasional
preservation of Ac and Bn groups. Virtual deprotection allows
for a fairer representation of properties, as the deprotected
structure contains all the heavy atoms that would be included
in a final fragment, whereas a protecting group would distort
the overall molecular properties, including shape. Any products isolated as a mixture of diastereomers were included as
separate diastereomers so as to not omit any synthesised
molecules which may offer a different coverage of 3D chemical space to their counterpart stereoisomer. Fragment screening libraries routinely comprise racemates of chiral
compounds, and enantiomers are identical in terms of 3D
descriptors and physicochemical properties – for these only
one enantiomer needs to be considered. All data was
imported into Molecular Operating Environment (MOE) version 2019.0102. SMILES were generated and each structure
was annotated with a unique identifier and the name of the
article from which the compound originated. The database
was exported as a CSV file and imported into PipeLine Pilot
8.5.0.200 (2011, Accelrys Software Inc.) for conformational
analysis using the BEST method in Catalyst using the Rel
option (maximum 255 conformers per compound), with a
preceding wash step at pH = 7.4. For each conformation, PMI
values were calculated within PipeLine Pilot. The output was
saved as an SD file and imported into KNIME version 4.2.1
and PBF values were calculated using the Erlwood node.
Other descriptors were calculated in KNIME from the neutral
species using RDKIT (MW, HAC, HBD, HBA, TPSA, cLogP
(calculated as SLogP), Fsp3
) or Pipeline Pilot (nRot).
Results & discussion
Physicochemical properties
For each library collection, key physicochemical properties
were calculated using the open-source software KNIME (exact
values are presented in the Supplementary Data). The data are
expressed in the form of radar plots, in which the mean and
range of each collection is given (Fig. 2). The original rule of
three (RO3) [36] cut-offs for fragments are shown in the first
entry in Fig. 2 as a visual aid and state the following:
molecular weight (MW) 300 g mol1
; number of hydrogen
bond acceptors (HBA) 3; number of hydrogen bond donors
(HBD) 3; cLogP 3; as well as two less strict limits of
number of rotatable bonds (nRot) 3 and total polar
surface area (TPSA) 60 A˚ 2
. Whilst not formally part of
the RO3, we also incorporated the heavy atom count
(HAC) as an additional useful parameter in the analyses, as
this is also routinely used for guiding fragment design. This
parameter has the advantage of removing the increased
weighting that heavier atoms contribute towards the MW
cut-off compared to the overall size of the fragment (e.g.
chlorine atoms).
From the analysis in Fig. 2, as had already been estimated
qualitatively, it becomes evident that the majority of the 25
libraries are RO3-compliant. Several libraries exceed the RO3
on the basis of average rotatable bond count and TPSA –
properties which are accepted to be more loosely defined [61]
cf. the more strictly adhered-to MW and HAC. It appears that
in the development of the chemistries and libraries it is not
always easy to adhere to the specific rules of rotatable bond
count and TPSA, although from our analysis it is not clear if
this is an inherent challenge for 3D fragments. Libraries such
as that of the Nelson [41] and Spring [51] groups (Table 1,
entries 4 and 14) are shown to have a wide range of calculated
molecular properties, with ranges spanning both considerably above and below the mean values. This suggests that
such collections possess a considerable amount of chemical
diversity. Interestingly, the bi-/spirocyclic heterocycle library
[51] from Spring and co-workers (Table 1, entry 14) focused
on a DOS approach in order to maximise scaffold diversity –
this seems to translate into a diverse set of properties, as
shown by the corresponding radar plot. In contrast to this,
the data such as that invoked from the Bull group’s oxetane
libraries [39,40] (Table 1, entry 3) and the Willand group’s
spirohydantoin library [44] (Table 1, entry 7) have a much
more narrow spread of calculated properties, with most
compounds falling around the mean value in all shown
parameters. These papers focus on fewer main scaffolds,
therefore decreasing the potential for diversity in the
properties of the library, as reflected in the small ranges of
their corresponding radar plots. The attention of such
publications, importantly, is concerned with developing novel synthetic chemistry with which to generate 3D fragments.
Thus, we speculate that the narrow distribution may be
attributed to the focus of many of these publications lying
in the development of novel synthetic chemistry to
efficiently generate fragment scaffolds with attractive features, such as accessible growth vectors and controlled
stereochemistry.
Drug Discovery Today: Technologies | Vol. xxx, No. xx 2019
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
6 www.drugdiscoverytoday.com
Vol. xxx, No. xx 2019 Drug Discovery Today: Technologies |
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
Fig. 2. Radar plots showing calculated physicochemical properties and molecular features of each library collection. The data was generated using KNIME
or PipeLine Pilot with the neutral chemical species. Axes were scaled according to minimum and maximum values across the full set of 897 compounds
(cLogP: [-1.9;4.5], HAC: [7;27], HBA: [0;8], HBD: [0;4], MW: [95;455 Da], nRot: [0;10], TPSA:[10;140 Å2
]). cLogP was calculated as SLogP. Mean average is
depicted by the blue line. Ranges, defined by the minimum and maximum values, are defined by the grey areas. The rule of three (complemented by HAC =
20) is exemplified as the first entry to provide context for scale interpretation. Exact values can be found in the Supplementary Data.
www.drugdiscoverytoday.com 7
Many of the fragment sets have low average cLogP values,
ranging from the lowest value of –0.12 for Willand’s spirohydantoin library [44] (Table 1, entry 7) up to the highest
value of 3.08 for Bull’s imidazolines [38] (Table 1, entry 2),
with 20 of the 25 libraries (80%) possessing average cLogP <
2.00, and 12 of the 25 libraries (48%) having average cLogP <
1.00. This demonstrates the low predicted lipophilicity of the
discussed 3D fragment collections, and suggests a favourable
factor in minimising the risk of poor aqueous solubility [62].
3D analyses of the novel libraries
Of the 25 libraries, ten calculate the average Fsp3 value for
their library collection, whilst 14 and two use PMI and PBF
analysis, respectively, with some papers using multiple parameters (Table 1). To compare the distribution of Fsp3 values
and PBF scores offered by each fragment collection (exact
values are presented in the Supplementary Data), we generated box plots for each library (Fig. 3A and B). From these
data, it can be seen that the fragment libraries tend to have a
library set which is more populated around the mean values
of these two descriptors, particularly for the PBF descriptor,
with quite small interquartile ranges (i.e. the range populated
by 50% of the values, represented by the box) and only a
limited number of examples of wide ranges (represented by
long whiskers). This implies that groups tend to supply only a
handful of the ‘extremities’ of their library sets, likely for
exemplification, which are considerable outliers compared to
the bulk of their designed library.
In terms of Fsp3
, 19 of the 25 libraries have a mean Fsp3
score 0.45, and even 20 of the 25 possess a mean Fsp3 score
0.42 suggesting that, to some extent, three-dimensionality
can be associated with a higher proportion of sp3
-hybridised
carbon atoms despite this metric lacking any other information. Notably, ten of the 25 libraries give the Fsp3 metric,
indicating that this descriptor remains popular amongst the
fragment community.
In terms of the PBF scores of the data sets, there appears to
be a very narrow range of values despite the theoretical upper
limit being infinite, with a range of PBF averages from 0.28 to
only 0.83. This is likely an effect of the small variation in
molecular size amongst the libraries discussed. Similarly, only
14 of the total 25 libraries possess a mean PBF score of 0.60 or
above, meaning that almost half of the collections do not
meet the proposed PBF cut-off criterion [30]. It is apparent
from Fig. 3A and B that there is little to no correlation
between Fsp3 and PBF for the fragments examined, and this
indicates that Fsp3 is a poor measure of 3D shape. Despite
offering a convenient single digit metric which takes molecular conformations into consideration, in contrast to Fsp3
,
PBF analysis was only reported in two of the 25 publications,
and so is applied quite uncommonly in the recent literature.
Next, we generated PMI plots for each set of fragments
described in a single publication. For each set of fragments,
per library, two PMI plots were generated: one using the
ground state conformation of each molecule (Fig. 3C), and
another one using higher energy conformations accessible up
to a value of 1.5 kcal/mol (data not shown), as proposed by
O’Brien and co-workers [31] as a way of mapping conformationally accessible 3D space. In the majority of cases, there is
negligible difference between the distribution of ground state
and higher energy molecular conformations – with most
higher energy points simply clustering around the original
ground state point. As one would expect, this observation
may be attributed to the rigidity of the fragment members,
with most being either (bi)cyclic- or spiro-compounds, which
have limited capability of bond rotation and therefore a lower
degree of rotational freedom. One such example of this is in
the work of Spring and co-workers [46] (Table 1, entry 9)
which involves fused bicyclic ring systems – a very rigid
structure – translating into a rather clustered PMI plot. Conversely, library sets which contain more rotatable bonds
without such restrictions often differ more in the distribution
of their ground state and higher energy conformations. Only
the ground state conformations are included in this review
for ease of comparison. Notably, this descriptor is the most
information-rich method for assessment of 3D character and
proves to be the most popular, with 14 of the 25 publications
incorporating it to some extent in their own library analyses.
As the definition of what constitutes a 3D fragment is
somewhat loose, the interpretation of PMI data remains fairly
subjective. Broadly speaking, any molecule that is off the roddisc axis in a PMI plot may be considered to be 3D. More
specifically, molecules with a PNPR 1.07 in PMI plots,
therefore off the rod-disc axis, may be deemed to be 3D (vide
supra). Interestingly, 23 of the 25 libraries possess a mean
PNPR 1.07. Indeed, out of the total 897 fragments included in this study, the vast majority (88%) of compounds
possess a PNPR 1.07, with varying distributions across
each PMI plot (exact values are presented in the Supplementary Data). Many collections are successful in occupying the
more 3D region towards the top right corner of the plots. It is
notable that in some cases, exemplified by Bull’s oxetane [39]
and Pomerantz’ thiazepan(on)e [59] libraries (Table 1, entries
3 and 23), the compounds tend to cluster around the same
area of the plot. Often, library members share a common core
which largely dictates the overall molecular shape of the
fragment. The clustered collections tend to have a greater
proportion of structural elements in common, leading to a
higher degree of shape similarity because the difference between each derivative is smaller. As such, collections which
offer scaffold diversity as well as peripheral diversity may offer
much more variation in molecular shape compared to those
with only one larger common core. As a consequence, fragment libraries which incorporate more scaffold-based diversity into their design (notably DOS approaches) boast a
broader coverage of the plot and greater shape diversity, such
Drug Discovery Today: Technologies | Vol. xxx, No. xx 2019
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
8 www.drugdiscoverytoday.com
Vol. xxx, No. xx 2019 Drug Discovery Today: Technologies |
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
Fig. 3. (A) Box plot showing the distribution of Fsp3 values of each library collection, calculated with KNIME using RDKIT nodes. The box represents the
middle 50% of the samples and median value, with whiskers ranging to the maximum and minimum values; (B) Box plot showing the distribution of PBF
scores for the ground state conformations of each library collection, calculated with KNIME using Erlwood nodes; (C) PMI plots for the ground state
conformations of each library collection, calculated using PipeLine Pilot.
www.drugdiscoverytoday.com 9
as those of the Young and Spring groups [45,51], as well as the
Clausen group’s fluorinated library [56] (Table 1, entries 8, 14,
and 19). It is important to emphasise that for libraries with
lower scaffold diversity it is still possible to achieve shape
diversity, as shown by Cox’s bridged pyrrolidines [58] and
Bull’s cyclopropanes [49] (Table 1, entries 21 and 12), which
boast broad coverage of the PMI plot (Fig. 3) despite comprising but one or two scaffolds, respectively. This can be
achieved through exploration of regiochemical and stereochemical diversity, and in such cases PMI analysis can serve as
a useful tool in designing a diverse fragment set prior to
synthesis.
Next, we generated a scatter plot showing SNPR values and
the corresponding PBF scores for all 3D fragments collected
(Fig. 4). It should be noted that whilst PMI analysis is sizeindependent, PBF analysis varies with molecular size and is
best used to compare compounds of similar mass or HAC.
The data show a significant correlation – albeit tending
toward the PBF-axis – reinforcing the idea that if a fragment
shows a high degree of three-dimensionality using one deDrug Discovery Today: Technologies | Vol. xxx, No. xx 2019
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
Fig. 3. (Continued).
10 www.drugdiscoverytoday.com
scriptor, it is likely to also have a high score in the other
descriptor. From these data, a modest 431 out of the total 897
compounds (48%) possess a PBF score 0.60 [30], and so by
this metric only around half of the designed library members
are deemed to be 3D. This may be slightly biased, however, in
that it is more commonplace for PMI-guided design to drive
library generation (Table 1), and therefore it is more likely
that the 3D fragments exceed the PMI cut-off (88%, vide
supra) compared to the PBF cut-off. Indeed, 14 of the 25
libraries use PMI analysis in some form, compared to only two
which use PBF. In addition, 430 of the 897 fragments (48%)
fulfill the criteria of both descriptors and are therefore
deemed 3D according to the two orthogonal descriptors.
Only one fragment (0.2%) does not fulfill a SNPR 1.07
whilst satisfying a PBF score 0.60, indicating that if a
molecule possesses a PBF score exceeding 0.60, then it is
highly likely to have SNPR 1.07. However, this is not true
for the reverse. This could be an indication that the PMI cutoff may be too loose compared to that of the PBF analysis, and
so a greater proportion of compounds surpass its requirement
for three-dimensionality.
In terms of application, five of the 25 publications (20%)
present some form of screening data for the generated fragments, showing a diverse set of hits that are active against
different targets and possess accessible growth vectors for
further elaboration [45,53,56,59,60]. The majority of publications discussed do not include screening data as they focus
predominantly or entirely on library design and synthesis. It
is likely that screening results will follow in the near future,
allowing for better judgement of the effect of increased threedimensionality on the success of fragment screening.
Conclusion
A standardised analysis of the physicochemical and 3D properties of 25 published synthetic 3D fragment libraries has
been conducted. Most libraries tend to be RO3-compliant,
particularly in terms of abiding by the MW (and HAC) cut-off,
whilst being more flexible in the other parameters. Each of
the descriptors through which the 3D shape and diversity of
fragment libraries may be assessed possesses its own strengths
and weaknesses. This review has inspected molecular shape
using these descriptors across a range of scaffolds in the recent
literature, thereby allowing for better comparison, as bias
towards any one scaffold is removed. Fsp3 is the simplest
of the discussed descriptors and the two most popular shapebased methods currently deployed are PBF and PMI analyses,
with the latter being preferable likely due to it being more
information-rich in terms of molecular shape. Overall, PMI
and Fsp3 seem to be the two descriptors that are most used in
publications regarding 3D fragment synthesis. Fsp3 lacks
conformational information, which is a key factor in the
assessment of three-dimensionality, and our analysis conVol. xxx, No. xx 2019 Drug Discovery Today: Technologies |
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
PBF
∑NPR
Drug Discovery Today: Technologies
Fig. 4. Scatter plot depicting the SNPR values and corresponding PBF scores for each compound examined. 3D descriptor cut-off values of SNPR 1.07
and PBF score 0.60 are shown by dashed lines
www.drugdiscoverytoday.com 11
firms the previously identified [30,31] lack of correlation
between the Fsp3 and PBF/PMI descriptors. We have shown
that the analysed libraries possess high SNPR and PBF values,
with most libraries exceeding the established cut-offs for
three-dimensionality. The best assessment will likely be made
when several descriptors are used in conjunction. Whilst it is
important to achieve diversity in molecular shape, a library
can also be judged on its range of molecular physicochemical
properties, with the ideal collection possessing both. Scaffold
diversity-based synthetic approaches such as DOS, at the
fragment level, appear to successfully attain a high level of
both shape and chemical diversity. The valuable efforts towards sophisticated synthetic methodology, which do not
necessarily focus on library design, tend to afford more
restricted libraries. Complementary to these approaches, biocatalysis may offer an additional gateway into new 3D fragment space [63]. It is our expectation that the collective
pursuit for 3D fragments will continue and that the fruits
of these creative efforts to bolster the Screening Library 3D character of fragment screening libraries willreveal whether 3D fragments can
be translated into successful screening cascades and clinical
candidates in the near future.
Conflict of interest
The authors declare no conflict of interest.
Acknowledgments
We acknowledge funding from the European Union’s Framework Programme for Research and Innovation Horizon 2020
(2014–2020) under the Marie-Skoldowska-Curie grant agreement number 675899 (‘‘Fragment based drug discovery Network, FRAGNET’’), the Dutch Research Council under
Applied and Engineering Sciences grant number 18019
(‘‘Ready for growth: a new generation of highly versatile
fragment libraries’’) and The Royal Society (Industry Fellowship, INFR1191028).
Appendix A. Supplementary data
Supplementary material related to this article can be found,
in the online version, at doi:https://doi.org/10.1016/j.ddtec.
2021.05.001.
References
[1] Lamoree B, Hubbard RE. Current perspectives in fragment-based lead
discovery (FBLD). Essays Biochem 2017;61:453–64. http://dx.doi.org/
10.1042/EBC20170028.
[2] Erlanson DA, de Esch IJP, Jahnke W, Johnson CN, Mortenson PN.
Fragment-to-lead medicinal chemistry publications in 2018. J Med Chem
2020;63:4430–44. http://dx.doi.org/10.1021/acs.jmedchem.9b01581.
[3] Jahnke W, Erlanson DA, de Esch IJP, Johnson CN, Mortenson PN, Ochi Y,
et al. Fragment-to-lead medicinal chemistry publications in 2019. J Med
Chem 2020;63:15494–507. http://dx.doi.org/10.1021/acs.
jmedchem.0c01608.
[4] Bollag G, Tsai J, Zhang J, Zhang C, Ibrahim P, Nolop K, et al. Vemurafenib:
the first drug approved for BRAF-mutant cancer. Nat Rev Drug Discov
2012;11:873–86. http://dx.doi.org/10.1038/nrd3847.
[5] Kim A, Cohen MS. The discovery of vemurafenib for the treatment of
BRAF-mutated metastatic melanoma. Expert Opin Drug Discov
2016;11:907–16. http://dx.doi.org/10.1080/17460441.2016.1201057.
[6] Souers AJ, Leverson JD, Boghaert ER, Ackler SL, Catron ND, Chen J, et al.
ABT-199, a potent and selective BCL-2 inhibitor, achieves antitumor
activity while sparing platelets. Nat Med 2013;19:202–8. http://dx.doi.
org/10.1038/nm.3048.
[7] Murray CW, Newell DR, Angibaud P. A successful collaboration between
academia, biotech and pharma led to discovery of erdafitinib, a selective
FGFR inhibitor recently approved by the FDA. MedChemComm
2019;10:1509–11. http://dx.doi.org/10.1039/C9MD90044F.
[8] Zhang C, Ibrahim PN, Zhang J, Burton EA, Habets G, Zhang Y, et al. Design
and pharmacology of a highly specific dual FMS and KIT kinase inhibitor.
Proc Natl Acad Sci 2013;110:5689–94. http://dx.doi.org/10.1073/
pnas.1219457110.
[9] Tap WD, Wainberg ZA, Anthony SP, Ibrahim PN, Zhang C, Healey JH,
et al. Structure-guided blockade of CSF1R kinase in tenosynovial giant-cell
tumor. N Engl J Med 2015;373:428–37. http://dx.doi.org/10.1056/
NEJMoa1411366.
[10] Erlanson DA. Fragments in the clinic: 2020 edition. Pract Fragm Blog 2021
[Accessed 18 January 2021]http://practicalfragments.blogspot.com/2020/
03/fragments-in-clinic-2020-edition.html.
[11] Sauer WHB, Schwarz MK. Molecular shape diversity of combinatorial
libraries: a prerequisite for broad bioactivity. J Chem Inf Comput Sci
2003;43:987–1003. http://dx.doi.org/10.1021/ci025599w.
[12] Hann MM, Leach AR, Harper G. Molecular complexity and its impact on
the probability of finding leads for drug discovery. J Chem Inf Comput Sci
2001;41:856–64. http://dx.doi.org/10.1021/ci000403i.
[13] Meyers J, Carter M, Mok NY, Brown N. On the origins of threedimensionality in drug-like molecules. Future Med Chem 2016;8:1753–
67. http://dx.doi.org/10.4155/fmc-2016-0095.
[14] (a) Lovering F, Bikker J, Humblet C. Escape from flatland: increasing
saturation as an approach to improving clinical success. J Med Chem
2009;52:6752–6. http://dx.doi.org/10.1021/jm901241e; (b) Lovering F.
Escape from Flatland 2: Complexity and promiscuity. Medchemcomm
2013;4:515–9. http://dx.doi.org/10.1039/c2md20347b.
[15] Keserá GM, Hann MM. What is the future for fragment-based drug
discovery? Future Med Chem 2017;9:1457–60. http://dx.doi.org/10.4155/
fmc-2017-0106.
[16] Prosser KE, Stokes RW, Cohen SM. Evaluation of 3-dimensionality in
approved and experimental drug space. ACS Med Chem Lett
2020;11:1292–8. http://dx.doi.org/10.1021/acsmedchemlett.0c00121.
[17] Yan A, Gasteiger J. Prediction of aqueous solubility of organic compounds
based on a 3D structure representation. J Chem Inf Comput Sci
2003;43:429–34. http://dx.doi.org/10.1021/ci025590u.
[18] Milroy LG, Brunsveld L, Ottmann C. Stabilization and inhibition of
protein-protein interactions: the 14-3-3 case study. ACS Chem Biol
2013;8:27–35. http://dx.doi.org/10.1021/cb300599t.
[19] Valenti D, Hristeva S, Tzalis D, Ottmann C. Clinical candidates
modulating protein-protein interactions: the fragment-based experience.
Eur J Med Chem 2019;167:76–95. http://dx.doi.org/10.1016/j.
ejmech.2019.01.084.
[20] Johnson JA, Nicolaou CA, Kirberger SE, Pandey AK, Hu H, Pomerantz
WCK. Evaluating the advantages of using 3D-enriched fragments for
targeting BET bromodomains. ACS Med Chem Lett 2019;10:1648–54.
http://dx.doi.org/10.1021/acsmedchemlett.9b00414.
[21] Morley AD, Pugliese A, Birchall K, Bower J, Brennan P, Brown N, et al.
Fragment-based hit identification: thinking in 3D. Drug Discov Today
2013;18:1221–7. http://dx.doi.org/10.1016/j.drudis.2013.07.011.
[22] Walters WP, Green J, Weiss JR, Murcko MA. What do medicinal chemists
actually make? A 50-year retrospective. J Med Chem 2011;54:6405–16.
http://dx.doi.org/10.1021/jm200504p.
[23] Murray CW, Rees DC. Opportunity knocks: organic chemistry for
fragment-based drug discovery (FBDD). Angew Chem Int Ed 2016;55:488–
92. http://dx.doi.org/10.1002/anie.201506783.
Drug Discovery Today: Technologies | Vol. xxx, No. xx 2019
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
12 www.drugdiscoverytoday.com
[24] Blakemore DC, Castro L, Churcher I, Rees DC, Thomas AW, Wilson DM,
et al. Organic synthesis provides opportunities to transform drug
discovery. Nat Chem 2018;10:383–94. http://dx.doi.org/10.1038/s41557-
018-0021-z.
[25] St. Denis JD, Hall RJ, Murray CW, Heightman TD, Rees DC. Fragmentbased drug discovery: opportunities for organic synthesis. RSC Med Chem
2021. http://dx.doi.org/10.1039/D0MD00375A.
[26] Bostro¨m J, Brown DG, Young RJ, Keseru¨ GM. Expanding the medicinal
chemistry synthetic toolbox. Nat Rev Drug Discov 2018;17:709–27.
http://dx.doi.org/10.1038/nrd.2018.116.
[27] Caplin MJ, Foley DJ. Emergent synthetic methods for the modular
advancement of sp3
-rich fragments. Chem Sci 2021;12:4646–60. http://
dx.doi.org/10.1039/D1SC00161B.
[28] Kombo DC, Tallapragada K, Jain R, Chewning J, Mazurov AA, Speake JD,
et al. 3D molecular descriptors important for clinical success. J Chem Inf
Model 2013;53:327–42. http://dx.doi.org/10.1021/ci300445e.
[29] Tajabadi FM, Campitelli MR, Quinn RJ. Scaffold flatness: reversing the
trend. Springer Sci Rev 2013;1:141–51. http://dx.doi.org/10.1007/s40362-
013-0014-7.
[30] Firth NC, Brown N, Blagg J. Plane of best fit: a novel method to
characterize the three-dimensionality of molecules. J Chem Inf Model
2012;52:2516–25. http://dx.doi.org/10.1021/ci300293f.
[31] Downes TD, Jones SP, Klein HF, Wheldon MC, Atobe M, Bond PS, et al.
Design and synthesis of 56 shape-diverse 3D fragments. Chem A Eur J
2020;26:8969–75. http://dx.doi.org/10.1002/chem.202001123.
[32] Morrison CN, Prosser KE, Stokes RW, Cordes A, Metzler-Nolte N, Cohen
SM. Expanding medicinal chemistry into 3D space: metallofragments as
3D scaffolds for fragment-based drug discovery. Chem Sci 2020;11:1216–
25. http://dx.doi.org/10.1039/c9sc05586j.
[33] Reddy Guduru SK, Chamakuri S, Raji IO, MacKenzie KR, Santini C, Young
DW. Synthesis of enantiomerically pure 3-substituted piperazine-2-acetic
acid esters as intermediates for library production. J Org Chem
2018;83:11777–93. http://dx.doi.org/10.1021/acs.joc.8b01708.
[34] Chamakuri S, Jain P, Reddy Guduru SK, Arney JW, Mackenzie KR, Santini
C, et al. Synthesis of enantiomerically pure 6-substituted-piperazine-2-
acetic acid esters as intermediates for library production. J Org Chem
2018;83:6541–55. http://dx.doi.org/10.1021/acs.joc.8b00854.
[35] Jain P, Raji IO, Chamakuri S, MacKenzie KR, Ebright BT, Santini C, et al.
Synthesis of enantiomerically pure 5-substituted piperazine-2-acetic acid
esters as intermediates for library production. J Org Chem 2019;84:6040–
64. http://dx.doi.org/10.1021/acs.joc.9b00148.
[36] Congreve M, Carr R, Murray C, Jhoti H. A ‘Rule of Three’ for fragmentbased lead discovery? Drug Discov Today 2003;8:876–7. http://dx.doi.org/
10.1016/S1359-6446(03)02831-9.
[37] Hung AW, Ramek A, Wang Y, Kaya T, Wilson JA, Clemons PA, et al. Route
to three-dimensional fragments using diversity-oriented synthesis. Proc
Natl Acad Sci U S A 2011;108:6799–804. http://dx.doi.org/10.1073/
pnas.1015271108.
[38] Davis OA, Hughes M, Bull JA. Copper-catalyzed N-arylation of 2-
imidazolines with aryl iodides. J Org Chem 2013;78:3470–5. http://dx.
doi.org/10.1021/jo400120r.
[39] Morgan KF, Hollingsworth IA, Bull JA. 2-(Aryl-sulfonyl)oxetanes as
designer 3-dimensional fragments for fragment screening: synthesis and
strategies for functionalisation. Chem Commun 2014;50:5203–5. http://
dx.doi.org/10.1039/c3cc46450d.
[40] Morgan KF, Hollingsworth IA, Bull JA. Studies on the synthesis, stability
and conformation of 2-sulfonyl-oxetane fragments. Org Biomol Chem
2015;13:5265–72. http://dx.doi.org/10.1039/c5ob00549c.
[41] Foley DJ, Doveston RG, Churcher I, Nelson A, Marsden SP. A systematic
approach to diverse, lead-like scaffolds from a,a-disubstituted amino
acids. Chem Commun 2015;51:11174–77. http://dx.doi.org/10.1039/
c5cc03002a.
[42] Tran NC, Dhondt H, Flipo M, Deprez B, Willand N. Synthesis of
functionalized 2-isoxazolines as three-dimensional fragments for
fragment-based drug discovery. Tetrahedron Lett 2015;56:4119–23.
http://dx.doi.org/10.1016/j.tetlet.2015.05.035.
[43] Palmer N, Peakman TM, Norton D, Rees DC. Design and synthesis of
dihydroisoquinolones for fragment-based drug discovery (FBDD). Org
Biomol Chem 2016;14:1599–610. http://dx.doi.org/10.1039/
C5OB02461G.
[44] Prevet H, Flipo M, Roussel P, Deprez B, Willand N. Microwave-assisted
synthesis of functionalized spirohydantoins as 3-D privileged fragments
for scouting the chemical space. Tetrahedron Lett 2016;57:2888–94.
http://dx.doi.org/10.1016/j.tetlet.2016.05.065.
[45] Wang Y, Wach J-Y, Sheehan P, Zhong C, Zhan C, Harris R, et al. Diversityoriented synthesis as a strategy for fragment evolution against GSK3b.
ACS Med Chem Lett 2016;7:852–6. http://dx.doi.org/10.1021/
acsmedchemlett.6b00230.
[46] Twigg DG, Kondo N, Mitchell SL, Galloway WRJD, Sore HF, Madin A, et al.
Partially saturated bicyclic heteroaromatics as an sp3
-enriched fragment
collection. Angew Chem Int Ed 2016;55:12479–83. http://dx.doi.org/
10.1002/anie.201606496.
[47] Prescher H, Koch G, Schuhmann T, Ertl P, Bussenault A, Glick M, et al.
Construction of a 3D-shaped, natural product like fragment library by
fragmentation and diversification of natural products. Bioorgan Med
Chem 2017;25:921–5. http://dx.doi.org/10.1016/j.bmc.2016.12.005.
[48] Chalyk BA, Isakov AA, Butko MV, Hrebeniuk KV, Savych OV, Kucher OV,
et al. Synthesis of 6-azaspiro[4.3]alkanes: innovative scaffolds for drug
discovery. Eur J Org Chem 2017;2017:4530–42. http://dx.doi.org/
10.1002/ejoc.201700536.
[49] Chawner SJ, Cases-Thomas MJ, Bull JA. Divergent synthesis of
cyclopropane-containing lead-like compounds, fragments and building
blocks through a cobalt catalyzed cyclopropanation of phenyl vinyl
sulfide. Eur J Org Chem 2017;2017:5015–24. http://dx.doi.org/10.1002/
ejoc.201701030.
[50] Hassan H, Marsden SP, Nelson A. Design and synthesis of a fragment set
based on twisted bicyclic lactams. Bioorganic Med Chem 2018;26:3030–3.
http://dx.doi.org/10.1016/j.bmc.2018.02.027.
[51] Mateu N, Kidd SL, Kalash L, Sore HF, Madin A, Bender A, et al. Synthesis of
structurally diverse N-substituted quaternary-carbon-containing small
molecules from a,a-disubstituted propargyl amino esters. Chem A Eur J
2018;24:13681–87. http://dx.doi.org/10.1002/chem.201803143.
[52] Garner P, Cox PB, Rathnayake U, Holloran N, Erdman P. Design and
synthesis of pyrrolidine-based fragments that sample three-dimensional
molecular space. ACS Med Chem Lett 2019;10:811–5. http://dx.doi.org/
10.1021/acsmedchemlett.9b00070.
[53] Zhang R, McIntyre PJ, Collins PM, Foley DJ, Arter C, von Delft F, et al.
Construction of a shape-diverse fragment set: design, synthesis and screen
against aurora-A kinase. Chem A Eur J 2019;25:6831–9. http://dx.doi.org/
10.1002/chem.201900815.
[54] Sveiczer A, North AJP, Mateu N, Kidd SL, Sore HF, Spring DR. Spirocycles as
rigidified sp3
-rich scaffolds for a fragment collection. Org Lett
2019;21:4600–4. http://dx.doi.org/10.1021/acs.orglett.9b01499.
[55] King TA, Stewart HL, Mortensen KT, North AJP, Sore HF, Spring DR.
Cycloaddition strategies for the synthesis of diverse heterocyclic
spirocycles for fragment-based drug discovery. Eur J Org Chem
2019;2019:5219–29. http://dx.doi.org/10.1002/ejoc.201900847.
[56] Troelsen NS, Shanina E, Gonzalez-Romero D, Dankova´ D, Jensen ISA,
Sniady KJ, et al. The 3F library: fluorinated Fsp3
-rich fragments for
expeditious 19F NMR based screening. Angew Chem Int Ed 2020;59:2204–
10. http://dx.doi.org/10.1002/anie.201913125.
[57] Hanby AR, Troelsen NS, Osberger TJ, Kidd SL, Mortensen KT, Spring DR.
Fsp3
-rich and diverse fragments inspired by natural products as a
collection to enhance fragment-based drug discovery. Chem Commun
2020;56:2280–3. http://dx.doi.org/10.1039/c9cc09796a.
[58] Cox B, Zdorichenko V, Cox PB, Booker-Milburn KI, Paumier R, Elliott LD,
et al. Escaping from flatland: substituted bridged pyrrolidine
fragments with inherent three-dimensional character. ACS Med Chem
Lett 2020;11:1185–90. http://dx.doi.org/10.1021/
acsmedchemlett.0c00039.
[59] Pandey AK, Kirberger SE, Johnson JA, Kimbrough JR, Partridge DKD,
Pomerantz WCK. Efficient synthesis of 1,4-thiazepanones and 1,4-
thiazepanes as 3D fragments for screening libraries. Org Lett
2020;22:3946–50. http://dx.doi.org/10.1021/acs.orglett.0c01230.
[60] Kidd SL, Fowler E, Reinhardt T, Compton T, Mateu N, Newman H, et al.
Demonstration of the utility of DOS-derived fragment libraries for rapid
Vol. xxx, No. xx 2019 Drug Discovery Today: Technologies |
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001
www.drugdiscoverytoday.com 13
hit derivatisation in a multidirectional fashion. Chem Sci 2020;11:10792–
801. http://dx.doi.org/10.1039/d0sc01232g.
[61] Ko¨ster H, Craan T, Brass S, Herhaus C, Zentgraf M, Neumann L, et al. A
small nonrule of 3 compatible fragment library provides high hit rate of
endothiapepsin crystal structures with various fragment chemotypes. J
Med Chem 2011;54:7784–96. http://dx.doi.org/10.1021/jm200642w.
[62] Colclough N, Hunter A, Kenny PW, Kittlety RS, Lobedan L, Tam KY, et al.
High throughput solubility determination with application to selection of
compounds for fragment screening. Bioorg Med Chem 2008;16:6611–6.
http://dx.doi.org/10.1016/j.bmc.2008.05.021.
[63] Ramsden JI, Cosgrove SC, Turner NJ. Is it time for biocatalysis in fragmentbased drug discovery? Chem Sci 2020;11:11104–12. http://dx.doi.org/
10.1039/D0SC04103C.
Drug Discovery Today: Technologies | Vol. xxx, No. xx 2019
DDTEC-627; No of Pages 14
Please cite this article in press as: Hamilton DJ, et al. Escape from planarity in fragment-based drug discovery: A physicochemical and 3D property analysis of synthetic 3D fragment
libraries, Drug Discov Today: Technol (2021), https://doi.org/10.1016/j.ddtec.2021.05.001