Design and tests of prospective property predictions for novel antimalarial 2-aminopropylaminoquinolones

Article abstract of DOI:10.1007/s10822-020-00333-x

There is a pressing need to improve the efficiency of drug development, and nowhere is that need more clear than in the case of neglected diseases like malaria. The peculiarities of pyrimidine metabolism in Plasmodium species make inhibition of dihydroorotate dehydrogenase (DHODH) an attractive target for antimalarial drug design. By applying a pair of complementary quantitative structure–activity relationships derived for inhibition of a truncated, soluble form of the enzyme from Plasmodium falciparum (s-PfDHODH) to data from a large-scale phenotypic screen against cultured parasites, we were able to identify a class of antimalarial leads that inhibit the enzyme and abolish parasite growth in blood culture. Novel analogs extending that class were designed and synthesized with a goal of improving potency as well as the general pharmacokinetic and toxicological profiles. Their synthesis also represented an opportunity to prospectively validate our in silico property predictions. The seven analogs synthesized exhibited physicochemical properties in good agreement with prediction, and five of them were more active against P. falciparum growing in blood culture than any of the compounds in the published lead series. The particular analogs prepared did not inhibit s-PfDHODH in vitro, but advanced biological assays indicated that other examples from the class did inhibit intact PfDHODH bound to the mitochondrial membrane. The new analogs, however, killed the parasites by acting through some other, unidentified mechanism 24–48?h before PfDHODH inhibition would be expected to do so.

Full text of DOI:10.1007/s10822-020-00333-x

Journal of Computer-Aided Molecular Design
https://doi.org/10.1007/s10822-020-00333-x
Design and tests of prospective property predictions for novel
antimalarial 2‑aminopropylaminoquinolones
Robert D. Clark1 · Denise N. Morris2 · Gary Chinigo3,6 · Michael S. Lawless1 · Jacques Prudhomme4 ·
Karine G. Le Roch4 · Maria José Lafuente5 · Santiago Ferrer5 · Francisco Javier Gamo5 · Robert Gadwood3 ·
Walter S. Woltosz1
Received: 9 April 2020 / Accepted: 21 July 2020
© The Author(s) 2020
Abstract
There is a pressing need to improve the efciency of drug development, and nowhere is that need more clear than in the case
of neglected diseases like malaria. The peculiarities of pyrimidine metabolism in Plasmodium species make inhibition of
dihydroorotate dehydrogenase (DHODH) an attractive target for antimalarial drug design. By applying a pair of comple-
mentary quantitative structure–activity relationships derived for inhibition of a truncated, soluble form of the enzyme from
Plasmodium falciparum (s-PfDHODH) to data from a large-scale phenotypic screen against cultured parasites, we were able
to identify a class of antimalarial leads that inhibit the enzyme and abolish parasite growth in blood culture. Novel analogs
extending that class were designed and synthesized with a goal of improving potency as well as the general pharmacokinetic
and toxicological profles. Their synthesis also represented an opportunity to prospectively validate our in silico property
predictions. The seven analogs synthesized exhibited physicochemical properties in good agreement with prediction, and
fve of them were more active against P. falciparum growing in blood culture than any of the compounds in the published
lead series. The particular analogs prepared did not inhibit s-PfDHODH in vitro, but advanced biological assays indicated
that other examples from the class did inhibit intact PfDHODH bound to the mitochondrial membrane. The new analogs,
however, killed the parasites by acting through some other, unidentifed mechanism 24–48 h before PfDHODH inhibition
would be expected to do so.
Keywords Antimalarial · ADME · Dihydroorotate dehydrogenase · Drug design · PBPK · QSAR
Introduction
Electronic supplementary material The online version of this
There were approximately 228 million malaria cases in
article (https://doi.org/10.1007/s10822-020-00333-x) contains
2018, with 405,000 deaths attributed to malaria [1]. Infec-
supplementary material, which is available to authorized users.
tion by Plasmodium falciparum and P. vivax are the most
common causes of the disease in humans, with the former
being responsible for the greatest mortality. Growing resist-
ance to currently available antimalarial drugs makes identi-
fcation of new compounds with novel modes of action and
activity against resistant parasites a matter of great urgency
[2–4].
* Robert D. Clark
bob@simulations-plus.com
1
Simulations Plus, Inc., 42505 10th Street West, Lancaster,
CA 93534-7059, USA
2
Cognigen Corporation, a Simulations Plus Company,
Bufalo, NY, USA
Finding new molecular targets is one way to reduce the
risk of cross-resistance developing after introduction of a
new antimalarial. Fortunately, biochemical idiosyncrasies
of the parasite provide several distinct targets that reduce
the risk of undesirable “of-target” efects. Unfortunately,
the expected return on investment for “neglected diseases”
like malaria is too low to motivate large-scale commercial
3
Kalexsyn, Inc., Kalamazoo, MI, USA
4
Department of Molecular, Cell and Systems Biology,
University of California, Riverside, CA, USA
5
Tres Cantos Medicines Development Campus-Diseases
of the Developing World, GlaxoSmithKline, Tres Cantos,
Madrid, Spain
6
Present Address: Pfzer Inc., Groton, CT, USA
Vol.:(0123456789)
1 3

Journal of Computer-Aided Molecular Design
development of compounds directed at those disease targets.
A few pharmaceutical companies have made substantial
contributions to public-sector drug discovery eforts (par-
ticularly in the form of screening results), but most work in
the area has been done by academic groups and nonproft
organizations.
drug design program. PBPK simulations using predicted
ADMET properties as input parameters were carried out
to determine whether virtual compounds were likely to
have good enough oral absorption and slow enough sys-
temic clearance to sustain a blood concentrations high
enough to kill the parasite given a dosing regimen accept-
able for existing antimalarial drugs (vide infra). Surviv-
ing virtual compounds were synthesized and characterized
experimentally.
Funding constraints and limited synthesis resources make
in silico and collaborative approaches particularly attrac-
tive for such noncommercial applications. Computational
methods being employed range from constructing quantita-
tive models of the relationship of molecular structure to the
specifc biological activity of interest (QSARs) or to more
general molecular properties related to in vivo absorption,
distribution, metabolism, excretion, and toxicity (ADMET
QSPRs). Physiologically based pharmacokinetic (PBPK)
simulations are also potentially useful, in that they are able
to combine individual system-specifc properties with drug-
specifc information in order to anticipate how a compound
will behave in vivo. Such simulations are needed to take
into account complex interactions between physicochemi-
cal properties—e.g., lipophilicity, solubility, and permeabil-
ity. The goal here is to kill the parasites, of course, which
makes such modeling applications somewhat diferent from
those in which one is trying to adjust some physiological
imbalance. The situation is similar to oncology, though the
adversary in the case of malaria is—fortunately—more con-
sistent in its presentation. We were targeting the blood-stage
parasite, so the goal was to get compounds to red blood cells
in the systemic circulation and keep them there; distribution
to peripheral tissues was not particularly desirable.
The methodology described herein focuses on the use
of in silico tools for identifying novel lead compounds for
synthesis in a way that exploits QSAR and PBPK mod-
eling to minimize the bench resources needed. The work-
fow can be characterized as follows:
1. Use enzymatic screening data to generate a predictive
QSAR model for an attractive target.
2. Apply the QSAR obtained to a large and diverse library
of structures and associated phenotypic activity to
identify a lead series predicted to exhibit good ADMET
properties in aggregate.
3. Create unique combinations of active substituents from
discovered scafolds.
4. Predict activity, ADMET properties, and PK profles
based on QSAR and PBPK models.
5. Choose and synthesize attractive candidates from the
analogs identifed.
6. Measure biological activities and ADMET properties of
the synthesized compounds.
7. Iterate as needed.
Eforts to use virtual screening to identify new antimalari-
als have had limited success. Zhang et al. [5], for example,
built QSAR models using training data from a 3,133-com-
pound library that contained 158 confrmed actives. The
models obtained were used to screen a commercial data-
base (ChemBridge, San Diego, CA) in order to identify new
potential leads. From this virtual screen, 176 compounds
representing 22 unique scafolds were identifed and tested
for growth inhibition. Of these, 7 were active against P. fal-
ciparum chloroquine-susceptible (3D7) and chloroquine-
resistant (K1) strains ex vivo with a concentration required
to kill half of the organisms (XC50) of 1 μM or less [5].
Here, we used QSAR models and PBPK simulations
to pick an attractive lead series from “hits” in a pheno-
typic assay, then extended that series by generating a
virtual library of novel analogs. An array of predicted
properties—PfDHODH inhibition, physiochemical prop-
erties, and pharmacokinetic profles—were used to select
a handful of candidates for further evaluation. The full
range of predicted properties taken into consideration are
spelled out in Supplementary Data Table S2. The ADMET
Risks [6] they highlight can reduce oral bioavailability or
pose toxicity problems, which makes them relevant in any
The work described here covers a single iteration of
this workfow that provided us with an opportunity to
assess the ability of our models to make good activity
and ADMET property predictions for novel compounds.
It also represents a proof-of-concept exercise of our lead
discovery and optimization tools and analytical methods
developed over the past several years. Our hope was that
we might, in the process, also contribute something of
value to the search for novel antimalarials.
As a proof-of-concept project, the scale of our work
was necessarily limited. The scope was nonetheless broad
and deep enough that we encountered several challenges
typical of larger drug development projects. Those chal-
lenges included needing to revise our original design to
incorporate more practical synthesis schemes, to revise
synthesis targets to work around failed reactions, and to
encountering an of-target mode of action. Fortunately, in
our case the of-target activity was one that results in more
rapid death of the parasite than inhibiting our intended
target does.
1 3

Journal of Computer-Aided Molecular Design
Table 1 Properties and
Statistic
Model A (1×9a)
Trainb
Model B (2×29)
performance of the Ki models
Verifyc
Testd
Train
Verify
Test
MAE
SRCC
Q2
0.51
0.77
–
0.52
0.78
0.83
0.44
0.76
0.68
0.29
0.92
–
0.37
0.88
0.90
0.43
0.88
0.71
RMSE root mean square error, MAE mean absolute error, SRCC Spearman’s rank correlation coefcient,
Q2 predictive relevance for the verifcation and test sets
aANN architecture indicated by number of neurons x number of input descriptors
bTraining set
cInternal test set used trigger early stopping
dHold out test set [6]
(fewer false positives) and model B being more sensitive
(more prone to generating false negatives). Hence, using
them together to identify potent compounds worked better
than using either model alone. Plotting the observed versus
predicted values from the models illustrated the ability of
both models to adequately predict the Ki values from both
the training set that was used to build the models and the test
set used to validate the models (Fig. 1).
Lead selection and compound design
Plasmodium species cannot salvage preformed pyrimi-
dine bases for nucleic acid synthesis as its human hosts
can. Dihydroorotate dehydrogenase (DHODH) is a critical
enzyme in the de novo pyrimidine synthesis pathway in the
parasite and, thus, a potential target for antimalarial drug
therapies [7, 8]. The enzyme from Plasmodium species is
located in the mitochondrion and utilizes ubiquinone (also
known as coenzyme Q) as a cosubstrate.
Selecting a lead series
Predicting DHODH inhibition
The developed QSAR models were used to analyze struc-
tures for 13,533 antimalarials identifed in a phenotypic
screen of nearly 2 million compounds that was carried out
by GlaxoSmithKline (GSK). The published data included
P. falciparum growth inhibition for the chloroquine-sensi-
tive 3D7 (PubChem AID 2306) as well as for the multid-
rug resistant strain Dd2 (PubChem AID 2302). Data from
a counter screen for mammalian cytotoxicity was available
from PubChem AID 2303 [19].
Data on PfDHODH inhibition were collected from the lit-
erature for a structurally diverse group of compounds. These
included triazolopyrimidines [7, 9], diphenylureas [10],
and 2-cyano-3-hydroxy acrylanilides [11], among others
[12–15]. Quantitative data were not available for inhibition
of the intact enzyme, which is bound to the mitochondrial
membrane by its hydrophobic N-terminus in vivo [16]. A
total of 89 of those compounds had been assayed against
the soluble recombinant form of PfDHODH (s-PfDHODH)
from which the hydrophobic “tail” has been removed. The
very low solubility of CoQ-10 limits its utility in vitro, so
an artifcial ubiquinone (dodecylubiquinone [DQ]) is usually
used as substrate instead.
There were some important limitations in working with
this data set. First, evaluating efcacy against parasites cul-
tured in human erythrocytes provides no direct information
on a compound’s mode of action. Secondly, only results for
the active compounds (those that inhibited parasite growth
by at least 80% at a concentration of 2 μM) were reported.
The lack of data on inactives posed a considerable risk of
synthesis resources being wasted on seemingly “novel”
analogs that in fact had already been synthesized but not
reported due to their lack of activity. Lastly, many of the
compounds in the screening data set exhibited undesirable
physical properties and/or undesirable substructures (e.g.,
multiple halogenated thiophene rings) that would cause them
to be avoided by many pharmaceutical companies [20].
Candidate lead series from the active compounds in
the phenotypic screening data set were identified and
structural classes were generated in MedChem Studio™
[21] based on shared substructures. Unlike most other
approaches [22], the method used in MedChem Studio
The IC50 data for these 89 compounds were converted
to inhibition constants (Ki’s) by assuming competitive
inhibition against DQ and applying the method described
by Cheng and Prusof [17]. Two artifcial neural network
ensemble (ANNE) regression models were constructed
(henceforth referred to as Model A and Model B) using
ADMET Modeler™ module of ADMET Predictor® [18].
The models were built on distinct 20% hold-out test sets
(see Supplementary Methods for details). That and the use
of diferent random number seeds led to somewhat difer-
ent descriptor sets, complexities and performance statistics
(Table 1). The two models complemented each other well,
with Model A yielding more conservative activity estimates
1 3

Journal of Computer-Aided Molecular Design
Fig. 1 Performance plots for the ANNE regression models developed
for in vitro PfDHODH inhibition built from literature data. The plot
for Model A is on the left and the plot for Model B is on the right.
Labeled points correspond to data from Supplementary Table S1: (a)
Tz18; (b) P05; (c) G47; (d) D10; and (e) Tz11
does not cluster compounds (in this case, actives) into
exclusive subsets based on the degree of pairwise overlap
in their substructural fngerprints. Rather, it generates non-
exclusive classes of molecules that share a maximal com-
mon substructure. In most cases, this shared substructure
is, itself, a scafold in the medicinal chemistry sense. If
not, it can readily be modifed to become one. The non-
exclusivity is advantageous for lead series identifcation
because it is typically easier to see cases where overlap-
ping classes can proftably be combined by merging the
scafolds that defne them.
Results for three classes at an intermediate stage of the
series defnition process are shown in Fig. 2. Two of the
classes—diphenyl ureas (DPUs) and triazolopyrimidines
(Tzs)—represent PfDHODH inhibitor lead series that had
already been explored or were being actively explored by
other research groups [24, 27].
The 2-(3-aminopropylamino)-4-quinolone (APAQ) class
was particularly promising because its scafold bore a dis-
tinct resemblance to ubiquinone, which is a cosubstrate for
PfDHODH in vivo (Fig. 3). They also resembled N-hydroxy-
2-dodecyl-4-quinolone (HDQ) (Fig. 3) which Dong et al.
had determined to have good antimalarial activity, with PfD-
HODH as its primary target [25]. Although they reported
that HDQ inhibited parasite growth in culture and were able
to show that PfDHODH was the site of action, they did not
have access to soluble enzyme. Hence the XC50 obtained
in culture could not be translated into an in vitro Ki for the
enzyme. HDQ was therefore not included in the data set used
to build the PfDHODH inhibition models. Two quinolones
found among the “hits” reported by Patel et al. [8] were
initially grouped with some of the APAQs. These lack the
defning 2-amino substitution and so did not appear in the
fnal lead series, which was comprised of 113 actives.
None of the APAQs were reported to be cytotoxic to the
mammalian HepG2 cells used in the companion AID 2303
screen, which suggests that the intrinsic risk of mammalian
toxicity is small. That said, many of them had discouragingly
high ADMET Risk scores. In particular, the two most active
examples (Fig. 3) violated 7 and 9 of the standard ADMET
Risk rules, respectively. Both are large and were predicted to
be excessively hydrophobic, to be highly bound in plasma, to
exhibit low water solubility, and to be acutely toxic to rats. In
addition, CID 44535189 has an excessively large number of
Classes were evaluated as potential lead series based on
the QSAR-predicted inhibitory potencies against PfDHODH
and the degree of parasite growth inhibition observed in the
phenotypic screen for chloroquine-sensitive P. falciparum
3D7 (PubChem AID 2306). ADMET Risk™, which indi-
cates how many of an array of ADMET property rules are
violated by a compound [6], was also taken into consid-
eration. The rule thresholds are calibrated against a subset
of the World Drug Index (WDI) that is enriched in orally
delivered commercial drugs similar to that used by Lipinski,
et al., for anticipating solubility and oral absorption prob-
lems [23]. Ninety percent of the WDI reference set have an
ADMET Risk score less than seven.
Low group averages were preferred for both predicted Ki
and ADMET Risk, but the spread in each profle was also
important. Lack of variation in potency indicates a “lack of
SAR”, which makes it difcult to enhance activity against
the target by modifying structure. Similarly, a lack of vari-
ation in ADMET Risk suggests that it will be difcult to
engineer out potential liabilities by changing structures. The
exception would be a class wherein all ADMET Risk scores
are very low, and that was never the case in this dataset.
1 3

Journal of Computer-Aided Molecular Design
Fig. 2 Comparing property distributions across three representative
classes: triazolopyrimidines (Tzs), aminopropylaminoquinolones
(APAQs) and diphenylureas (DPUs). The portion of each representa-
tive structure highlighted in blue corresponds to the class scafold.
Growth inhibition is shown as %inhibition vs. P. falciparum strains
3D7 and DD2, which are chloroquine-sensitive and -resistant, respec-
tively. pKi_pred values are the negative log of the predicted Ki in
µM, so a larger number indicates higher potency; their distribution
for Models A and B are shown. ADMET Risk is a measure of likely
development liabilities that can range from 0 to 24 [6]
Fig. 3 Scafold and structures of
compounds associated with the
APAQ lead series
rotatable bonds and was predicted to exhibit chronic toxicity
in mice. Besides the fve likely liabilities shared with CID
44535189, CID 44534046 was predicted to be metabolized
by CYP2D6 and CYP3A4, to inhibit CYP3A4, and to bind
to the human ether-a-go-go (hERG) gene product.
and property profles. Substituents from each R-group were
drawn from the known actives in the class. Doing so
increases the likelihood that the structures produced will be
synthetically accessible, since at least one example already
exists where that substituent was placed at that position. In
addition: only single substitutions were allowed at the dis-
tal nitrogen; only simple substituents (hydrogen, halogen,
methyl, or trifuoromethyl) were allowed on the quinolone
ring; and substitution was restricted on the distal aromatic
rings. The resulting constrained combinatorial R-group
explosion produced a virtual library of ~ 99,000 novel
analogs.
Analog generation and selection
The predictions of high lipophilicity, low water solubility,
high susceptibility to metabolism by cytochrome P450s
(especially CYP2D6), and hepatotoxicity were common
among the actives in the APAQ class. Substituents from the
less risky APAQs were combinatorially reshufed in silico
to generate lead candidates with more favorable activity
Attractive lead analogs from the virtual library were
selected based on their predicted PfDHODH inhibitory
1 3

Journal of Computer-Aided Molecular Design
power in both QSAR models, their predicted ADMET prop-
erties, and their expected ease of synthesis. Compounds for
which the predicted PfDHODH Ki for Model A was greater
than 0.1 µM and with ADMET Risk>4 were set aside, as
were compounds that were out-of-scope1 with respect to
Model A. Though Model B was useful for identifying good
lead series, it failed to provide much further discrimination
between APAQ analogs, hence whether its prediction for a
particular molecule was out-of-scope or not was disregarded.
As is often the case, descriptors for some of the com-
pounds ultimately selected for synthesis based on activity
predictions fell well outside the range of descriptors for com-
pounds used to train one or more of the metabolism models
available at the time. Predictions for such compounds were
considered “out-of-scope” extrapolations and, therefore, not
necessarily reliable. Compounds with in-scope predictions
were favored in selecting candidates for further analysis but
ignoring those for which any prediction was out-of-scope
was impractical. Such limitations in coverage are common
for any novel chemistry, but they are also a major motivation
for continuously refning and expanding ADMET models to
improve their predictive performance.
aminopropylamino bridge to a gem-cyclopentyl ring. No
such compound existed among the GSK actives, though
there were several derived from trans 1-amino-2-amino-
methylcyclopentane [19, 27]. A 1,2-ring is expected to adopt
a quite diferent 3D conformation from that for a 2,2-ring,
thereby placing the terminal group at quite diferent posi-
tions in space.
After the initial set of lead compounds were selected,
their expected pharmacokinetic profiles were evaluated
by applying the in silico predictions and standard human
physiologies in GastroPlus® [28] to predict plasma concen-
tration-versus-time profles. Property estimates were taken
from ADMET Predictor and dosing regimens were similar to
those used for existing antimalarial drugs, e.g., chloroquine.
Target plasma concentrations were based on predicted Ki
values.
An initial set of diverse synthesis candidates (Supplemen-
tary Table S2) was selected from among those predicted to
have good potency as well as an acceptable ADMET Risk
score and favorable simulated in vivo PK profiles. The
twelve synthesis targets produced by this process had struc-
tural variations at three diferent positions: the R1 substitu-
ent on the quinolone end of the scafold; the R2 and R3 sub-
stituents of of the distal nitrogen of the aminopropylamino
bridge; and the R4 and R5 substituents of of the central car-
bon in the bridge. The beginning of the reduction-to-practice
phase was announced in a press release in September 2011
[29] and bids were solicited from several contract synthe-
sis companies. Four of the targets distributed are shown in
Fig. 4.
Once analogs predicted to have relatively low potency
or discouraging ADMET properties had been fltered out,
approximately 34,001 analogs still remained. Of those,
17,334 bore symmetrical central diamine moieties, a fea-
ture expected to greatly simplify synthesis. ADMET Predic-
tor now includes a synthetic difculty score analogous to
that described to the synthetic accessibility score described
by Ertl and Schufenhauer [26]. That functionality was not
available when this project began, however, so assessment
was carried out manually by visual inspection of analogs
displayed in a tile view, i.e., in a gridded rather than a row
format. In retrospect, the manual process was reasonably
efective. The synthetic difculty scores for the symmetri-
cal diamine analogs ranged from 1.95 to 5.0 (median 3.62,
where 10 is most difcult) and the fnal dozen candidates
(vide infra) ranged from 2.5 to 4.15 (median 3.35), indicat-
ing that the fnal candidates were fairly representative of the
initial pool in terms of synthetic accessibility.
The individual compounds proposed were synthetically
feasible, but the number of separate intermediates and vari-
ations in reaction conditions required would have been pro-
hibitively expensive. Instead, variations were restricted to
either end of the molecule and we focused on a single kind
of aminopropylamino bridge. Doing so made it likely that
similar reaction conditions and intermediates could be used
to make multiple analogs. The convergent synthesis plan
greatly reduced the potential for complications and budget
overruns. Though not often dwelt on in the literature, such
considerations are often a critical practical consideration for
any molecular design project; hence their explicit inclusion
here. In the end, we settled on seven analogs built around a
simplifed scafold bearing the novel gem-cyclopentyl group
at the central carbon of the bridge. Diversity was provided
by placing three diferent substituents on the distal nitrogen
(i.e., R2) (Fig. 5).
Marginal aqueous solubility was a central concern that
had to be balanced against model predictions that substi-
tution at the central carbon of a 1,3-diaminopropyl bridge
should improve activity and reduce CYP metabolism.
This and other ADMET Risks prompted us to manually
expand the library by introducing small point changes to
some marginal analogs. One such change was “mutat-
ing” a gem-cyclohexyl ring at the central carbon of the
Kalexsyn, Inc. (Kalamazoo, MI) synthesized a revised
list of seven target analogs. Carbonyl precursors suitable for
reductive amination were available or accessible for some of
the desired R2 substituents, considerably simplifying their
synthesis.
1
A prediction was fagged as “out of scope” when any of its descrip-
tor values fell more than 10% outside the range of the corresponding
values seen for the data set upon which the model was trained.
1 3

Journal of Computer-Aided Molecular Design
Fig. 4 Four of the 12 APAQ
synthesis targets initially put out
for bids
Fig. 5 Initial scafold used
for R-Group explosion versus
the fnal, simplifed scafold
shared by the analogs that were
synthesized
Scheme 1 Syntheses of three
variations on the free amino
APAQ scafold 7
5
7a: X = H
7b: X = Cl
7c: X = OMe
5 +
6a: X = H
6b: X = Cl
6c: X = OMe
synthesis required a distinct approach. Fortunately, both
approaches were relatively straightforward. What is pro-
vided here is an overview for illustrative purposes. That
said, it is complete enough to give a sense of how much
efort was involved in generating the seven targeted ana-
logs. More details are provided in the Supplementary
Materials.
Experimental results and discussion
The candidates chosen under the plan outlined above
struck a reasonable balance between structural diversity
and being accessible via reasonably convergent and paral-
lel synthetic methodologies. Despite our hopes and best
intentions, however, the particular substituents targeted for
the distal nitrogen still needed some tweaking as regards
how the coupling was carried out. Similarly, although
the 6-methoxy analog moiety is very similar in structure
to that in the 6-chloro and unsubstituted quinolones, its
Synthesis
The general synthesis procedure used is outlined in
Scheme 1. gem-Dicyanocyclopentane was prepared by
1 3

Journal of Computer-Aided Molecular Design
Scheme 2 Synthesis of
2-chloro-4,6-dimethoxy qui-
noline
6c
Scheme 3 Synthesis of targeted
APAQs 8–11
7a
8
7a
9
10a: X = H
10b: X = Cl
R = Boc
R = H
7a-c
10c: X = OMe
11a: X =H
11b: X = Cl
11c: X = OMe
condensation of malononitrile with dibromobutane and
reduced to the diamine 5 using lithium hydride in tetrahy-
drofuran. Subsequent reaction with 2-chloro-4-methoxyqui-
nolines yielded 7a–c after hydrolysis in HCl.
dihydroquinolone reacted cleanly. Deprotection with
ZnBr2 in dichloromethane (DCM) or transesterification
in formic acid followed by hydrolysis in NaOH afforded
the targeted analogs 11a–c (Scheme 3).
2-Chloro-4,6-dimethoxyquinoline (6c) was prepared by
condensation of malonic acid with p-anisidine, yielding
2,4-dichloro-6-methoxy quinoline [30], which was converted
to the desired dimethoxy compound by treatment with meth-
oxide in methanol (Scheme 2).
Analogs bearing an unsubstituted imidazolinone ring
in place of the dimethylthiophene ring in 8 were attrac-
tive in terms of predicted activity and ADMET properties.
Unfortunately, several attempts to prepare and purify them
failed. Some N-protected intermediates were generated,
but all broke down during deprotection or during purifca-
tion. The 1,3-dimethylimidazolinones 12a and 12b were
synthesized in their stead (Scheme 4). The added methyl
groups were predicted to reduce solubility somewhat,
but absorption was expected to be enhanced. Moreover,
the methyl groups were predicted to be cleaved of in the
liver to produce the original targets. This change in targets
necessitated the use of N,N’-dimethylimidazolinone-4-car-
boxaldehyde, which was prepared from ethylene glycol
and N,N’-dimethyl urea, then formylated with POCl3 in
dimethylformamide. Condensation with the respective free
amino precursor yielded the desired APAQs (Scheme 4).
Reductive alkylation of the distal amino group in 7a
with thiophene carbaldehyde to yield 8 proceeded readily
with cyanoborohydride in methanol at room temperature,
whereas preliminary formation of the imine and heating
were required to generate the tetrahydrocarbazole analog 9
and the tetrahydroquinolines 10a–c from the corresponding
ketones (Scheme 3).
Protection of the indole nitrogen in the former case
was not necessary. In fact, difficulties with deprotect-
ing the product made it counterproductive. Blocking the
amino group in the tetrahydroquinolone, on the other
hand, was essential. Though reductive amination with the
free amine failed, the t-butyloxycarbonyl(Boc)-protected
1 3

Journal of Computer-Aided Molecular Design
Scheme 4 Synthesis of imida-
zolinone analogs 12a and 12b
12a: X =H
12b: X = Cl
7a-b
Table 2 Predicted and measured physicochemical properties of the candidates
Compound
8
9
11a
11b
RMSE
Name
SLP0005
SLP0003
SLP0004
SLP0006
S+Sw (µg/mL)
obsd. solubility
1.8
33
0.32
0.76
3.4
32
1.2
22
11-fold
0.63
S+logP
obsd. logP
4.7
4.2
5.05
4.4
4.3
3.5
5.0
5.45
S+pKa1
5.24
4.95
5.24
4.80
5.29
4.22
5.14
4.75
0.63
obsd. pKa1
S+pKa2
8.01
8.33
7.63
7.36
7.74
7.80
7.55
7.55
0.21
obsd. pKa2
S+logD6.8
3.44
2.68
4.46
3.76
3.54
2.46
4.36
4.73
0.77
obsd. logD6.8
logP logarithm of the octanol:water partition coefcient, logD6.8 logP at pH 6.8, obsd observed, pKa1 negative logarithm of the frst dissociation
constant, pKa2 negative logarithm of second dissociation constant, RMSE, root mean squared error, S+ denotes a proprietary model from Simu-
lations Plus, Inc., Sw aqueous solubility
overall performance of the models in ADMET Predictor
is good, but interpretation is more complicated than for
the physicochemical properties. These are not predictions
of simple intrinsic clearance, as several of the relevant
concentrations at half-maximal metabolic velocity (Km’s)
are predicted to be low enough for partial saturation to
be an issue (Supplementary Table S3). This underscores
the desirability of having afnity models (here, for Km)
available to put data on microsomal stability obtained at a
single concentration into proper context.
ADMET properties: predicted vs. experimental
To become a drug candidate, a lead molecule must possess
many favorable pharmacological properties, including good
solubility, metabolic stability, and low toxicity as well as
activity. Hence, ADMET QSARs and QSPRs provide a valu-
able complement to activity models in avoiding unfavorable
ADMET profles, provided the predictions they provide are
accurate and reliable. Measured physicochemical properties
of the most active synthesized compounds were determined
experimentally and are compared with the predictions from
ADMET Predictor 9.0 in Table 2. Most of the observed and
predicted properties have a root mean squared error (RMSE)
of 0.8 log units or less. Aqueous solubility was an exception,
but not an unexpected one. That is because the compounds
were all isolated as foams that subsequently solidifed to
glassy solids, the solubilities of which can be quite difcult
to measure accurately.
The results for CYP2C9 and CYP2C19 in Table 3 also
illustrate why it is important to make regression models
for enzyme activity contingent on classifcation models
that distinguish substrates from nonsubstrates [6, 31].
Nonsubstrates are necessarily absent from the data used
to construct the regression models, and the classifcation
models confdently identify them [32] as being out-of-
scope for the corresponding regression models. The clear-
ance predictions provided represent the values expected
if they were substrates, which the classifcation models
confdently predict they are not.
Rates of in vitro clearance determined from a panel
of recombinant microsomes containing human CYPs
were measured, as was that by human liver microsomes
(HLMs). The results obtained are shown in Table 3. The
1 3

Journal of Computer-Aided Molecular Design
Table 3 Predicted and measured rates of metabolism by CYPs in vitro
Property
8
9
11a
11b
CYP
Name
SLP0005
SLP0003
SLP0004
SLP0006
1A2
Substratea
No (60%)
(6.9)
Yes (48%)
4.9
No (41%)
(5.5)
Yes (43%)
10.7
b
Pred. CL1µM
2-folde
Obsd. CL1µM
Substratea
10
9.3c
7.3c
NDd
2C9
2C19
2D6
3A4
No (89%)
(137)
No (77%)
(231)
No (96%)
(214)
No (76%)
(245)
b
Pred. CL1µM
Obsd. CL1µM
Substratea
4.1d
NDd
1.8d
NDd
(CLint predicted
as if they were
substrates)
No (94%)
(532)
No (93%)
(115)
No (87%)
(642)
No (97%)
(571)
b
Pred. CL1µM
Obsd. CL1µM
Substratea
14.6
3.8
3.5
2.2c
Yes (82%)f
Yes (63%)f
Yes (63%)f
Yes (82%)f
b
Pred. CL1µM
330
31
381
10.6
420
30
447
18
20-fold
Obsd. CL1µM
Substratea
Yes (92%)
41
Yes (98%)
135
Yes (98%)
155
Yes (98%)
84
b
Pred. CL1µM
3-fold
Obsd. CL1µM
256
179
298
263
HLM
Pred. CL
Obsd. CL
171
220
138
120
117
230
250
620
1.8-fold
CLint intrinsic clearance, CYP cytochrome P450, HLM human liver microsomes, ND not detected, Obsd,observed, Pred predicted. aPredicted to
be a substrate (yes/no; percent confdence)
bClearance at 1 µM expressed as µL/min/mg HLM protein. Clearance predictions for compounds predicted not to be substrates are set of by
parentheses
cPossibly a substrate
dUnlikely to be a substrate
eFold-errors calculated from the root mean square errors (RMSE) in the log for compounds predicted to be substrates
fPredicted to be an inhibitor as well as a substrate
The situation is somewhat diferent for CYP1A2. Here
the substrate classifcation predictions are uncertain—the
confdences are near toss-up (41–60%), but they are still
consistent with the low to nonexistent experimental clear-
ance values for this isoform.
CYP kinetics are complicated by the potential for auto-
inhibition [33]. The APAQs were not tested as inhibitors,
but all were predicted to inhibit CYP2D6. That may explain
the consistent overestimation of CYP2D6 clearance for
CYP2D6, for which the predicted afnities are high—i.e.,
for which the Km’s are all less than 1 μM (Supplementary
Table S3). All four analogs are also predicted to be CYP3A4
inhibitors, but the corresponding Km’s are 15 μM or above,
making autoinhibition unlikely to be a problem at 1 μM.
Pharmacokinetic predictions
The intrinsic clearances for the most active APAQs would
have posed a risk high enough to discourage further develop-
ment of these compounds had they not exhibited acceptable
pharmacokinetic profles using the other property predic-
tions. In fact, high predicted microsomal clearance was the
Fig. 6 Human concentration–time profle expected for 9 based on
PBPK simulation using GastroPlus. Pharmacokinetic parameters
were taken from experimental values where available and estimated
using the QSAR models in ADMET Predictor otherwise. Conc
concentration, Ki predicted inhibition constant for PfDHODH when
plasma protein binding is taken into account
1 3

Journal of Computer-Aided Molecular Design
Table 4 Antimalarial activity of APAQs in asynchronous blood cul-
reason several otherwise attractive candidates that failed to
yield acceptable pharmacokinetic profles were set aside dur-
ing analog selection. The result for compound 9 is shown
in Fig. 6. Here the predicted fraction unbound in plasma
(fup <3%) and high volume of distribution (Vd; estimated at
7 L/kg) ofset the relatively high predicted intrinsic clear-
ance. Basically, too little of the compound is likely to get
into the liver for hepatic metabolism to be a problem. As
a result, the bioavailability is predicted to be high enough
to produce a favorable in vivo profle, given an appropriate
dosing schedule—here, a schedule similar to that currently
used for chloroquine.
ture
Compound
Pred. PfDHODH XC50 (µM)a,b
Resistance
ratio ( )
Ki (µM)
3d7(−) Dd2(+)
12a
12b
11a
11c
8
0.049
0.051
0.023
0.037
0.037
0.025
0.038
10.0
1.61
0.55
0.37
0.30
0.106 0.21
0.037 0.24
0.89
0.85
46
6.4
2.3
1.78
1.47
4.6
3.9
4.1
4.8
5.0
2.0
6.6
5.2
10.1
9
11b
CID 44534046c 0.112
CID 44535189c 0.077
4.6
8.6
Both fup and Vd are in fact fagged as ADMET Risks by
default, because they can make it harder for a drug to get
from the plasma into peripheral tissues. That is not relevant
to the blood phase of the Plasmodium life cycle, but would
afect the ability of the compound to attack liver stages. The
fact that they can ofset potential frst-pass metabolic liabili-
ties is one good reason to examine pharmacokinetic simula-
tion results for compounds of interest that raise piece-wise
risk fags [31].
PfDHODH dihydroorotate dehydrogenase from Plasmodium falcipa-
rum, Ki inhibition constant, Pred predicted
aConcentration required to reduce parasite growth rate by 50%
b(−) and (+) denote chloroquine-susceptible and -resistant strains,
respectively
cMost active APAQs in the GSK data set
The QSAR predictions of Vd and fup for 9 are similar to
the values predicted for chloroquine (10.6 L/kg2 and 9%,
respectively), which is widely used to treat malaria, and thus
are not unrealistic pharmacokinetic parameters.
used by GSK [19]. Table 4 shows the experimental results
obtained in asynchronous cultures, expressed as the con-
centration required to reduce growth by 50% (XC50). With
the exception of the imidazolinones 12a and 12b, the new
compounds were more potent than the most active APAQs
in the GSK data set.
Biological results
With the exception of 12a, the novel analogs met the
minimum potency threshold (better than 2 μM) for parasite
proliferation called for in the Medicines for Malaria Venture
(MMV) guidelines at the time [37], with 11b satisfying the
ideal criteria with an XC50 below 100 nM. Similar antima-
larial potencies were obtained in replicate assays carried out
at UC Riverside and at Tres Cantos. The compounds were
also tested in the chloroquine-resistant Dd2 strain, which
contains an efux pump [38]. It was gratifying to see that
the ratio of XC50′s between the resistant and susceptible
strains—i.e., the resistance ratio (Table 4)—was lower for
all of the APAQs than the value of 15.5 reported for chlo-
roquine [39].
All aminoquinolones synthesized for this project were
assayed for ex vivo antimalarial activity in blood culture and
for inhibition of s-PfDHODH. Characterization in terms of
kill rate is very resource-intensive, as is determining whether
or not parasites can be rescued from growth inhibition by
transfection with Saccharamyces cerevisiae DHODH (ScD-
HODH) DNA (see the section below on “Inhibition of intact
PfDHODH”). These assays were, therefore, only carried out
for the two most active candidates for which sufcient mate-
rial was available—8 and 11b. In addition, historical data for
some of the APAQs in the PubChem dataset, including ScD-
HODH rescue and ex vivo activity, were made available by
the Tres Cantos Medicines Development Campus-Diseases
of the Developing World, GSK (Tres Cantos) [19].
Speed of killing time course assay
XC50s for cultured parasites
The parasite reduction ratio (PRR) is an indicator of how
rapidly parasites are killed by antimalarial treatment [40].
Preliminary assays carried out at Tres Cantos for 8 and 11b
yielded 48-h parasite in vitro reduction ratios (PPRs) of
approximately 3 log units, i.e., 99.9% of the parasites were
killed in the frst 48 h of exposure (1 lifecycle). This was
slightly better than pyrimethamine, which was included as
a positive control. Such rapid killing of the parasite is very
desirable for antimalarials [41]. Little or no lag time in the
Antimalarial activity in blood culture was determined as
previously described [35, 36] at the University of Califor-
nia, Riverside, under conditions essentially identical to those
2
The actual Vd for chloroquine is difcult to determine experimen-
tally, but is probably much higher [34].
1 3

Journal of Computer-Aided Molecular Design
onset of killing was evident for either compound, compared
to the 24–48 h time lag reported for triazolopyrimidine
inhibitors of PfDHODH [42].
bind differs substantially from species to species and
changes upon inhibitor binding [45]. These observations
suggest that the absence of inhibition by APAQs in the
in vitro assay could be due to the loss of the hydrophobic
N-terminal residues and may not refect the activity in liv-
ing parasites.
Inhibition of s‑PfDHODH in vitro
Identifcation of a compound’s mode of action is important
in drug discovery because it opens the door to structure-
based drug design and because it helps alert researchers to
undesirable of-target activities that might cause concern.
In some cases, it highlights opportunities for synergistic
inhibition of multiple targets by a single compound, i.e.,
for polypharmacology [43]. PfDHODH was the intended
target for this proof-of-principle project, so we wanted to
know whether the identifed lead series (in general) and
our new analogs (in particular) inhibited the enzyme in
intact parasites.
Inhibition of intact PfDHODH
Fortunately, there is a way to address this question, though
the method is somewhat indirect. P. falciparum parasites
engineered to express DHODH from the yeast Saccharamy-
ces cerevisiae (ScDHODH) use fumarate as an oxidizing
agent instead of ubiquinone, so their growth is not inhibited
by (i.e., they are “rescued from”) inhibitors that compete
with ubiquinone. Painter et al. used this technique to validate
PfDHODH as an antimalarial drug target [46].
The literature data set used to build our PfDHODH
inhibition models was based on in vitro assays carried out
using the truncated, soluble form of the enzyme introduced
by Baldwin et al. [10]. Patel et al. subsequently used it
to screen a library of 208,000 compounds and validated
the 38 “hits” obtained by testing for inhibition of para-
site growth in blood culture [8] The seven compounds we
designed and synthesized were poor inhibitors of s-PfD-
HODH with in vitro IC50 > 100 μM when assayed under
those conditions at the Tres Cantos laboratory.
Unbeknownst to the Simulations Plus and U.C. Riverside
groups, the Tres Cantos group had already tested several
2-aminoquinolone actives from the GSK antimalarial data
set for rescue by ScDHODH transfection. Their previously
unpublished work showed that parasites were rescued from
growth inhibition by 10 of 102 APAQs assayed, with some
additional compounds giving an equivocal result, i.e., par-
tial rescue. An additional sixteen 2-amino-4-quinolones hav-
ing diferent bridging groups were also assayed; three of
these were clearly subject to rescue, whereas three yielded
ambiguous results. Illustrative example structures are shown
in Fig. 7.
This result was somewhat unexpected, because of the
active quinolones in the QSAR model training set (com-
pounds P01 and P02 in Supplemental Table S1) and the
potency of HDQ mentioned above. The substituents at the
C2 position in the APAQs are longer and bulkier than are
the styryl and pentyl groups in P01 and P02, but they are
still much smaller than the 30- to 50-carbon polyisopre-
noid tail of the natural substrate, CoQ10.
The APAQs from which parasites were rescued by ScD-
HODH transfection did inhibit s-PfDHODH in vitro, but
with higher in vitro IC50′s (i.e., with lower afnity) than
expected given the phenotypic assay results—3.1 and
5.2 μM for the most potent examples. The most potent GSK
analog (CID 44,537,350) had an in vitro IC50 of 0.86 μM. It
bore a 4-(2-pyridyl)butylamino group in place of the 3-ami-
nopropyl amino group in our series.
The solubility of s-PfDHODH results from the removal
of the hydrophobic tail that anchors the enzyme to the
mitochondrial membrane, where its native cosubstrate
resides in vivo. A compound that inhibits the truncated,
soluble enzyme in vitro is likely to inhibit the full-length,
membrane-bound form. The converse—that an inhibitor
of the full-length enzyme will necessarily inhibit the trun-
cated one—is not necessarily true, especially for inhibitors
like the APAQs that resemble ubiquinone more than they
resemble dihydroorotate [16]. Atovaquone, for example,
inhibits the full-length enzyme from P. falciparum with
a Ki of 27 µM after solubilization in Triton X-100 [44],
whereas the corresponding value for the truncated enzyme
is reportedly> 500 µM when assayed under similar condi-
tions [45]. Binding to the s-RrDHODH from rat is much
tighter (Ki ~0.08 µM) and intermediate for the correspond-
ing human enzyme ((Ki ~ 2 µM). The organization of the
N-terminal “stump” where competitive CoQ inhibitors
The limitation of this test is that the failure of transfection
to rescue parasites from the other APAQs tested—includ-
ing 8 and 11b—does not imply that those compounds do
not inhibit the intact Plasmodium enzyme. It does imply,
however, that PfDHODH inhibition is not their exclusive
mode of antimalarial action. Basically, 8 and 11b start kill-
ing the parasite in culture immediately (see above), whereas
specifc PfDHODH inhibitors take 24–48 h to do so [47]. A
somewhat analogous situation exists for atovaquone, which
is much more potent in blocking respiratory oxidation of
ubiquinol than in inhibiting PfDHODH directly [46]. In
that case, the time frame and the end result—the parasite is
starved of the CoQ it needs to synthesize pyrimidine nucleo-
tides—are the same for both targets,only the target itself is
diferent.
1 3

Journal of Computer-Aided Molecular Design
Fig. 7 Active GSK APAQs
from which P. falciparum
grown in blood culture were
rescued by transfection with
ScDHODH
Fig. 8 Malarial methionyl
t-RNA synthetase (PfMRS)
inhibitors with antimalarial
activity
The bottom line is that APAQs as a class evidently do
inhibit PfDHODH ex vivo but that, at least for our most
potent analogs, some other, more rapid-onset mode of action
is responsible for their antimalarial activity.
blocking PfMRS was the compounds’ mode of action in the
intact parasite.
In fact, the rapid knockdown and absence of a lag time
in the PRR experiments argues against any kind of protein
synthesis inhibition being involved, at least for 8 and 11b [50].
Other suggested modes of action
The malarial methionyl t-RNA synthetase (PfMRS) has also
been suggested as a target. A patent for 2-aminoquinolones as
antibiotics was fled in 1999 [27, 48] and the mode of action
was characterized as being inhibition of bacterial MRS. How-
ever, the 26 compounds characterized as MRS inhibitors in
that patent were not among the actives in the Gamo et al. data
set [19], which argues against the idea that MRS might be
the target in malaria. Subsequent to the work described here,
two other APAQs characterized as bacterial MRS inhibitors
(REP3123 and REP8839 in Fig. 8) were shown to inhibit the
malarial enzyme in vitro and to be potent antimalarials ex vivo
[49].
Conclusions
The 2-(3-aminopropylamino)-4-quinolones generated by the
model-driven in silico design approach described here are
potent antimalarial lead drug candidates. The methodology
used to design compounds that target PfDHODH yielded
lead candidates with excellent activity against P. falcipa-
rum parasites as well as acceptable ADMET properties and
favorable PK profles. Compounds from the frst design
iteration were predicted to be more potent in blood culture
than structurally related literature compounds and their
experimentally determined potencies were consistent with
that expectation. It seems likely that further iterations and
in vivo characterization would be productive.
Those authors did not test for activity against PfDHODH,
however, nor did they present conclusive evidence that
1 3

Journal of Computer-Aided Molecular Design
Code availability ADMET Predictor® and GastroPlus® are commer-
cially available from Simulations Plus, Inc., Lancaster CA 93534 (https
://www.simulations-plus.com).
The results presented here do not pin down the primary
antimalarial mode of action for the APAQs, but do indicate
that it is probably neither PfDHODH nor protein synthesis.
Fortunately, the lack of cytotoxicity the lead series showed
against mammalian cells in culture suggests that inhibition
of the primary target, whatever it is, is not intrinsically
problematic with respect to human beings. If that is indeed
the case, having multiple activities should improve pros-
pects for further development iterations centered around
this class of chemistry, since hitting multiple targets is a
good way to slow development of resistance.
Compliance with ethical standards
Conflict of interest Robert Clark, Michael Lawless, and Walter Wol-
tosz are employees of Simulations Plus, Inc. and own stock and/or op-
tions in the company. The authors declare no other conficts of interest
or competing interests beyond those implicit in their employment by
their respective institutions.
Open Access This article is licensed under a Creative Commons Attri-
bution 4.0 International License, which permits use, sharing, adapta-
tion, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source,
provide a link to the Creative Commons licence, and indicate if changes
were made. The images or other third party material in this article are
included in the article’s Creative Commons licence, unless indicated
otherwise in a credit line to the material. If material is not included in
the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a
copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
This successful proof-of-principle study highlights
how useful in silico tools can be in de novo design and
for identifying lead compounds from phenotypic screens.
The example chosen illustrates how including predictive
PBPK simulations early in the workfow can reduce the
risk of relying too heavily on any single ADMET property
flter when selecting analogs for synthesis. Applied more
generally, the overall approach has the potential to reduce
the number of compounds that need to be synthesized to
get from discovery through optimization, to make animal
testing more efcient, and to reduce attrition in clinical
trials. Beyond being useful as a proof-of-principle, the
exercise provided prospective validation of several kinds
of in silico property predictions as well as an example
of the kind of unexpected of-target activity commonly
encountered in drug development projects—in this case,
a desirable one.
References
1. World Health Organization (2019) World malaria report 2019.
https://www.who.int/malaria/publications/world-malaria-repor
t-2019/report/en/. Accessed 7 Apr 2020
2. Wells TNC, van Huijsduijnen RH, Van Voorhis WC (2015)
Malaria medicines: a glass half full? Nat Rev Drug Discov
14:424–442
Acknowledgements The authors are indebted to the Medicines for
Malaria Venture for logistical support and to Jeremy Burrows and
Xavier Ding for orchestrating and facilitating the transfer of materials
to Tres Cantos for assay; to Richard Lloyd of GlaxoSmithKline for
his support and encouragement throughout the antimalarial project; to
Mike Melnick for his synthesis support; and to Kayla Andrews for her
careful reading of an earlier version of the manuscript and editorial
suggestions. We would also like to thank our reviewers for their careful
reading of the manuscript and thoughtful comments.
3. Cui L, Mharakurwa S, Ndiaye D, Rathod PK, Rosenthal PJ (2015)
Antimalarial drug resistance: literature review and activities
and fndings of the ICEMR network. Am J Trop Med Hyg 93(3
Suppl):57–68
4. Bhattacharjee AK, Geyer JA, Woodard CL, Kathcart AK, Nich-
ols DA, Prigge ST, Li Z, Mott BT, Waters NC (2004) A three-
dimensional in silico pharmacophore model for inhibition of Plas-
modium falciparum cyclin-dependent kinases and discovery of
diferent classes of novel Pfmrk specifc inhibitors. J Med Chem
47:5418–5426
5. Zhang L, Fourches D, Sedykh A, Zhu H, Golbraikh A, Ekins S,
Clark J, Connelly MC, Sigal M, Hodges D, Guiguemde A (2013)
Discovery of novel antimalarial compounds enabled by QSAR-
based virtual screening. J Chem Inf Model 53:475–492
6. Lawless MS, Waldman M, Fraczkiewicz R, Clark RD (2016)
Using cheminformatics in drug discovery. In: Nielsch U, Fuhr-
mann U, Jaroch S (eds) Handbook of experimental pharmacol-
ogy: new approaches to drug discovery. Springer, Switzerland,
pp 139–168
Author contributions The Simulations Plus and Cognigen groups did
the modeling, analog design, and pharmacokinetic simulations work
described herein and drafted the manuscript. The Kalexsyn group
helped refne the list of synthesis targets, then planned and carried out
the synthesis of the fnal set of cyclopentyl-1,1-dimethylene analogs.
The UC Riverside and Tres Cantos groups carried out XC50 determina-
tions in blood culture. The Tres Cantos GSK group also conducted the
kill rate and ScDHODH rescue assays. Synthesis and academic work
were underwritten by Simulations Plus, Inc.
7. Phillips MA, Rathod PK (2010) Plasmodium dihydroorotate dehy-
drogenase: a promising target for novel anti-malarial chemother-
apy. Infect Disord Drug Targets 10:226–239
Funding No external funding was used for the research described
herein.
8. Patel V, Booker M, Kramer M, Ross L, Celatka CA, Kennedy LM,
Dvorin JD, Duraisingh MT, Sliz P, Wirth DF, Clardy J (2008)
Identifcation and characterization of small molecule inhibitors
of Plasmodium falciparum dihydroorotate dehydrogenase. J Biol
Chem 283:35078–35085
Data availability Our supply of synthesized material is exhausted at
present, but the experimental procedures required to synthesize more
are provided. NMR spectra and any details not included in the supple-
mentary materials are available from the authors upon request.
9. Phillips MA, Gujjar R, Malmquist NA, White J, El Mazouni F,
Baldwin J, Rathod PK (2008) Triazolopyrimidine-based dihy-
droorotate dehydrogenase inhibitors with potent and selective
1 3

Journal of Computer-Aided Molecular Design
activity against the malaria parasite Plasmodium falciparum. J
Med Chem 51:3649–3653
chain of Plasmodium falciparum and its inhibitors. Bioorg Med
Chem Lett 19:972–975
10. Baldwin J, Michnof CH, Malmquist NA, White J, Roth MG,
Rathod PK, Phillips MA (2005) High-throughput screening for
potent and selective inhibitors of Plasmodium falciparum dihy-
droorotate dehydrogenase. J Biol Chem 280:21847–21853
11. Davies M, Heikkila T, McConkey GA, Fishwick CWG, Parsons
MR, Johnson AP (2009) Structure-based design, synthesis, and
characterization of inhibitors of human and Plasmodium falcipa-
rum dihydroorotate dehydrogenases. J Med Chem 52:2683–2693
12. Heikkilä T, Ramsey C, Davies M, Galtier C, Stead AM, Johnson
AP, Fishwick CW, Boa AN, McConkey GA (2007) Design and
synthesis of potent inhibitors of the malaria parasite dihydrooro-
tate dehydrogenase. J Med Chem 50:186–191
26. Ertl P, Schufenhauer A (2009) Estimation of synthetic accessibility
score of drug-like molecules based on molecular complexity and
fragment contributions. J Cheminform 1:8
27. Berge JM, Brown P, Elder JS, Forrest AK, Hamprecht DW, Jarvest
RL, McNair DJ, Sheppard RJ (2001) Quinolones used as MRS
inhibitors and bactericides. US Patent 6320051
28. GastroPlus (2020) Simulations Plus, Inc., Lancaster, CA. https://
www.simulations-plus.com
29. Simulations Plus, Inc., Press release (9/13/2011) https://www.irdir
ect.net/prviewer/release/id/2168302
30. Okubo T, Kumagi T, Nakamura T, Abe K, Amada Y, Ishizaka T,
Sun X-M, Sekiguchi Y, Sasako S, Shimizu T, Nagatsuka T (2011)
Aminopyrrolidine compound. US Patent US8044068B2
31. Clark RD (2018) Predicting mammalian metabolism and toxicity of
pesticides in silico. Pest Manage Sci 74:1992–2003
13. Gujjar R, Marwaha A, El Mazouni F, White J, White KL, Crea-
son S, Shackleford DM, Baldwin J, Charman WN, Buckner FS,
Charman S (2009) Identifcation of a metabolically stable triazo-
lopyrimidine-based dihydroorotate dehydrogenase inhibitor with
antimalarial activity in mice. J Med Chem 52:1864–1872
14. Gujjar R, El Mazouni F, White KL, White J, Creason S, Shackl-
eford DM, Deng X, Charman WN, Bathurst I, Burrows J, Floyd
DM (2011) Lead-optimization of aryl and aralkyl amine based tria-
zolopyrimidine inhibitors of Plasmodium falciparum dihydrooro-
tate dehydrogenase with anti-malarial activity in mice. J Med Chem
54:3935–3949
32. Clark RD, Liang W, Lee AC, Lawless MS, Fraczkiewicz R, Wald-
man M (2014) Using beta binomials to estimate classifcation uncer-
tainty for ensemble models. J Cheminform 6:34
33. Houston JB, Kenworthy KE (2000) In vitro-in vivo scaling of CYP
kinetic data not consistent with the classical Michaelis–Menten
model. Drug Metab Dispos 28:246–254
34. Looareesuwan S, White NJ, Chanthavanich PO, Edwards G, Nicholl
DD, Bunch C, Warrell DA (1986) Cardiovascular toxicity and dis-
tribution kinetics of intravenous chloroquine. Br J Clin Pharmacol
22:31–36
15. The Binding Database (2012). https://www.bindingdb.org
16. Hansen M, Le Nours J, Johansson EV, Antal T, Ullrich A, Löfer M,
Larsen S (2004) Inhibitor binding in a class 2 dihydroorotate dehy-
drogenase causes variations in the membrane-associated N-terminal
domain. Protein Sci 13:1031–1042
35. Daub ME, Prudhomme J, Le Roch K, Vanderwal CD (2015) Syn-
thesis and potent antimalarial activity of kalihinol B. J Am Chem
Soc 137:4912–4915
17. Cheng YC, Prusof WH (1973) Relationship between the inhibition
constant (Ki) and the concentration of inhibitor which causes 50 per
cent inhibition (I50) of an enzymatic reaction. Biochem Pharmacol
22:3099–3108
36. Prudhomme J, McDaniel E, Ponts N, Bertani S, Fenical W, Jensen
P, Le Roch K (2008) Marine actinomycetes: a new source of com-
pounds against the human malaria parasite. PLoS ONE 3:e2335
37. Ray S, Madrid PB, Catz P, LeValley SE, Furniss MJ, Rausch LL,
Guy RK, DeRisi JL, Iyer LV, Green CE, Mirsalis JC (2010) Devel-
opment of a new generation of 4-aminoquinoline antimalarial com-
pounds using predictive pharmacokinetic and toxicology models. J
Med Chem 53:3685–3695
18. ADMET Predictor (2020) Simulations Plus, Inc., Lancaster, CA.
https://www.simulations-plus.com
19. Gamo FJ, Sanz LM, Vidal J, De Cozar C, Alvarez E, Lavandera JL,
Vanderwall DE, Green DV, Kumar V, Hasan S, Brown JR (2010)
Thousands of chemical starting points for antimalarial lead identi-
fcation. Nature 465:305–310
38. Martin RE, Marchetti RV, Cowan AI, Howitt SM, Bröer KK (2009)
Chloroquine transport via the malaria parasite’s chloroquine resist-
ance transporter. Science 325:1680–1682
20. Ekins S, Williams AJ (2010) When pharmaceutical companies
publish large datasets: an abundance of riches or fool’s gold? Drug
Discov Today 15:812–815
39. Kumar R, Dufy S, Avery VM, Davis RA (2017) Synthesis of anti-
malarial amide analogues based on the plant serrulatane diterpenoid
3,7,8-trihydroxyserrulat-14-en-19-oic acid. Biorg Med Chem Lett
27:4091–4095
21. MedChem Studio (2020) Simulations Plus, Inc., Lancaster, CA.
https://www.simulations-plus.com; this functionality has since been
subsumed into the MedChem Studio Module of ADMET Predictor
22. Ploufe D, Brinker A, McNamara C, Henson K, Kato N, Kuhen
K, Nagle A, Adrián F, Matzen JT, Anderson P, Nam TG (2008) In
silico activity profling reveals the mechanism of action of antima-
larials discovered in a high-throughput screen. Proc Natl Acad Sci
105:9059–9064
40. White NJ (2011) The parasite clearance curve. Malar J 10:278
41. Burrows JN, Duparc S, Gutteridge WE, van Huijsduijnen RH,
Kaszubska W, Macintyre F, Mazzuri S, Möhrle JJ, Wells TN (2017)
New developments in anti-malarial target candidate and product pro-
fles. Malar J 16:26
42. Phillips MA, Lotharius J, Marsh K, White J, Dayan A, White KL,
Njoroge JW, El Mazouni F, Lao Y, Kokkonda S, Tomchick DR
(2015) A long-duration dihydroorotate dehydrogenase inhibitor
(DSM265) for prevention and treatment of malaria. Sci Transl Med
7:296ra111
23. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ (2001) Experi-
mental and computational approaches to estimate solubility and per-
meability in drug discovery and development settings. Adv Drug
Deliv Rev 46:3–26
43. Chaudhari R, Tan Z, Zhang S (2017) Overview of drug polyphar-
macology and multitargeted molecular design. In: Wilson AGE
(ed) Comprehensive medicinal chemistry III. Elsevier, Oxford, pp
259–275
24. Coteron JM, Marco M, Esquivias J, Deng X, White KL, White J,
Koltun M, El Mazouni F, Kokkonda S, Katneni K, Bhamidipati R
(2011) Structure-guided lead optimization of triazolopyrimidine-
ring substituents identifes potent Plasmodium falciparum dihy-
droorotate dehydrogenase inhibitors with clinical candidate poten-
tial. J Med Chem 54:5540–5561
44. Seymour KK, Lyons SD, Phillips L, Rieckmann KH, Christopher-
son RI (1994) Cytotoxic efects of inhibitors of de novo pyrimi-
dine biosynthesis upon Plasmodium falciparum. Biochemistry
33:5268–5274
25. Dong CK, Patel V, Yang JC, Dvorin JD, Duraisingh MT, Clardy J,
Wirth DF (2009) Type II NADH dehydrogenase of the respiratory
1 3

Journal of Computer-Aided Molecular Design
45. Hurt DE, Widom J, Clardy J (2006) Structure of Plasmodium falci-
parum dihydroorotate dehydrogenase with a bound inhibitor. Acta
Cryst D62:312–323
49. Hussain T, Yogavel M, Sharma A (2015) Inhibition of protein
synthesis and malaria parasite development by drug targeting of
methionyl-tRNA synthetases. Antimicrob Agents Chemother
59:1856–1867
46. Painter HJ, Morrisey JM, Mather MW, Vaidya AB (2007) Specifc
role of mitochondrial electron transport in blood-stage Plasmodium
falciparum. Nature 446:88–91
50. Baragaña B, Hallyburton I, Lee MC, Norcross NR, Grimaldi R,
Otto TD, Proto WR, Blagborough AM, Meister S, Wirjanata G,
Ruecker A (2015) A novel multiple-stage antimalarial agent that
inhibits protein synthesis. Nature 522:315–320
47. Phillips MA, Gujjar R, Malmquist NA, White J, El Mazouni F, Bald-
win J, Rathod PK (2016) A triazolopyrimidine-based dihydrooro-
tate dehydrogenase inhibitor with improved drug-like properties for
treatment and prevention of malaria. ACS Infect Dis 2:945–957
48. Jarvest RL, Berge JM, Berry V, Boyd HF, Brown MJ, Elder JS,
Forrest AK, Fosberry AP, Gentry DR, Hibbs MJ, Jaworski DD,
O’Hanlon PJ, Pope AJ, Rittenhouse S, Sheppard RJ, Slater-Radosti
C, Worby A (2002) Nanomolar Inhibitors of Staphylococcus aureus
methionyl tRNA synthetase with Potent antibacterial activity against
gram-positive pathogens. J Med Chem 45:1959–1962
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional afliations.
1 3