3D-QSAR assisted identification of FABP4 inhibitors: An effective scaffold hopping analysis/QSAR evaluation

Article abstract of DOI:10.1016/j.bioorg.2018.11.045

Following on the recent publication of pharmacologically relevant effects, small molecule inhibitors of adipocyte fatty-acid binding protein 4 (FABP4) have attracted high interest. FABP4 is mainly expressed in macrophages and adipose tissue, where it regulates fatty acid storage and lipolysis, being also an important mediator of inflammation. In this regard, FABP4 recently demonstrated an interesting molecular target for the treatment of type 2 diabetes, other metabolic diseases and some type of cancers. In the past years, hundreds of effective FABP4 inhibitors have been synthesized. In this paper, a quantitative structure-activity relationship (QSAR) model has been produced, in order to predict the bioactivity of FABP4 inhibitors. The methodology has been combined with a scaffold-hopping approach, allowing to identify three new molecules that act as effective inhibitors of this protein. These molecules, synthesized and tested for their FABP4 inhibitor activity, showed IC₅₀ values between 3.70 and 5.59 μM, with a high level of agreement with the predicted values.

Full text of DOI:10.1016/j.bioorg.2018.11.045

Accepted Manuscript
3D-QSAR assisted identification of FABP4 inhibitors: an effective scaffold
hopping analysis/QSAR evaluation
Giuseppe Floresta, Agostino Cilibrizzi, Vincenzo Abbate, Ambra Spampinato,
Chiara Zagni, Antonio Rescifina
PII:
S0045-2068(18)31127-1
DOI:
https://doi.org/10.1016/j.bioorg.2018.11.045
YBIOO 2653
Reference:
Bioorganic Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
4 October 2018
20 November 2018
24 November 2018
Please cite this article as: G. Floresta, A. Cilibrizzi, V. Abbate, A. Spampinato, C. Zagni, A. Rescifina, 3D-QSAR
assisted identification of FABP4 inhibitors: an effective scaffold hopping analysis/QSAR evaluation, Bioorganic
Chemistry (2018), doi: https://doi.org/10.1016/j.bioorg.2018.11.045
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

3D-QSAR assisted identification of FABP4 inhibitors: an effective scaffold hopping
analysis/QSAR evaluation.
Giuseppe Florestaa,b,c,*, Agostino Cilibrizzic,d, Vincenzo Abbated, Ambra Spampinatoa, Chiara Zagnia,
Antonio Rescifinaa,*
a Department of Drug Sciences, University of Catania, V.le A. Doria 6, 95125 Catania, Italy
b Department of Chemical Sciences, University of Catania, V.le A. Doria, 95125 Catania, Italy
c
Institute of Pharmaceutical Science, King’s College London, Stamford Street, London SE1 9NH,
UK
d
King’s Forensics, School of Population Health & Environmental Sciences, King’s College
London, Franklin-Wilkins Building, 150 Stamford Street, London SE1 9NH
Corresponding authors
E-mail address: giuseppe.floresta@unict.it (G. Floresta) and arescifina@unict.it (A. Rescifina)
Graphical Abstract
1

Highlights
 A 3D-QSAR model was produced for the FABP4 inhibitors
 A scaffold hopping suggested 3,000 potentially active FABP4 inhibitors
 Three synthesized molecules have shown an IC50 between 3.70 and 5.59 M
Keywords: FABP4 inhibitors, aP2, A-FABP, 3D-QSAR, scaffold-hopping, BMS309403 analogs,
thiazole, triazole, Forge and Spark software.
2

Abstract
Following on the recent publication of pharmacologically relevant effects, small molecule inhibitors
of adipocyte fatty-acid binding protein 4 (FABP4) have attracted high interest. FABP4 is mainly
expressed in macrophages and adipose tissue, where it regulates fatty acid storage and lipolysis,
being also an important mediator of inflammation. In this regard, FABP4 recently demonstrated an
interesting molecular target for the treatment of type 2 diabetes, other metabolic diseases and some
type of cancers. In the past years, hundreds of effective FABP4 inhibitors have been synthesized. In
this paper, a quantitative structure-activity relationship (QSAR) model has been produced, in order
to predict the bioactivity of FABP4 inhibitors. The methodology has been combined with a
scaffold-hopping approach, allowing to identify three new molecules that act as effective inhibitors
of this protein. These molecules, synthesized and tested for their FABP4 inhibitor activity, showed
IC50 values between 3.70 and 5.59 M, with a high level of agreement with the predicted values.
3

1. Introduction
Fatty acids (FAs) are a class of carboxylic acids that carries out different vital functions in the
organism [1]. Recently, it was proved that chronically elevated plasma fatty acids lead to specific
physiological disorders [2]. For instance, FAs elevated levels are connected with type 2 diabetes
[3], obesity [4] and atherosclerosis [5]. Due to the poor solubility in water of FAs, their trafficking
requires a cluster of specific carrier proteins, such as albumin, lipocalins and fatty acid-binding
proteins (FABPs), that facilitates their transport increasing their water solubility [6]. The adipocyte
FABP (A-FABP, aP2 or FABP4) is highly expressed in adipocytes and it is regulated by
peroxisome-proliferator-activated receptor-c agonists, as well as by the levels of insulin and fatty
acids [7]. Studies in FABP4 knockout mice have shown that FABP4 has a key role in many aspects
of the metabolic syndrome [8, 9], showing that the lack of this protein partially prevents the
advancement of insulin resistance and obesity in mice.
Therefore, pharmacological agents that inhibit FABP4 function can mimic the phenotype of
FABP4-deficient mice and small molecules that inhibit FABP4 might be useful candidates for the
treatment of metabolic syndromes. FABPs play also an important role in carcinogenesis [10].
Modified FABPs expression were described for prostate, bladder, renal cell carcinoma and other
types of cancer cells [11-13], although FABPs biological function in cancer remains mostly unclear
[14].
Recently, a variety of effective FABP4 inhibitors have been developed [15], but unfortunately, none
of them is currently in the clinical research phases, and only BMS309403 (Figure 1) has been
systematically studied in both in vitro and in vivo animal models [16-18]. In Table 1 are reported
the different classes of FABP4 inhibitors [15], in Figure 1 and Table 2 some representative
examples and their biological evaluations, respectively.
Table 1
Various chemical classes of FABP4 inhibitors.
Chemical Classification
Chemical Structure
Pyrazole derivatives
4

Oxazole derivatives
Imidazole derivatives
Indole derivatives
Benzimidazole derivatives
Thiophene and thiazoles derivatives
Pyrimidine, bicyclic pyridine and quinoxaline derivatives
Urea and carbamoyl derivatives
5

Fig. 1. Structures of different FABP4 selective inhibitors.
Table 2
Experimental biological activity of FABP4 inhibitors.
Compound
Reference
IC50 or Ki (M)
BMS309403
0.71a
0.67b
0.45a
1.00a
0.001b
4.00a
1.30a
0.033b
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
HTS01037
1
2
3
4
5
6
a Experimental IC50; b Experimental Ki
6

Computer-aided drug design shows a promising and effective tool for the identification of FABP4
inhibitors. Indeed, different structure-based computational approaches have been already performed
in the context of FABP4 research, for instance virtual screening studies have been reported with
different libraries of compounds, leading to important biological results [27, 28]. In line with our
recent interest in the development of QSAR models and related applications, in order to identify
novel hit compounds, herein we report the development of a 3D-QSAR approach for FABP4 [29-
35]. Moreover, the 3D-QSAR model was combined with a scaffold-hopping analysis, allowing the
design of new potent molecules able to interact with the binding site and inhibit the FABP4. The
3D-QSAR model for the determination of FABP4 endpoints was developed using the software
Forge, while Spark was adopted to produce a scaffold-hopping analysis and generate a virtual
library of FABP4 inhibitors. This 3D-QSAR model was used to elucidate SAR data and predict
FABP4 endpoints for new molecules. Three of the ligands suggested by the scaffold-hopping
analysis were synthesized and tested in vitro yielding IC50 values between 3.70 and 5.59 M.
2. Results and discussion
2.1. Statistical analysis
The regression method used in Forge was PLS (SIMPLS algorithm) [36-40]. As presented in the
Data in Brief related article, two different alignment methods were initially tested (i.e., structure-
based and ligand-based). The predictive ability of the generated QSAR model was confirmed by
several statistical tests. The cross-validation regression coefficient (q2) was calculated based on the
PRESS (Prediction error sum of squares) and SSY (Sum of squares of deviation of the experimental
values from their mean):
Yexp = experimental activity of training set compound
Ypred = predicted activity of training set compound
Ymean = mean values of the activity of training set compound
The statistical results of both models are reported in Table 3 and Figure 2.
7

Table 3
3D-QSAR model statistics of the different alignment protocols.
Alignment protocols
Leave-one-out q2
Training set r2
0.90
3D-QSAR Based on Docked Poses (structure based)
3D-QSAR Based on Forge Alignments (ligand based)
0.38
0.64
0.92
Fig. 2. 3D-QSAR model statistics: (up) 3D-QSAR based on docked poses (structure based); (down) 3D-QSAR based
on Forge alignments (ligand based).
As demonstrated above (Fig. 2, Table 3), the ligand-based approach results to be more reliable (r2 =
0.92, q2 = 0.64) than the structure-based alignment, which has an increased noise (r2 = 0.90, q2 =
0.38). At this point, the ligand-based 3D-QSAR align model was further validated with a set of
external compounds (i.e. test set). Indeed, a more reliable estimation of robustness comes from
separate training and test sets. Out of 120 molecules, we randomly choose 96 molecules (covering
the whole range of activities of the compounds) as a training set to build the model, while the
remaining 24 compounds served as a test set to evaluate the model.
The statistical reliability of this model was also validated by the determination of the r2test, using
the following equation:
Ypredtest = predicted activity of test set compound by QSAR equation
Ytest = experimental activity of test set compound
8

Ymean = mean values of the activity of training set compound
The results are presented in Figures 3 and 4. The 11-components model shows both good predictive
and descriptive capability as it is shown by the good r2 (0.99) and q2 (0.69) [41] values for the
training and the cross-validated training sets. The plot of experimental vs. predicted activity for the
compounds, in both the training set and the cross-validated training set (q2 = 0.69), shows a
reasonable distribution of the values. The plot of experimental vs. predicted activity for the
compounds in the test set is still reasonably good with only few outliers and a good cross-validated
r2 of 0.73.
Fig. 3. The 11-component 3D-QSAR model – Experimental vs. Predicted activity of the compounds in the training set.
Fig. 4. The 11-component 3D-QSAR model – Experimental vs. Predicted activity of the compounds in the test set.
The graphical interpretation of the model is shown in Figure 5, where the 3D-QSAR model results
are superposed to the structure of the BMS309403. In this model representation, the larger the
points are and the stronger is the correlation between the electrostatic/steric fields in that position.
This 3D-QSAR model is clearly described by both steric and electrostatic effects and well describes
the behavior of the different classes of FABP4 inhibitors inside the binding pocket of FABP4. To
reveal the key features of the series under investigation, a structure-activity relationship (SAR)
study was also performed through activity-atlas (AA) visualization software [42]. Figure 6
illustrates the results of AA calculations, where the AA map is superimposed to BMS309403 lead
9

compound. The different colors on the map highlight electrostatic, hydrophobic and shape features.
A more positive electrostatic field increases the receptor-affinity in the red area, whereas in the blue
region a more negative electrostatic field increases the affinity. Green and violet areas represent the
steric and bulk/hydrophobic interactions. A steric/bulk interaction improves the binding affinity in
the green area. On the other hand, in the violet area, a steric/bulk interaction decreases the affinity.
Fig. 5. Electrostatic and Steric coefficients for the FABP4 model superposed to the BMS309403. A) Steric coefficients.
B) Electrostatic coefficients. C) Contributions to the predicted affinity for BMS309403.
Fig. 6. The AA FABP4 model map is superimposed to the BMS309403. Molecular insight of SAR mechanism models,
revealing the different lead optimization sites of active compounds. Red color shows positive field region controlling
the activity and blue color the negative ones. Green color shows favorable shape/hydrophobic regions and purple color
the unfavorable ones.
2.2. Scaffold-hopping analysis
10

To design novel hit compounds with FABP4 inhibitory activity, by using the above-reported
information from the QSAR model, we decided to use a bioisostere and fragment replacement
software tool (i.e. Spark; Cresset group) and produce a scaffold hopping analysis to generate a
virtual library of FABP4 ligands [43]. As reported in Figure 7, Different portions of the
BMS309403 were studied in the bioisosteric replacement. Once the virtual library was created, the
newly designed molecules were scored assuming that if their calculated/aligned fields are highly
similar to that of the original compounds present in the 3D-QSAR model, the resulting biological
activity would have been comparable [44, 45]. The bioisosteric replacement was performed through
the same 178,558 fragments for each part, which derive from ChEMBL and Zinc databases [46, 47]
(see the Data in Brief related article). Five hundred compounds were generated for each substitution
producing 3,000 hits in total, which were then scored in the 3D-QSAR model.
The first top-scoring compounds (best predicted pIC50 values) resulting from the bioisosteric
replacement of series 1–6 are reported in Figure 8, while the full set of compounds are reported as
SMILES in the Data in Brief related article. Overall, the results indicate that the bioisosteric
replacement and the following 3D-QSAR model evaluation generate compounds with a suitable
chemical structure for the inhibition of FABP4.
Fig. 7. Bioisosteric replacement of the lead compound BMS309403.
11

Fig. 8. First top-scored compounds derived from series 1–6, with regards to FABP4 predicted pIC50.
The results of the scaffold-hopping analysis were used for the design of a novel series of
BMS309403-based derivatives. In particular, we focused our attention on the results of the series 3
and 4. In series 3, the most potent compound was predicted as a thiazole derivate of the
BMS309403, where the central pyrazole is substituted by a thiazole. In series 4, the most potent
predicted compound was the term with a triazole as the aromatic linker, which can be easily
introduced through a click chemistry reaction. Using these information, we designed the new series
of compounds AST_1–3 (Fig. 9). The molecules were imported into Forge and aligned/evaluated
with the 3D-QSAR model. Interestingly, AST_1–3 were predicted to have a pIC50 of 5.4, 5.9 and
5.6, respectively, and demonstrated an IC50 of 3.98, 1.25 and 2.51 M. To prove the effective
predicting capabilities of our model, we decide to set up a synthetic route that affords AST_1–3.
12

Fig. 9. Structures of AST_1–3.
2.3. Synthesis
The synthesis (Scheme 1) starts from commercially available aldehydes (i.e. benzaldehyde,
naphthaldehyde, and 4-chloro benzaldehyde) and the Grignard reagent generated in situ by benzyl
bromide and magnesium. The alcohol 10 was firstly oxidized with PCC (Pyridinium
chlorochromate) to give compounds 11 and then monohalogenated, at the -position, with bromine
under acidic conditions (compounds 12). The 2‐ aminobenzene‐ 1‐ carbothioamide 14 was
generated from the respective amide 13 and the Lawesson’s reagent. Subsequently, the thioamide
was reacted with intermediate 12 to give the thiazole derivatives 15. The butynoic acid 17 was
obtained by oxidation of 3-butyn-1-ol 16. Lastly, the treatment of 15 with t-BuONO (tert-Butyl
nitrite) and azidotrimethylsilane afforded the azide intermediate and the final analogs were
assembled by using 17 via azide-alkyne Huisgen cycloaddition (i.e. copper (I) catalyzed variant,
where organic azides and terminal alkynes are united to afford 1,4-regioisomers of 1,2,3-triazoles as
single products).
13

Scheme 1. Synthetic scheme for the molecules AST_1–3.
2.4. FABP4 inhibition evaluation
FABP4 inhibitory activity was determined with a commercially available screening assay kit, by
measuring the decreasing in fluorescence of a detection reagent (which is supplied in the kit) when
displaced by an inhibitor of FABP4. Specifically, the detection reagent exhibits increased
fluorescence when bound to FABP4. Binding of the detection reagent is monitored by exciting at
370 nm and measuring the emission at 470 nm. Therefore, any powerful inhibitor of the protein,
which binds to the same binding pocket, can displace the detection reagent, thereby reducing the
fluorescence read-out. Inhibition activity data are expressed as IC50 (µM) and results obtained with
AST_1–3 are reported in Table 4. Arachidonic acid, a known powerful ligand of FABP4, was used
as a positive control and revealed an IC50 of 3.06 µM. Our set of compounds, AST_1–3, showed
IC50 values of 5.59, 3.70, and 4.31, respectively.
14

Table 4
Measured IC50 values for Arachidonic acid and AST_1–3 in FABP4.
IC50 (µM)
3.06 ±0.22
5.59 ±0.79
3.70 ±0.34
4.31 ±0.67
Compounds
Arachidonic acid
AST_1
AST_2
AST_3
3. Conclusion
FABP4 provides an attractive therapeutic and diagnostic target for a variety of diseases as
demonstrated in a variety of animal models [48-51]. FABP4 inhibitors could also be effective for
the treatment of cancer patients via the inhibition or reduction of early-stage tumors and metastasis.
Similarly, they show effective as biomarkers for the diagnosis of different types of cancers [52-58].
In addition, FABP inhibitors could also be used for the treatment of neurological disorders [59-62].
In this study, a 3D-QSAR model for known FABP4 inhibitors was produced and used to study a
large library of molecules as possible inhibitors of FABP4, by means of scaffold-hopping analysis.
The efficient predictive capabilities of the QSAR model was also proved through the synthesis of
three new compounds as FABP4 inhibitors, using the information obtained from the QSAR
equation. FABP4 binding affinities of the newly designed compounds (AST_1–3) were measured
by means of an established displacement assay, showing that they are FABP4 inhibitors in the low
micromolar range (i.e. IC50 values between 3.70 and 5.59 M). Moreover, a large number of
different molecules were also theorized to be as effective as the three synthesized and can be a
useful starting point for the design of others FABP4 inhibitors. Our results widely demonstrate that
simple modifications performed on different sites of the gold standard BMS309403 can retain or
even improve FABP4 inhibitory activity of the analogs. These findings, of course, should be
validated experimentally but are considered a reasonable starting point for the researchers in the
field. Surely, these findings will lead our future research in the field of the development of FABP4
inhibitors. Considering this transporter as a very promising target, the ultimate goal is to finally
bring optimized molecules into clinical research phase.
4. Experimental Section
4.1. Molecular modeling and biological data
15

The 120 structures used to build and evaluate the 3D-QSAR model were chosen among the
structures previously published [19, 21, 22, 24, 25, 63, 64]. All molecules possess an excellent
selectivity toward the FABP4 and the range of activities is broad enough to be able to build an
efficient model. The 2D chemical structures were built by Marvin Sketch, and all structures were
subjected to molecular mechanics energy minimization using the MMFF94 force field present in the
same software [65]. Once obtained the 3D structures of all compounds the geometry was also
optimized at the semi-empirical level of theory using the PM3 Hamiltonian [66], as implemented in
MOPAC 2016 package [67]. The protonation states of the molecules were calculated assuming a
pH of 7.0. Forge 10.4.2 (Cresset BioMolecular Discovery Ltd., https://www.cresset-group.com/)
was used for the building of the 3D-QSAR model. Spark 10.4.0 was used for the scaffold-hopping
analysis (Cresset BioMolecular Discovery Ltd., https://www.cresset-group.com/). Forge’s
parameters used for conformation hunt/alignment and to build/validate the QSAR model are
reported in the Data in Brief related article.
4.2. Chemistry
All chemicals were purchased from Merk, Sigma Aldrich, and Acros Organics and were reagent
grade or better. Solvents were purchased from Fisher Scientific, Sigma Aldrich, and VWR. Silica
gel for column chromatography was purchased from Merck. Samples were dried in a vacuum oven
(Gallenkamp, rated to ≤250 °C) connected to a vacuum pump (BOC-Edwards.). Pre-coated
aluminum sheets (silica gel 60 F254, Merck) were used for thin-layer chromatography (TLC) and
1
spots were visualized under UV light. H NMR and 13C NMR spectra were recorded on Varian
UNITY Inova spectrometer using CDCl3 or DMSO-d6 as a solvent and tetramethylsilane (TMS) as
1
an internal standard, at 200 or 500 MHz for H NMR and 125 MHz for 13C NMR. 13C spectra were
1H decoupled and multiplicities were determined by the APT pulse sequence. Chemical shift (δ)
values are given in ppm. All of the NMR experiments were analyzed using the General NMR
Analysis Toolbox (GNAT) [68]. Mass spectra were run at King’s College London (UK) on a
Thermofisher LCQ DECA XP ion trap mass spectrometer or a Waters - Micromass ZQ - Single
quadrupole mass spectrometer. HPLC analyses were performed on Agilent 1100 series HPLC with
quadrupole pump, vacuum degasser, variable wavelength detector (set to 254 or 281 nm), using a
gradient water/acetonitrile as reported in the supporting information. Elemental analyses for C, H,
N, and S were performed on a Carlo Erba Elemental Analyzer Mod. 1108 apparatus.
4.2.1 General procedure for the synthesis of alcohols 10
16

A solution of benzyl bromide 8 (0.02 mol) in Et2O (50 mL) was dropwise added to a stirred
suspension of Mg (0.025 mol) over a period of 2 h at room temperature. The mixture was stirred for
3 additional hours more at room temperature to produce benzyl Grignard 9 and cooled to –40 °C. A
solution of aldehyde 7 (0.01mol) in Et2O (50 mL) was added over a period of 1 h. The mixture was
stirred for 2 h at –40 °C, gradually brought to room temperature and stirred for 3 h more. After
addition of saturated aqueous NH4Cl (10 mL), the mixture was extracted with EtOAc. The
combined organic extracts were washed with water, brine and dried over Na2SO4. Solvent removal
under reduced pressure and column chromatography of the residue (silica gel, 93/7,
cyclohexane/EtOAc) afforded the pure alcohol 10. Yields 92–99%.
4.2.2 General procedure for the synthesis of ketones 11
A solution of 10 (8.0 mmol) was added to a stirred suspension of pyridinium chlorochromate (12
mmol) in CH2Cl2 (60 mL). The mixture was stirred at room temperature for nearly 3 h until the
disappearance of the starting material (monitored by TLC), diluted with diethyl ether (50 mL) and
filtered through a column of celite. Solvent removal of the filtrate under reduced pressure and
column chromatography of the residue (silica gel, 95/5, cyclohexane/EtOAc) afforded pure 11.
Yields 95–99%.
4.2.3 General procedure for the synthesis of -bromoketone 12
Bromine (3.8 mmol) was added dropwise to a solution of 11 (2.5 mmol) in freshly distilled CHCl3
(10 mL). The mixture was heated under reflux until conversion was complete (monitored by TLC).
After cooling, the solvent was evaporated under reduced pressure to provide a residue that was
purified via column chromatography (silica gel, cyclohexane) to yield pure -bromoketone 12.
Yields 70–85%.
4.2.4 General procedure for the synthesis of 2-Aminothiobenzamide 14
Lawesson’s reagent (8.0 mmol) was added to a solution of 13 (14.7 mmol) in THF (75 mL). The
mixture was stirred under N2 at room temperature for 24h. The solvent was evaporated, and the
residue partitioned between EtOAc (50 mL), and 1 N HCl (30 mL). Aqueous sat. NaHCO3 was
added to the aqueous layer until pH= 8–9. This basic solution was extracted with EtOAc (30 mL).
The combined organic layers were dried (Na2SO4) and filtered. The solvent was evaporated and the
resulting solid residue was recrystallized from toluene to afford pure 14 as a yellow solid; yield
72%.
17

4.2.5 General procedure for the synthesis of thiazole 15
The 2-aminothiobenzamide 14 (1.5 mmol) was added to a solution of -bromoketone 12 (1.0
mmol) in methanol. The mixture was refluxed overnight. After cooling, the solvent was evaporated
under reduced pressure to provide a residue that was purified by column chromatography (silica gel,
99/1, cyclohexane/EtOAc) to yield pure thiazoles 15. 15_1: yield 75%, yellow solid, m.p.: 168–170
1
°C. H NMR (200 MHz, Chloroform-d) δ 6.22 (bs, 2H), 6.58 – 6.88 (m, 2H), 7.04–7.47 (m, 9H),
7.57 (m, 3H). 13C NMR (50 MHz, chloroform-d) δ 113.63, 116.77, 116.82, 127.60, 127.68, 127.92,
128.16, 128.55, 128.63, 128.87, 129.58, 129.83, 130.59, 145.64, 149.16, 166.96. 15_2: yield 76%,
yellow solid, m.p.: 178–180 °C. 1H NMR (200 MHz, chloroform-d) δ 6.42–6.85 (m, 2H), 7.17–7.54
13
(m, 9H), 7.54–7.89 (m, 6H), 7.99–8.22 (m, 1H), 8.68 (s, 1H). C NMR (50 MHz, chloroform-d) δ
110.92, 115.13, 126.18, 126.27, 126.89, 127.69, 127.79, 128.19, 128.34, 128.78, 129.26, 129.72,
130.87, 131.32, 131.94, 132.56, 132.86, 133.37, 147.61, 148.92.
15_3: yield 81%, yellow solid, m.p.: 172–173 °C.
4.2.6 General procedure for the synthesis of 3-butynoic acid 17
A solution of 21 g (0.3 mol) of 3-butyn-1-ol 16 in acetone (300 mL) was added dropwise to a stirred
solution of chromium trioxide (60 g, 0.6mol) in 750 mL of 10 N H2S04 at 0 °C. The reaction
mixture was allowed to warm to room temperature and the stirring was continued for 1.5 h. The
liquid was decanted into a mixture of EtOAc and H2O. The aqueous layer was extracted five times
with EtOAc. The extracts were combined, washed twice with saturated NaCl solution, dried
(Na2SO4) and evaporated. The resulting solid residue was recrystallized from hexane to afford pure
17 as a white solid, yield 93%.
4.2.7 General procedure for the synthesis of final compounds AST_1–3
The suitable thiazole 15 (0.54 mmol) was dissolved in THF (2 mL) in a 25 mL round-bottomed
flask and cooled to 0 °C in an ice-bath. t-BuONO (0.81 mmol) and TMSN3 (0.65 mmol) were
sequentially added to this mixture under stirring. The resulting solution was reacted at room
temperature for 2 h. 3-Butinoic acid 17 (0.54 mmol), an aq. solution (0.2 mL) of CuSO4 (0.027
mmol), sodium ascorbate (0.108 mmol) were then added and the reaction was stirred overnight at
room temperature. The crude mixture was concentrated under vacuum and the product was then
precipitated by addition of water to give the final compounds AST_1–3. AST_1: white solid, m.p.:
1
238–240 °C. H NMR (500 MHz, DMSO-d6) δ 3.62 (s, 2H), 7.11–7.47 (m, 10H), 7.56 (d, J = 7.5
Hz, 1H), 7.73 (m, 2H), 8.08–8.47 (m, 2H). Anal. Calcd. for C25H18N4O2S: C, 68.48; H, 4.14; N,
12.78; S, 7.31; found C, 68.11; H, 4.10; N, 12.89; S, 7.22. AST_2: white solid, m.p.: 241–243 °C.
18

1H NMR (500 MHz, DMSO-d6) δ 3.57 (s, 2H), 7.28 (d, J = 7.9 Hz, 2H), 7.36 (m, 3H), 7.43–7.52
(m, 2H), 7.55 (d, J = 7.9 Hz, 1H), 7.64–7.74 (m, 1H), 7.74–7.83 (m, 3H), 7.83–7.91 (m, 1H), 7.95
(s, 1H), 8.26 (d, J = 6.7 Hz, 2H). Anal. Calcd. for C29H20N4O2S: C, 71.29; H, 4.13; N, 11.47; S,
1
6.56; found C, 71.54; H, 4.18; N, 11.38; S, 6.33. AST_3: white solid, m.p.: 241–242 °C. H NMR
(500 MHz, chloroform-d) δ 3.98 (s, 2H), 7.13 (t, J = 7.8 Hz, 1H), 7.17–7.24 (m, 2H), 7.25–7.37 (m,
5H), 7.44 (t, J = 1.9 Hz, 1H), 7.50 (dd, J = 7.8, 1.3 Hz, 1H), 7.59 (td, J = 7.7, 1.5 Hz, 1H), 7.66 (td,
J = 7.7, 1.5 Hz, 1H), 7.89 (s, 1H), 8.15 (dd, J = 7.7, 1.5 Hz, 1H). 13C NMR (50 MHz, chloroform-d)
δ 29.68, 126.94, 127.93, 128.11, 128.72, 128.86, 128.96, 129.35, 129.53, 130.35, 130.70, 134.07,
136.05, 160.12. Anal. Calcd. for C25H17ClN4O2S: C, 63.49; H, 3.62; N, 11.85; S, 6.78; found C,
63.34; H, 3.61; N, 11.92; S, 6.72.
4.3. FABP inhibitory activity assays
To analyze the inhibitory activity of FABP4 ligands, a displacement assay was utilized as described
by the Cayman’s instruction, FABP4 Inhibitor/Ligand Screening Assay Kit, Item 10010231. The
samples of compounds AST_1–3 for activity determination were prepared as a stock solution (1
mM) in DMSO. On the day of activity assay, the compounds were all diluted in phosphate buffer
solution (PBS, pH 7.4) to different concentrations (50, 25, 12.5, 6.25, 3.12, 1.56 and 0.15 M).
Appropriate concentrations of DMSO in PBS was used as control. The detection reagent (FABP
Assay Detection Reagent, Item 10010376) was used as provided by the Cayman’s kit. The diluted
Detection Reagent probe was mixed with FABP4 protein present in the kit and incubated for 10 min
at room temperature. Compounds were then added and equilibrated for another 10 min. Lastly, the
fluorescence signal was recorded at 470 nm (i.e. emission, with the excitation fixed at 370 nm) with
a CytoFluor® Series 4000 Fluorescence Multi-Well Plate Reader. The plotted displacement curves
for Arachidonic acid (positive control) and AST_1–3 are reported in the supporting information.
Appendix A. Supplementary data
Supplementary data related to this article can be found at https://doi.org/
Acknowledgments
Free academic licenses from Cresset and ChemAxon for their suites of programs are gratefully
acknowledged.
19

References
[1] U.N. Das, Essential Fatty acids - a review, Curr Pharm Biotechnol, 7 (2006) 467-482.
[2] G. Boden, Free fatty acids (FFA), a link between obesity and insulin resistance, Front. Biosci., 3
(1998) 169-175.
[3] R.A. DeFronzo, Dysfunctional fat cells, lipotoxicity and type 2 diabetes, Int. J. Clin. Pract.
Suppl., (2004) 9-21.
[4] S.G. Sheth, F.D. Gordon, S. Chopra, Nonalcoholic steatohepatitis, Ann. Intern. Med., 126
(1997) 137-145.
[5] G. Boden, Obesity and free fatty acids, Endocrinol Metab Clin North Am, 37 (2008) 635-646,
viii-ix.
[6] J. Storch, A.E. Thumser, The fatty acid transport function of fatty acid-binding proteins,
Biochim. Biophys. Acta, 1486 (2000) 28-44.
[7] M.I. Queipo-Ortuno, X. Escote, V. Ceperuelo-Mallafre, L. Garrido-Sanchez, M. Miranda, M.
Clemente-Postigo, R. Perez-Perez, B. Peral, F. Cardona, J.M. Fernandez-Real, F.J. Tinahones,
J. Vendrell, FABP4 dynamics in obesity: discrepancies in adipose tissue and liver expression
regarding circulating plasma levels, PLoS One, 7 (2012) e48605.
[8] B.R. Thompson, A.M. Mazurkiewicz-Munoz, J. Suttles, C. Carter-Su, D.A. Bernlohr,
Interaction of Adipocyte Fatty Acid-binding Protein (AFABP) and JAK2 AFABP/aP2 AS A
REGULATOR OF JAK2 SIGNALING, Journal of Biological Chemistry, 284 (2009) 13473-
13480.
[9] A. Adida, F. Spener, Adipocyte-type fatty acid-binding protein as inter-compartmental shuttle
for peroxisome proliferator activated receptor gamma agonists in cultured cell, Bba-Mol Cell
Biol L, 1761 (2006) 172-181.
[10] K.M. Nieman, H.A. Kenny, C.V. Penicka, A. Ladanyi, R. Buell-Gutbrod, M.R. Zillhardt, I.L.
Romero, M.S. Carey, G.B. Mills, G.S. Hotamisligil, S.D. Yamada, M.E. Peter, K. Gwin, E.
Lengyel, Adipocytes promote ovarian cancer metastasis and provide energy for rapid tumor
growth, Nat Med, 17 (2011) 1498-1503.
[11] A. Tolle, S. Suhail, M. Jung, K. Jung, C. Stephan, Fatty acid binding proteins (FABPs) in
prostate, bladder and kidney cancer cell lines and the use of IL-FABP as survival predictor in
patients with renal cell carcinoma, BMC Cancer, 11 (2011) 302.
[12] H. Uehara, T. Takahashi, M. Oha, H. Ogawa, K. Izumi, Exogenous fatty acid binding protein 4
promotes human prostate cancer cell progression, Int J Cancer, 135 (2014) 2558-2568.
[13] A. Yang, H. Zhang, Y. Sun, Y. Wang, X. Yang, X. Yang, H. Zhang, W. Guo, G. Zhu, J. Tian,
Y. Jia, Y. Jiang, Modulation of FABP4 hypomethylation by DNMT1 and its inverse interaction
20

with miR-148a/152 in the placenta of preeclamptic rats and HTR-8 cells, Placenta, 46 (2016)
49-62.
[14] D. Lee, K. Wada, Y. Taniguchi, H. Al-Shareef, T. Masuda, Y. Usami, T. Aikawa, M. Okura,
Y. Kamisaki, M. Kogo, Expression of fatty acid binding protein 4 is involved in the cell growth
of oral squamous cell carcinoma, Oncol. Rep., 31 (2014) 1116-1120.
[15] G. Floresta, V. Pistara, E. Amata, M. Dichiara, A. Marrazzo, O. Prezzavento, A. Rescifina,
Adipocyte fatty acid binding protein 4 (FABP4) inhibitors. A comprehensive systematic
review, Eur J Med Chem, 138 (2017) 854-873.
[16] A. Bosquet, J. Girona, S. Guaita-Esteruelas, M. Heras, P. Saavedra-Garcia, N. Martinez-
Micaelo, L. Masana, R. Rodriguez-Calvo, FABP4 inhibitor BMS309403 decreases saturated-
fatty-acid-induced endoplasmic reticulum stress-associated inflammation in skeletal muscle by
reducing p38 MAPK activation, Biochim Biophys Acta Mol Cell Biol Lipids, 1863 (2018) 604-
613.
[17] C. Look, I. Morano, M. Ehrhart-Bornstein, S.R. Bornstein, V. Lamounier-Zepter, BMS309403
directly suppresses cardiac contractile function, Naunyn Schmiedebergs Arch Pharmacol, 384
(2011) 255-263.
[18] M.Y. Lee, H. Li, Y. Xiao, Z. Zhou, A. Xu, P.M. Vanhoutte, Chronic administration of
BMS309403 improves endothelial function in apolipoprotein E-deficient mice and in cultured
human endothelial cells, Br J Pharmacol, 162 (2011) 1564-1576.
[19] T. Okada, M. Hiromura, M. Otsuka, S. Enomoto, H. Miyachi, Synthesis of BMS-309403-
related compounds, including [(1)(4)C]BMS-309403, a radioligand for adipocyte fatty acid
binding protein, Chem Pharm Bull (Tokyo), 60 (2012) 164-168.
[20] A.V. Hertzel, K. Hellberg, J.M. Reynolds, A.C. Kruse, B.E. Juhlmann, A.J. Smith, M.A.
Sanders, D.H. Ohlendorf, J. Suttles, D.A. Bernlohr, Identification and characterization of a
small molecule inhibitor of Fatty Acid binding proteins, J. Med. Chem., 52 (2009) 6024-6031.
[21] T. Barf, F. Lehmann, K. Hammer, S. Haile, E. Axen, C. Medina, J. Uppenberg, S. Svensson, L.
Rondahl, T. Lundback, N-Benzyl-indolo carboxylic acids: Design and synthesis of potent and
selective adipocyte fatty-acid binding protein (A-FABP) inhibitors, Bioorg. Med. Chem. Lett.,
19 (2009) 1745-1748.
[22] R. Ringom, E. Axen, J. Uppenberg, T. Lundback, L. Rondahl, T. Barf, Substituted
benzylamino-6-(trifluoromethyl)pyrimidin-4(1H)-ones: a novel class of selective human A-
FABP inhibitors, Bioorg. Med. Chem. Lett., 14 (2004) 4449-4452.
21

[23] X. Liu, X. Huang, W. Lin, D. Wang, Y. Diao, H. Li, X. Hui, Y. Wang, A. Xu, D. Wu, D. Ke,
New aromatic substituted pyrazoles as selective inhibitors of human adipocyte fatty acid-
binding protein, Bioorg. Med. Chem. Lett., 21 (2011) 2949-2952.
[24] H. Cai, Q. Liu, D. Gao, T. Wang, T. Chen, G. Yan, K. Chen, Y. Xu, H. Wang, Y. Li, W. Zhu,
Novel fatty acid binding protein 4 (FABP4) inhibitors: virtual screening, synthesis and crystal
structure determination, Eur. J. Med. Chem., 90 (2015) 241-250.
[25] F. Lehmann, S. Haile, E. Axen, C. Medina, J. Uppenberg, S. Svensson, T. Lundback, L.
Rondahl, T. Barf, Discovery of inhibitors of human adipocyte fatty acid-binding protein, a
potential type 2 diabetes target, Bioorg. Med. Chem. Lett., 14 (2004) 4445-4448.
[26] Q. Xu, L. Huang, J. Liu, L. Ma, T. Chen, J. Chen, F. Peng, D. Cao, Z. Yang, N. Qiu, J. Qiu, G.
Wang, X. Liang, A. Peng, M. Xiang, Y. Wei, L. Chen, Design, synthesis and biological
evaluation of thiazole- and indole-based derivatives for the treatment of type II diabetes, Eur. J.
Med. Chem., 52 (2012) 70-81.
[27] Y. Wang, W.K. Law, J.S. Hu, H.Q. Lin, T.M. Ip, D.C. Wan, Discovery of FDA-approved
drugs as inhibitors of fatty acid binding protein 4 using molecular docking screening, J. Chem.
Inf. Model., 54 (2014) 3046-3050.
[28] Y. Zhou, T. Nie, Y. Zhang, M. Song, K. Li, M. Ding, K. Ding, D. Wu, Y. Xu, The discovery of
novel and selective fatty acid binding protein 4 inhibitors by virtual screening and biological
evaluation, Bioorg Med Chem, (2016).
[29] G. Floresta, O. Apirakkan, A. Rescifina, V. Abbate, Discovery of High-Affinity Cannabinoid
Receptors Ligands through a 3D-QSAR Ushered by Scaffold-Hopping Analysis, Molecules, 23
(2018).
[30] G. Floresta, V. Pittala, V. Sorrenti, G. Romeo, L. Salerno, A. Rescifina, Development of new
HO-1 inhibitors by a thorough scaffold-hopping analysis, Bioorg Chem, 81 (2018) 334-339.
[31] G. Floresta, E. Amata, M. Dichiara, A. Marrazzo, L. Salerno, G. Romeo, O. Prezzavento, V.
Pittala, A. Rescifina, Identification of Potentially Potent Heme Oxygenase 1 Inhibitors through
3D-QSAR Coupled to Scaffold-Hopping Analysis, Chemmedchem, 13 (2018) 1336-1342.
[32] L. Salerno, E. Amata, G. Romeo, A. Marrazzo, O. Prezzavento, G. Floresta, V. Sorrenti, I.
Barbagallo, A. Rescifina, V. Pittala, Potholing of the hydrophobic heme oxygenase-1 western
region for the search of potent and selective imidazole-based inhibitors, Eur J Med Chem, 148
(2018) 54-62.
[33] G. Floresta, A. Rescifina, A. Marrazzo, M. Dichiara, V. Pistara, V. Pittala, O. Prezzavento, E.
Amata, Hyphenated 3D-QSAR statistical model-scaffold hopping analysis for the identification
22

of potentially potent and selective sigma-2 receptor ligands, Eur J Med Chem, 139 (2017) 884-
891.
[34] A. Rescifina, G. Floresta, A. Marrazzo, C. Parenti, O. Prezzavento, G. Nastasi, M. Dichiara, E.
Amata, Sigma-2 receptor ligands QSAR model dataset, Data Brief, 13 (2017) 514-535.
[35] A. Rescifina, G. Floresta, A. Marrazzo, C. Parenti, O. Prezzavento, G. Nastasi, M. Dichiara, E.
Amata, Development of a Sigma-2 Receptor affinity filter through a Monte Carlo based QSAR
analysis, Eur J Pharm Sci, 106 (2017) 94-101.
[36] J.M. Luco, F.H. Ferretti, QSAR based on multiple linear regression and PLS methods for the
anti-HIV activity of a large group of HEPT derivatives, J Chem Inf Comput Sci, 37 (1997) 392-
401.
[37] A. Surribas, J.M. Amigo, J. Coello, J.L. Montesinos, F. Valero, S. Maspoch, Parallel factor
analysis combined with PLS regression applied to the on-line monitoring of Pichia pastoris
cultures, Anal Bioanal Chem, 385 (2006) 1281-1288.
[38] K. Tuppurainen, S.P. Korhonen, J. Ruuskanen, Performance of multicomponent self-
organizing regression (MCSOR) in QSAR, QSPR, and multivariate calibration: comparison
with partial least-squares (PLS) and validation with large external data sets, SAR QSAR
Environ Res, 17 (2006) 549-561.
[39] G. Palermo, P. Piraino, H.D. Zucht, Performance of PLS regression coefficients in selecting
variables for each response of a multivariate PLS for omics-type data, Adv Appl Bioinform
Chem, 2 (2009) 57-70.
[40] T. Asadollahi, S. Dadfarnia, A.M. Shabani, J.B. Ghasemi, M. Sarkhosh, QSAR models for
CXCR2 receptor antagonists based on the genetic algorithm for data preprocessing prior to
application of the PLS linear regression method and design of the new compounds using in
silico virtual screening, Molecules, 16 (2011) 1928-1955.
[41] A. Golbraikh, A. Tropsha, Beware of q2!, J Mol Graph Model, 20 (2002) 269-276.
[42] T. Cheeseright, M. Mackey, S. Rose, A. Vinter, Molecular field extrema as descriptors of
biological activity: definition and validation, J Chem Inf Model, 46 (2006) 665-676.
[43] P.H. Olesen, The use of bioisosteric groups in lead optimization, Current opinion in drug
discovery & development, 4 (2001) 471-478.
[44] A. Burger, Isosterism and bioisosterism in drug design, Prog Drug Res, 37 (1991) 287-371.
[45] G.A. Patani, E.J. LaVoie, Bioisosterism: A Rational Approach in Drug Design, Chem Rev, 96
(1996) 3147-3176.
[46] J.J. Irwin, B.K. Shoichet, ZINC--a free database of commercially available compounds for
virtual screening, J Chem Inf Model, 45 (2005) 177-182.
23

[47] A.P. Bento, A. Gaulton, A. Hersey, L.J. Bellis, J. Chambers, M. Davies, F.A. Kruger, Y. Light,
L. Mak, S. McGlinchey, M. Nowotka, G. Papadatos, R. Santos, J.P. Overington, The ChEMBL
bioactivity database: an update, Nucleic Acids Res, 42 (2014) D1083-1090.
[48] R. Abali, I. Temel Yuksel, M.A. Yuksel, B. Bulut, M. Imamoglu, V. Emirdar, F. Unal, S.
Guzel, C. Celik, Implications of circulating irisin and Fabp4 levels in patients with polycystic
ovary syndrome, J Obstet Gynaecol, (2016) 1-5.
[49] S.A. Abdelwahab, Y. Owada, N. Kitanaka, A. Adida, H. Sakagami, M. Ono, M. Watanabe, F.
Spener, H. Kondo, Enhanced expression of adipocyte-type fatty acid binding protein in murine
lymphocytes in response to dexamethasone treatment, Mol Cell Biochem, 299 (2007) 99-107.
[50] S. Bag, S. Ramaiah, A. Anbarasu, fabp4 is central to eight obesity associated genes: a
functional gene network-based polymorphic study, J Theor Biol, 364 (2015) 344-354.
[51] A.E. Thumser, J.B. Moore, N.J. Plant, Fatty acid binding proteins: tissue-specific functions in
health and disease, Curr Opin Clin Nutr Metab Care, 17 (2014) 124-129.
[52] S. Guaita-Esteruelas, J. Guma, L. Masana, J. Borras, The peritumoural adipose tissue
microenvironment and cancer. The roles of fatty acid binding protein 4 and fatty acid binding
protein 5, Mol Cell Endocrinol, (2017).
[53] K. Kawaguchi, A. Kinameri, S. Suzuki, S. Senga, Y. Ke, H. Fujii, The cancer-promoting gene
fatty acid-binding protein 5 (FABP5) is epigenetically regulated during human prostate
carcinogenesis, Biochem J, 473 (2016) 449-461.
[54] S. Guaita-Esteruelas, A. Bosquet, P. Saavedra, J. Guma, J. Girona, E.W. Lam, K. Amillano, J.
Borras, L. Masana, Exogenous FABP4 increases breast cancer cell proliferation and activates
the expression of fatty acid transport proteins, Mol Carcinog, 56 (2017) 208-217.
[55] M. Furuhashi, S. Saitoh, K. Shimamoto, T. Miura, Fatty Acid-Binding Protein 4 (FABP4):
Pathophysiological Insights and Potent Clinical Biomarker of Metabolic and Cardiovascular
Diseases, Clin Med Insights Cardiol, 8 (2014) 23-33.
[56] K. Kawaguchi, S. Senga, C. Kubota, Y. Kawamura, Y. Ke, H. Fujii, High expression of Fatty
Acid-Binding Protein 5 promotes cell growth and metastatic potential of colorectal cancer cells,
FEBS Open Bio, 6 (2016) 190-199.
[57] T. Yamamoto, M. Furuhashi, T. Sugaya, T. Oikawa, M. Matsumoto, Y. Funahashi, Y.
Matsukawa, M. Gotoh, T. Miura, Transcriptome and Metabolome Analyses in Exogenous
FABP4- and FABP5-Treated Adipose-Derived Stem Cells, PLoS One, 11 (2016) e0167825.
[58] J. Hao, F. Yan, Y. Zhang, A. Triplett, Y. Zhang, D.A. Schultz, Y. Sun, J. Zeng, K.A.T.
Silverstein, Q. Zheng, D.A. Bernlohr, M.P. Cleary, N.K. Egilmez, E. Sauter, S. Liu, J. Suttles,
24

B. Li, Expression of Adipocyte/Macrophage Fatty Acid-Binding Protein in Tumor-Associated
Macrophages Promotes Breast Cancer Progression, Cancer Res, 78 (2018) 2343-2355.
[59] X. Peng, K. Studholme, M.P. Kanjiya, J. Luk, D. Bogdan, M.W. Elmes, G. Carbonetti, S.
Tong, Y.H. Gary Teng, R.C. Rizzo, H. Li, D.G. Deutsch, I. Ojima, M.J. Rebecchi, M. Puopolo,
M. Kaczocha, Fatty-acid-binding protein inhibition produces analgesic effects through
peripheral and central mechanisms, Mol Pain, 13 (2017) 1744806917697007.
[60] G.G. Martin, D. Landrock, S. Chung, L.J. Dangott, D.R. Seeger, E.J. Murphy, M.Y. Golovko,
A.B. Kier, F. Schroeder, Fabp1 gene ablation inhibits high-fat diet-induced increase in brain
endocannabinoids, J Neurochem, 140 (2017) 294-306.
[61] H. Huang, A.L. McIntosh, G.G. Martin, D. Landrock, S. Chung, K.K. Landrock, L.J. Dangott,
S. Li, A.B. Kier, F. Schroeder, FABP1: A Novel Hepatic Endocannabinoid and Cannabinoid
Binding Protein, Biochemistry, 55 (2016) 5243-5255.
[62] M.W. Elmes, M. Kaczocha, W.T. Berger, K. Leung, B.P. Ralph, L. Wang, J.M. Sweeney, J.T.
Miyauchi, S.E. Tsirka, I. Ojima, D.G. Deutsch, Fatty acid-binding proteins (FABPs) are
intracellular carriers for Delta9-tetrahydrocannabinol (THC) and cannabidiol (CBD), J. Biol.
Chem., 290 (2015) 8711-8721.
[63] Y.T. Wang, C.H. Liu, H.L. Zhu, Fatty acid binding protein (FABP) inhibitors: a patent review
(2012-2015), Expert Opin Ther Pat, 26 (2016) 767-776.
[64] Y. Beniyama, K. Matsuno, H. Miyachi, Structure-guided design, synthesis and in vitro
evaluation of a series of pyrazole-based fatty acid binding protein (FABP) 3 ligands, Bioorg
Med Chem Lett, 23 (2013) 1662-1666.
[65] A. Cheng, S.A. Best, K.M. Merz, Jr., C.H. Reynolds, GB/SA water model for the Merck
molecular force field (MMFF), J Mol Graph Model, 18 (2000) 273-282.
[66] J.J. Stewart, Optimization of parameters for semiempirical methods IV: extension of MNDO,
AM1, and PM3 to more main group elements, J Mol Model, 10 (2004) 155-164.
[67] J.J. Stewart, MOPAC: a semiempirical molecular orbital program, J Comput Aided Mol Des, 4
(1990) 1-105.
[68] L. Castanar, G.D. Poggetto, A.A. Colbourne, G.A. Morris, M. Nilsson, The GNAT: A new tool
for processing NMR data, Magn Reson Chem, (2018).
25