Synthesis, docking, machine learning and antiproliferative activity of the 6-ferrocene/heterocycle-2-aminopyrimidine and 5-ferrocene-1H-Pyrazole derivatives obtained by microwave-assisted Atwal reaction as potential anticancer agents

Article abstract of DOI:10.1016/j.bmcl.2021.128240

A simple and fast methodology under microwave irradiation for the synthesis of 2-aminopyrimidine and pyrazole derivatives using Atwal reaction is reported. After the optimization of the reaction conditions, eight 2-aminolpyrimidines containing ferrocene a

Full text of DOI:10.1016/j.bmcl.2021.128240

Bioorg. Med. Chem. Lett. 48 (2021) 128240
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Synthesis, docking, machine learning and antiproliferative activity of the
6-ferrocene/heterocycle-2-aminopyrimidine and 5-ferrocene-1H-Pyrazole
derivatives obtained by microwave-assisted Atwal reaction as potential
anticancer agents
,
f
Eclair Venturini Filhoa, Jorge W.S. Pinaa, Mariana K. Antoniazia, Laiza B. Loureiroa
,
a
b
c,
d
c
˜
´
Marcos A. Ribeiro , Carlos B. Pinheiro , Celina J. Guimaraes , Fatima C.E. de Oliveira ,
,
*
Claudia Pessoac, Alex G. Tarantoe, Sandro J. Grecoa
a
´
Chemistry Department, Federal University of Espírito Santo, Vitoria, Espírito Santo CEP.:29075-910, Brazil
b
c
ˆ
Physical Department, Minas Gerais Federal University, Av. Antonio Carlos 6627, Pampulha, Belo Horizonte, Minas Gerais CEP.: 30161-970 Brazil
´
´
Department of Physiology and Pharmacology, Faculty of Medicine, Federal University of Ceara., Fortaleza, Ceara CEP 60430-275, Brazil
d Pharmacy Sector, Foundation of Oncology Control of the State of Amazonas, Manaus, Amazonas, CEP 69040-010, Brazil
e
˜
˜
˜
˜
Laboratory of Drug Design and Bioinformatics, Federal University of Sao Joao del-Rei, Sao Joao del-Rei, Minas Gerais CEP: 36307-352, Brazil
A R T I C L E I N F O
A B S T R A C T
Keywords:
A simple and fast methodology under microwave irradiation for the synthesis of 2-aminopyrimidine and pyrazole
derivatives using Atwal reaction is reported. After the optimization of the reaction conditions, eight 2-aminolpyr-
imidines containing ferrocene and heterocycles and three ferrocene pyrazoles were synthesized from the
respective chalcones in good yields. Eight compounds had their structure determined by X-ray diffraction. The
molecular hybrid 6a-h and 9a-c were tested on four cancer cell lines - HCT116, PC3, HL60 and SNB19 - where
four pyrimidine 6a, 6f-h and one pyrazole 9c derivatives show promising antiproliferative activity. In addition,
docking simulation and machine learning methods were carried out to explain the biological activity achieved by
the synthetized compounds.
Pyrimidine
Ferrocene
Atwal reaction
Antiproliferative activity
Computational studies
Cancer has become a threat to human beings as millions of people are
suffering by this disease in all world. Although efforts have been made in
reducing cancer incidence, the numbers of cancer patient continue to
increase, probably due to the population aging and greater exposure to
carcinogenic agents. Therefore, development of effective and safe anti-
cancer agents with less toxicity and high potency remains critically
important.
(pyridine-3-yl)pyrimidin-2-amine 3 and evaluated in vitro cytotoxicity
against a panel of human cancer cell lines namely, HeLa (human cervix),
NCI-H460 (human lung), PC-3 (human prostate), and NIH-3T3 (mouse
embryo fibroblasts) cell lines.2 Molecule 4 was prepared in the work of
Yoo and coworkers and presented antiproliferative activities against
HeLa by inhibitory activities in AXL kinase.3 Several other examples can
be easily found in the literature.4-9
In this context, 2-aminopyrimidine and its derivatives have a special
place in cancer chemotherapy for presenting anticancer activity against
different types of cancer and for acting through various mechanisms of
action. As examples, we can highlight the recent works of Ding and
coworkers that designed and synthesized a series of 2-aminopyrimidine
derivatives including the compound 2 as highly selective FGFR4 in-
hibitors in breast cancer cells (Fig. 1).1 In another example Alam
research group synthetized 4-(3-(aryl)-1-phenyl-1H-pyrazol-4-yl)-6-
The use of ferrocene in medicinal applications is an active research
area and several reports have shown that some ferrocene derivatives are
highly active in vitro and in vivo, against several diseases including
cancer.10-13 The anticancer potential of ferrocene derivatives was first
reported in the 1970s, by Brynes and co-workers, when they described
the antitumor activity of ferrocene compounds bearing amine or amide
groups against lymphocytic leukemia P-388.14 Since then efforts have
been made to synthesize ferrocene derivatives with anticancer activity,
* Corresponding author.
E-mail address: sandro.greco@ufes.br (S.J. Greco).
f
Currently at the State University of Campinas.
https://doi.org/10.1016/j.bmcl.2021.128240
Received 3 February 2021; Received in revised form 2 June 2021; Accepted 28 June 2021
Available online 2 July 2021
0960-894X/© 2021 Elsevier Ltd. All rights reserved.

E.V. Filho et al.
Bioorganic & Medicinal Chemistry Letters 48 (2021) 128240
Table 1
Synthesis of the precursor chalcones.
Compound
Chalcone
Time (h)
24
Solvent
MeOH
Yield (%)
97
5a
5b
5c
5d
12
MeOH
EtOH
95
89
97
120
21
MeOH
Fig. 1. Selected examples of 2-aminopyrimidines with anticancer activity.
5e
5f
3
MeOH
MeOH
54
45
27
5g
5h
3
2
MeOH
MeOH
90
75
media (Scheme 1).20
In the first step, a Claisen-Schmidt condensation reaction was used to
prepare chalcones following the standard methodology described in the
literature.19,21 This condensation procedure is attractive since it pre-
dominantly generates the (E)-isomer.22 The eight chalcones obtained,
the reaction times and the yield results for each one of them are shown in
Table 1.
Fig. 2. Molecular Hybrids pyrimidine and pyrazole/ferrocene with anti-
cancer activity.
For all chalcones, except 5c, an enolizable ketone (acetophenone or
p-nitro or p-amino acetophenone) was reacted in the presence of
aqueous sodium hydroxide in a methanol solution of the respective non-
enolizable aldehyde (ferrocene carbaldehyde or 2-phenyl-2H-1,2,3-tra-
zol-4-carbaldehyde or 1H-pyrrole-2-carbaldehyde or thiophene-2-
carbaldehyde). The aldehyde 2-phenyl-2H-1,2,3-triazole was prepared
in three steps from D-(+)-glucose, as previously reported.23
Using the Claisen-Schmidt condensation with NaOH as the base and
methanol as the solvent, the chalcone 5c was obtained with yield around
50% and a long reaction time (120 h) was required to consume the re-
agents. In other to improve both reaction time and yield in 5c produc-
tion, alternative methodologies were explored: reactions with different
bases (EtONa, tBuONa, DIPEA and DBU), base excess, reflux heating and
different solvents (EtOH, iPrOH, CH2Cl2 and THF) were tested. The best
result obtained in terms of yield was using EtOH, two equivalents of
NaOH and room temperature (89% yield).
Comparing the reaction times for obtaining the chalcones 5a-c and
5d-f, the effect of the substituent at the para position of the aromatic ring
is clearly perceived, since the chalcones 5b and 5e containing the nitro
group, strongly electrons withdrawer, increases the speed of the reac-
tion. This can be explained for the greater enolate stability.24
Through reaction times of chalcones 5a, 5d, 5g and 5h, we can also
note the influence of the aldehyde reactivity. The experimental results
show that the pyrrole aldehyde presents the most electrophilic carbonyl
group, which is consistent with the electronic effects characteristic of
these heterocycles.25
Scheme 1. Synthesis of 2-aminopyrimidines 6a-h.
and among the great structural diversity, molecular hybrids containing
the pyrimidine and pyrazole group can be highlighted. Some examples
are shown in Fig. 2.15–18
In this work we have explored the traditional and microwave assisted
methodology of the Atwal reaction to obtain 6-ferrocene/heterocycle-2-
aminopyrimidine and 5-ferrocene-1H-Pyrazole derivatives. Our strategy
was based on an easy and rapid two-step synthesis pathway shown in
Scheme 1. We started with the preparation of ferrocene and heterocycle
chalcones 5a-h using the classical Claisen-Schmidt condensation reac-
tion.19 Then, 2-aminopyrimidines 6a-h were prepared via Atwal reac-
tion by of the respective chalcones with the guanidine chloride in basic
Structures of compounds 5a-h are consistent with IR, 1H and 13C
NMR spectra. All compounds showed bands near 1630 cmꢀ 1 in the IR
–
spectra for the C O stretch of chalcones. In contrast with the litera-
–
ture,22 all compounds were obtained, after purification, as single
2

E.V. Filho et al.
Bioorganic & Medicinal Chemistry Letters 48 (2021) 128240
Fig. 3. Molecular representation and atomic labeling scheme of chalcones 5c
and 5f obtained from single crystal X-ray diffraction technique.
Scheme 2. Mechanistic proposal for Atwal reaction.
Table 2
Conventional and microwave heating Atwal reaction.
Considering the low yields obtained, different reaction conditions
were tested varying the solvent (H2O, MeOH, EtOH, tBuOH, THF, DMF
and DMSO) and the bases (NaOH, EtOK, tBuOK, DBU, Et3N, DIPEA and
Py). For synthesis of 6a, 6b, 6d, 6e and 6g the best combination was
NaOH/EtOH under reflux whereas for 6c, 6f and 6h better yields were
obtained using tBuOK/tBuOH. The detailed experimental procedures for
these reactions are described in the supplementary material.
Conventional Heating
Microwave Heating
Compound
Pyrimidine
Time
Yield (%)
Time
Yield (%)
6a
15 h
9 h
57a
1 h
45c
25b
20 min
62d
6b
6c
6d
4 h
3 h
63a
10b
10 min
15 min
15c
71d
All Atwal reactions described previously using NaOH/EtOH and
tBuOK/tBuOH as base/solvent combinations were performed under
microwave irradiation and the best results are shown in the Table 2.
These findings showed that under microwave assisted conditions the
reaction times decreased considerably, and similar or higher yields were
obtained when compared with the traditional heating method. This can
be explained by the first stage of the cyclocondensation reaction
involving an 1,4-addition between guanidine and chalcone. The formed
adduct is thermodynamically favored at higher temperatures,27 and the
transfer of energy to the reaction system in heating by microwave in-
creases the efficiency of the process.
35 h
12 h
23a
54b
2 h
10c
75d
40 min
6 h
3 h
55a
10b
2 h
3 h
78c
42d
6e
6f
1 h
82a
15b
30 min
35 min
57c
22d
13 h
The spectral data for the known pyrimidines 6a, 6b, 6g and 6h
shown in the supplementary information agree with those previously
reported.28-30 The novel 2-aminopyrimidine (6c, 6d, 6e and 6f) were
characterized by IR, NMR spectroscopy and mass spectrometry. IR
spectroscopic data showed the formation of new bands around
3500–3200 cmꢀ 1 typical NH2 group, besides the absence of the char-
acteristic stretching band of the carbonyl group. According to the 1H
NMR, the formation of pyrimidine ring was verified through the signals
referring to the amino group and the singlet attributed to the methinic
hydrogen present in the pyrimidine ring around 6.20 and 7.20 ppm,
respectively.
3 h
2 h
23 a
34b
1 h
52c
27d
50 min
6g
6h
5 h
1 h
35a
20b
20 min
45 min
58c
25d
4 h
1 h
10a
50b
15 min
15 min
42c
80d
Furthermore, according to the 13C NMR spectra the signal at 100.6
ppm is attributed to the methinic carbon of the pyrimidine, where the
couplings that corroborate with this characterization can be seen in the
HSQC and HMBC contour maps. In case of 6c the characteristic signs of
the ferrocene group were observed in 1H NMR data at 4.03 (5H), 4.41
(2H) and 5.01 ppm (2H), in addition to the chemical shifts in 13C NMR at
68.1, 69.9, 70.4 and 82.6 ppm referring to the unsubstituted cyclo-
pentadienyl ring of ferrocene. The 1H NMR of triazole pyrimidines 6d-f
showed an important singlet around 8.50 ppm referring to triazole
hydrogen, this being the most deshielding hydrogen in the molecule.
The 13C NMR signals present in the triazole ring are assigned around
135.0 and 148.0 ppm.
diastereomers. To triazole chalcones 5d-f one characteristic singlet was
observed for hydrogen in the 1,2,3-triazole ring, around 8.10 ppm. The
E-geometry of the double bonds in 5a-h were assigned with basis on
their 1H NMR spectra. For example, compound 5a showed a signal for its
vinyl hydrogen at 7.10 ppm with a coupling constant of 12.0 Hz, which
agrees with the geometry trans. Furthermore, crystallization of 5c and 5f
with hot methanol provided a single crystal for X-ray diffraction study
revealed the (E) geometry for these chalcones, which in solid state
adopts a s-trans conformation (Fig. 3).
Initially, the synthesized chalcones were used to obtain the respec-
tive 2-aminopyrimidines 6a-h by Atwal reaction under traditional
heating conditions (Table 2). In this procedure, the cyclocondensation
occurs usually using guanidine chloride, NaOH as the base and ethanol
as solvent. Low yields were observed for all compounds, except for 6e,
which was obtained in 82% yield. An explanation for this is, probably,
due to the formation of by-products as described by Chebanov and co-
workers.26
The electrospray ionization mass spectrometry (ESI-MS) analysis of
novel 2-aminopyrimidine 6c, 6d, 6e and 6f showed a molecular ion peak
in 371.09523 m/z, 315.13524 m/z, 437.19340 m/z and 330.14617 m/z,
respectively, corroborating the spectroscopic results.
The mechanism proposed is initiated by the conjugated addition of
3

E.V. Filho et al.
Bioorganic & Medicinal Chemistry Letters 48 (2021) 128240
Table 3
Anti-proliferative activities of selected compounds against HCT116, PC3, HL60
and SNB19 cell lines.
Compound
Cell lines - IC50 (µM)
HCT116
PC3
HL60
SNB19
6a
6b
6c
6d
6e
6f
102.10
>200
115.00
120.20
132.80
95.27
56.99
55.47
>200
> 200
3.12
100.50
>200
3.98
>200
>200
>200
>200
132.10
106.50
85.11
173.50
>200
>200
60.44
>200
75.79
66.47
> 200
68.19
70.26
29.33
>200
>200
6.81
96.40
142.30
>200
Scheme 3. Ferrocene pyrazole synthesis.
58.23
33.56
81.55
101.30
>200
6g
6h
9a
9b
9c
124.40
guanidine to chalcone, as suggested by Atwal (Scheme 2).20,26,31 Sub-
sequently, 1,2-addition occurs between other guanidine nitrogen to the
chalcone carbonyl, forming the hydroxylated cyclic intermediate 7. This
hydroxyl is eliminated via anchimeric assistance and E1cb mechanism to
former dihydropyrimidinone 8. This compound is found in imine/
enamine tautomeric equilibrium, that after is easily oxidized in solution,
leading to the formation of the corresponding amine pyrimidine.
To attest the Atwal methodology under microwave irradiation and
increase its scope, 5-ferrocene-1H-pyrazole hybrids were synthesized
using phenyl hydrazine instead of guanidine in acid media (Scheme 3).
The spectral data for the known ferrocene 1H-pyrazole 9a agree with
those previously reported.32 The novel hybrid 9b and 9c were charac-
terized by IR, NMR spectroscopy and mass spectrometry.
IR spectroscopic data showed the formation of new bands around
1590 and 1490 cmꢀ 1 typical
ν
(C N and C C) pyrazole ring. Besides,
–
–
–
–
compounds 9b and 9c shows characteristic bands around 1330 cmꢀ 1
attributed to
ν
(NO2) and bands around 3500–3200 cmꢀ 1 typical NH2
group, respectively. As well the absence of the stretching band of the
carbonyl group. According to the 1H NMR, 9b and 9c had similar
behavior for 1H-pyrazole ring signals. Both spectral shows evidence for
ring formation through the two signals around 3.60–3.90 ppm referring
to the methylene hydrogens also one signal around 5.00–5.20 ppm
attributed for methinic hydrogen, all signals with double doublet
coupling.
Furthermore, according to the 13C NMR spectra all ferrocene-1H-
pyrazole hybrids presents characteristic signals of the ferrocene group.
The 13C NMR signals present in the pyrazoline ring are assigned 40.97,
59.73 and 143.73 ppm for 9b and 42.08, 58.49 and 145.71 ppm for 9c.
The electrospray ionization mass spectrometry (ESI-MS) analysis of
novel ferrocene-pyrazole 9b and 9c showed a molecular ion peak in
450.09154 m/z and 422.13103 m/z, respectively.
The geometrical properties of some of the compounds obtained could
be investigated by single crystal X-ray diffraction techniques. The crystal
properties, the data collection, and the structure refinement parameters
as well as the geometrical parameters are summarized in the supple-
mentary information. The molecular structures of chalcones were pre-
viously shown in Fig. 2 and structures of the 6a, 6b, 6c, 6d, 9a and 9b
are shown in Fig. 4.
X-ray structure determination of compounds 6a, 6b, 6d, 9a and 9b
have been obtained from ethanol and 5c, 5f and 6c from methanol under
slow evaporation. Chalcone 5c and pyrimidine 6a and 6c crystallized in
orthorhombic system with P212121, Pbca and Pbcn space groups,
respectively, and compound 6b crystallized in a triclinic P1 space group.
Monoclinic crystal system was obtained to 5f, 6d, 9a and 9b in space
group P21, P21/n, P21/n and P21/c, respectively.
Further crystallographic details for the structure reported in this
paper may be obtained from the Cambridge Crystallographic Data
Center, on quoting the depository numbers CCDC- 2041688, 2042671,
1874652, 1874676, 1874661, 2042672, 2042673 and 2042416 for the
Fig. 4. Molecular representation and atomic labeling scheme of pyrimidine
hybrids 6a, 6b, 6c and 6d obtained from single crystal X-ray diffraction tech-
nique. Atomic displacement parameters at 70%. Hydrogen atoms labels were
omitted for sake of clarity.
4

E.V. Filho et al.
Bioorganic & Medicinal Chemistry Letters 48 (2021) 128240
Fig. 5. Inhibitory concentration means (IC50) of the synthesized compound 6a-6h and 9a-9c against HCT116, PC3, HL60 and SNB19 tumoral cell lines.
carcinoma), PC3 (prostate adenocarcinoma), HL60 (acute promyelo-
cytic leukemia) and SNB19 (astrocytoma) tumoral cell lines are shown
in the Fig. 5.
Molecular hybrids 9c, 6g and 6h, showed good antiproliferative
activities with IC50 of 3.12, 55.47 and 56.99 µM, respectively, for
HCT116. For the PC3 cancer cell line, 6f, 6g and 6h, exhibited good
antiproliferative activities with the IC50 values of 33.56, 58.23 and
81.55 µM, respectively. The compounds 6a, 6f and 6h, inhibited Pro-
myelocytic leukemia cell lines with IC50 of 3.98; 6.81 and 29.33 µM and
lastly, for SNB19 cancer cell line, compounds 9c with an IC50 value of
60.44 µM, 6g 85.11 µM and 6f 106.5 µM, presented the best results.
The screening data indicated that compound 9c with pyrazole core
and substituted with ferrocene and NH2Ph was more effective that the
similar compound 6c with pyrimidine core in the HCT116, HL60 and
SNB19 human cancer cell line (Scheme 4a). In addition, as previously
highlighted, 9c exhibited the best antiproliferative activities to HCT116
and SNB19 cell lines when compared to all compounds synthesized.
Furthermore, this hybrid showed predominant cytotoxic potential (3.12
to 124.4 μM) in all tested cell lines. The other pyrazole derivatives 9a
and 9b substituted with Ph and NO2Ph, respectively presented low in-
hibition activity.
The promising anticancer activity of the hybrids containing pyrimi-
dine and ferrocene 6a-c was confirmed only against the HL60 cancer
Scheme 4. Comparative summary of the biological activity of molecular hy-
brids: a) compares and proves that 1H-pyrazole is more active than the py-
rimidine core containing the same substituents in the HCT116, HL60 and
SNB19 cell lines; b) shows that among the four aminopyrimidines containing
the phenyl group on carbon C4 and different pharmacophore groups on carbon
C6 (ferrocene, triazole, thiophene and pyrrole), the one with ferrocene is the
most active against HL60; c) shows that among three aminopyrimidines con-
taining the phenyl group on carbon C4 and different heterocycles (triazole,
thiophene and pyrrole) on carbon C6, the one with thiophene is the most active
against PC3 human cancer line.
cells, with 6a being the most powerful – IC50 3,98 μM – among all
compounds. The Hybrid 6a when compared to structurally similar
compounds – 6d-triazole, 6g-thiophene, 6h-pyrrole – the synergetic
effect of ferrocene is evident (Scheme 4b). The other pyrimidine/
ferrocene hybrids 6b and 6c substituted on carbon C4 by p-NO2Ph and p-
NH2Ph showed low and moderate antiproliferative activities, respec-
tively. The pyrimidine derivatives with phenyl triazole on carbon C6 6d-
f showed only moderate anticancer activity and the most significant IC50
values were determined for the 6f with NH2Ph substituent in the pros-
tate carcinoma (58,23 μM). The hybrids 6g and 6h with thiophene and
compounds 5c, 5f, 6a-d and 9a-b, respectively.
pyrrole, respectively, were effective in practically all tested cell lines,
highlighting the compound 6g in the colon carcinoma, prostate carci-
noma, and astrocytoma cancers and hybrid 6h in the colon carcinoma,
prostate carcinoma, and Promyelocytic leukemia. It should be noted that
The newly synthesized compounds 6a-h and 9a-9c were screened for
their in vitro antiproliferative activities against four types of human
cancer cell lines including HCT116 (Human Colon Carcinoma), PC3
(Prostate Carcinoma), HL60 (Promyelocytic leukemia) and SNB19 (As-
trocytoma) by MTT based assay. The results are represented as the half
6g presented the best results, this is, the low IC50 values (33,56
prostate carcinoma – PC3 (Scheme 4c).
μM) to
maximal inhibitory concentration (IC50-μM) values in Table 3.
For computational studies, initially, all 2-aminopyrimidines 6a-h
and pyrazole 9a-c structures were generated using the academic version
Graph representation of inhibitory concentration mean (IC50) of the
compound 6a-h and 9a-c against HCT116 (human colorectal
5

E.V. Filho et al.
Bioorganic & Medicinal Chemistry Letters 48 (2021) 128240
Table 4
Grid box parameters for all molecular targets. All values are in Å.
Receptors
1 M17
4HLW
5LGE
5QGF
Center X
Center Y
Center Z
22.108
0.234
16.542
17.872
29.481
ꢀ 14.976
ꢀ 83.394
24.569
43.588
ꢀ 31.027
ꢀ 823.967
52.778
of Discovery Studio.33 Next, all compounds were refined by the run_-
mopac routine using the Eigenvector Following (EF) routine.34 Run_-
mopac can refine a set of ligands into the fold using the parametric
method 7 (PM7)35 of MOPAC.34 During the process of refinement, for
each molecule, run_mopac warms the uses if the number of alfa electrons
are different from beta electrons. This difference suggests that there is a
missing atom, for instance, avoiding human errors. Noteworthy that
Autodock tools were not able to generate the ligand pdbqt files property.
Thus, the pdb files were converted to pdbqt files using Open Babel
software.
Fig. 6. Redocking results: A, B, C and D correspond to 1 M17, 4HLW, 5LGE and
5QGF, respectively.
The receptors of HCT116, PC3, HL60 and SNB19 were obtained from
Protein Data Bank (PDB) under code 5QGF,36 5LGE,37 4HLW,38 1 M17.39
All receptors were prepared by construction of loop regions using
MODELLER implemented into MolAr. Next, hydrogen atoms were
included, and the structures were optimized by steepest descent and
conjugate gradient algorithms using Chimera. Following, all molecular
targets pdb files were converted o pdbqt through Autodock Tools. The
compounds were docked using Autodock Vina. The exhaustiveness and
all values of the grid box were adjusted for 24; the atomic coordinates of
the center of grid box were describe according to the Table 4.
At this point, it is worth noting that the therapeutic targets chosen
are well established in the literature for treatment of different cancer
types:40-44 The 1M17 (PDB code) is epidermal growth factor receptor
(EGFR) which is overexpressed in colorectal cancer cells.43 The EGFR is
a transmembrane receptor that transmits signals from one cell to
another to tell to grow and multiply or to die. Therefore, EGFR plays a
key role in cell development, proliferation, and differentiation within a
variety of tissues. Deregulation of their activity is extremely associated
with tumorigeneses, particularly in the lung, breast, ovarian and colo-
rectal cancer. The second target, 4HLW (PDB code) is a human
Androgen Receptor (AR), a member of the nuclear hormone receptor
(NHR) family. The BF3 (Binding Function 3 site of the AR) is a ligand-
dependent transcription factor with therapeutic relevance in prostate
cancer PC3 cells. Therapeutic targeting of the AR BF3 provides an op-
portunity to develop AR inhibitors with alternative mechanisms of ac-
tion and the potential to circumvent resistance to conventional
antiandrogens.42 Them, the 5LGE (PDB code) an isocitrate dehydroge-
nase (IDH) 1, is a metabolic enzyme having an important role in the
citric acid cycle. IDH1 R132H, the most frequent mutation of the IDH1
gene in glioma, results in missense replacement of arginine by histidine,
leading to production of the mutant enzyme that catalyzed the synthesis
of oncometabolite 2-hydroxyglutarate. IDH1 mutations are early events
in the development of gliomas, and thus the majority of diffuse astro-
cytoma’s and secondary glioblastomas.45 Lastly, 5QGF (PDB code) is a
Protein Tyrosine Phosphatase 1B (PTP1B). PTP1B is a critical regulator
of signaling pathways controlling metabolic homeostasis, cell prolifer-
ation, and immunity. Consequently, PTP1B inhibitory activity for the
treatment of cancer has been a fervent topic of great interest in the
pharmacology.40 Recent studies have shown that myeloid-specific
deficiency of PTP1B is sufficient to promote the development of acute
myeloid leukemia.46
Fig. 7. Hydrophobic surface docked results of the most active compounds with
respective cell line. Hydrogen atoms were omitted for a better visualization.
a good fit between docked and crystallographic ligand for the molecular
targets 5LGE, reaching root-mean-square deviation (RMSD) value close
to zero, and binding energy ꢀ 9.4 Kcal.molꢀ 1 (Fig. 6C). However, there
were not fitted between crystallographic ligand and docked of 1 M17,
4HLW and 5QGF (Fig. 6A, B and D). The redocking evaluation showed
that the grid box parameters were adjusted addressing the docked li-
gands correctly for the binding site. In addition, the binding energy of
redocked crystallographic ligands were ꢀ 7.0, ꢀ 6.2, ꢀ 9.4 and ꢀ 6.4 Kcal.
molꢀ 1 for 1 M17, 4HLW, 5LGE and 5QGF molecular targets, respec-
tively. These binding energies can be used as reference to select the
active compounds.
Following, all compounds were submitted against all receptors
through docking methodology.49,50 The docking results showed binding
energy from ꢀ 7.2 to ꢀ 9.0 Kcal.molꢀ 1 for 1 M17; ꢀ 6.3 to ꢀ 8.5 Kcal.
molꢀ 1 for 4HLW; ꢀ 6.8 to ꢀ 8.1 Kcal/mol for 5LGE; and ꢀ 6.3 to ꢀ 7.9
Kcal/mol for 5QGF. These findings suggest that these ligands can reach
the molecular targets. Noteworthy that all compounds had binding en-
ergy values less than the crystallographic ligand for 1 M17, 4HLW and
5QGF, with only one except, 9c for 5QGF. This data suggests that they
The docking methodology was evaluated by redocking.47 The run_-
mopac was implemented into Molecular Architect (MolAr)48, which is a
workflow integrating other software such as MODELLER and Dock 6
(Grid and Amber score), developed to perform structure based-drug
design experiments with a friendly interface. Figure 5 shows the
redocking of all molecular targets in this study. As can be seen, there was
6

E.V. Filho et al.
Bioorganic & Medicinal Chemistry Letters 48 (2021) 128240
Table 5
Data frame applied to build machine learning models.
Compound
Total Mol weight
cLogP
cLogS
Total Surface Area
Relative PSA
Polar Surface Area
Druglikenes
Binding Energy (Kcal/mol)
1M17
4HLW
5LGE
5QGF
6a
6b
6c
6d
6e
6f
356.229
402.234
371.244
314.351
359.348
329.366
253.328
236.277
404.293
449.290
419.308
2.35
2.41
1.67
2.62
1.69
1.94
3.31
2.10
3.86
2.94
3.18
ꢀ 3.526
ꢀ 5.833
ꢀ 3.602
ꢀ 3.732
ꢀ 4.192
ꢀ 3.808
ꢀ 4.436
ꢀ 3.231
ꢀ 4.272
ꢀ 4.732
ꢀ 4.348
190.53
265.66
199.05
245.3
0.19530
0.25457
0.26365
0.26914
0.35855
0.32027
0.29559
0.27178
0.06305
0.15739
0.11360
51.8
97.62
77.82
82.51
128.33
108.53
80.04
67.59
17.82
63.64
43.84
ꢀ 1.44
ꢀ 9.09
ꢀ 1.44
3.03
ꢀ 7.9
ꢀ 8.1
ꢀ 7.9
ꢀ 8.6
ꢀ 9.0
ꢀ 8.2
ꢀ 7.2
ꢀ 7.7
ꢀ 8.7
ꢀ 8.1
ꢀ 8.3
ꢀ 6.9
ꢀ 6.4
ꢀ 6.7
ꢀ 7.4
ꢀ 8.5
ꢀ 7.3
ꢀ 6.3
ꢀ 6.5
ꢀ 7.1
ꢀ 7.0
ꢀ 6.7
ꢀ 7.4
ꢀ 7.4
ꢀ 7.3
ꢀ 7.4
ꢀ 8.1
ꢀ 7.5
ꢀ 6.8
ꢀ 7.0
ꢀ 7.2
ꢀ 7.2
ꢀ 7.4
ꢀ 7.6
ꢀ 7.9
ꢀ 7.5
ꢀ 7.4
ꢀ 7.8
ꢀ 7.2
ꢀ 6.4
ꢀ 6.7
ꢀ 6.5
ꢀ 6.6
ꢀ 6.3
268.97
253.82
194.76
188.39
282.95
306.62
291.47
ꢀ 2.34
3.03
6g
6h
9a
9b
9c
1.46
0.52
1.29
ꢀ 3.90
1.25
An exhaustive combination of features was carried out to search the
best ML model using and KNN and LR algorithms for each cell line. The
significant models were found for SB19 and PC3 cell lines through KNN
and LR, respectively. In both models, the best features were cLogP,
druglikenes and the binding energy for the respective molecular target
(5LGE and 4HLW for SB19 and PC3, respectively). As a result, the KNN
model could achieve a precision of 1.00 and 0.67 to predict the active
and inactive compounds, respectively, with an accuracy of 0.83 and a
cross validation value of 0.67. These results means that KNN model had
a good prediction for active compounds (100%), but it had few efficient
to predicted inactive compounds. This weakness in the KNN model is
due to docking methodology, which generates false-positive results. In
addition, LR model had 0.59, 0.95 and 0.32 of regression coefficient,
cross validation value and root mean square error, respectively. The
ꢀ 0.48, 0.02, ꢀ 0.09 coefficient values were obtained for 4HLW, LogP
and druglikenes, respectively.
The Fig. 8 shows the correlation between real and predict biological
activity values. These findings suggest a good model able to biological
activity. For instance, the real log of biological active values was 9.22,
8.67 and 9.90; whereas the predicted values were 9.12, 9.33, 9.60,
respectively, for 6a, 6f and 9b compounds. Hence, two good ML models
were obtained, which can be used to predict the biological activity of
SB19 and PC3 cell lines, respectively.
Fig. 8. Correlation between real and predict biological activity for PC3 cell line
through linear regression using 4HMW, cLogP and druglines as features.
are more active than crystallographic ligand for the cell line previously
described.
Fig. 7 summarize the ligands with the best biological activity. As can
be see, the ferrocene 6a and 9c compounds showed the best biological
activity values for HCT116 (Fig. 7A), SNB19 (Fig. 7C) and HL60 cell
In conclusion, we developed a simple, fast, and efficient methodol-
ogy for the Atwal reaction under microwave irradiation to synthesize
eight 2-amino-4-phenylpyrimidine substituted at carbon C6 with het-
erocycles and ferrocene and three pyrazole derivatives with phenyl and
ferrocene in yield ranging from good to excellent – 52–80%. Further-
more, eight crystal structures were determined successfully.
lines (Fig. 7D), with binding energy of ꢀ 8.3, ꢀ 7.4 and ꢀ 7.6 Kcal.molꢀ 1
,
respectively. These values are low than the respective crystallographic
ligands (Fig. 6). In addition, as can be seen in the Fig. 6, the ferrocene
moiety can be form complex into hydrophobic and hydrophilic cavities
of the molecular targets. This feature can explain their biologic activity,
which the ferrocene compounds were activities in 3 of the 4 cell lines. In
contrast the pyrimidine 6g was the most active compounds for PC3 cell
line with binding energy of ꢀ 6.3 Kcal.molꢀ 1 performing hydrophilic and
hydrophobic molecular interaction with 4HLW molecular target.
To Machine Learning (ML) the Autodock Vina51 binding energy (BE)
and physical-chemistry features (descriptors) were obtained by Data-
warrior software.52 The parameters, cLog P, cLog S, Total Molecular
Weight, Relative Polar Surface Area (PSA), Drug likeness, Total Surface
and Polar Surface Area were used to develop ML models. Thus, super-
vised, and unsupervised models were generated using K-Nearest
Neighbors (KNN) and Linear Regression (LR) through Jupyter Note-
book.53 The data frame was per-processed using pandas and numpy li-
brary, which a Log function was applied for the biological values. No
missing values were found in the data frame. In addition, the compounds
with value equal to 20,000 were classified as inactive and the others as
active to build the KNN supervised model, which n value was set to 3.
The biological activity was the target to be predicted. In addition, the
structures with biological activity more than 2000 were removed from
the dataset. The binding energy, together with other descriptors, were
used to generate machine learning models, as shown in Table 5.
All hybrids screened for their in vitro anti-proliferative activities
against four cancer cell lines where it was possible to correlate the
anticancer activity with the structures of the synthesized hybrids. The
compounds 6g and 9c stand out showed predominant cytotoxic poten-
tial in all tested cell lines. Furthermore, the biological activity of PC3 cell
line could be estimated by docking simulation, whereas the biological
activity for HL60 and HCT116 could be estimated by ML model using LR
and RF, respectively. These calculations can be used to design new
compounds with anticancer activity.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Acknowledgments
The authors acknowledge the Conselho Nacional de Desenvolvi-
´
mento Científico e Tecnologico (CNPq 305117/2017-3 and 2020), the
Coordenadoria de Aperfeiçoamento de Pessoal do Nível Superior
7

E.V. Filho et al.
Bioorganic & Medicinal Chemistry Letters 48 (2021) 128240
21 Nielsen AT, Houlihan WJ. The Aldol Condesation. 2011.
˜
`
(CAPES), Fundaçao de Amparo a Pesquisa do Espírito Santo (FAPES),
22 Ducki S, Rennison D, Woo M, et al. Bioorg Med Chem. 2009;17:7698–7710.
23 Tadeu D, Gonzaga G, Braga L, et al. Eur J Med Chem. 2017;139:698–717.
24 Eicher T, Hauptmann S, Speicher A. The Chemistry of Heterocycles Structure,
Reactions, Synthesis, and Applications, Third Edit., 2012.
˜
`
Fundaçao de Amparo a Pesquisa de Minas Gerais (FAPEMIG APQ-
˜
˜
02742-17), Federal University of Sao Joao del-Rei (UFSJ/PPGBiotec/
˜
PPBE/PPGMQ-MG) and Fundaçao Cearense de Apoio ao Desenvolvi-
25 Joule JA, Keith Mills. Heterocyclic Chemistry, Fifth Edit., 2010.
´
mento Científico e Tecnologico (Funcap) for their financial support. A.G.
26 Chebanov VA, Gurley TW, Desenko SM. Azaheterocycles Based on ,
Carbonyls, 2008.
α,β-Unsaturated
T. is grateful to Mineriando Dados team.
27 Clayden J, Greeves N, Wothers P, Warren S. Organic Chemistry, New York, First
Edit., 2001.
Appendix A. Supplementary data
28 Rashinkar GS, Pore SB, Mote KB, Salunkhe RS. Indian J Chem. 2009;48:606–610.
29 Mariella RP, Zelko JJ. Master Sci. 1960;513:647–648.
30 El-Rayyes NR. J Heterocycl Chem. 1982;19:415–419.
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.bmcl.2021.128240.
31 Katritzky A. Advances in Heterocyclic Chemistry, First edit., 2016, vol. 119.
´
´
´
32 Zsoldos-Mady V, Ozohanics O, Csampai A, Kudar V, Frigyes D, Sohar P. J Organomet
Chem. 2009;694:4185–4195.
References
`
33 Dassault Systemes BIOVIA. 2016.
34 Stewart JJP. 2016, http://openmopac.net/MOPAC2016.html.
35 Dutra JDL, Filho MAM, Rocha GB, Freire RO, Simas AM, Stewart JJP. J Chem Theory
Comput. 2013;9:3333–3341.
1 Mo C, Zhang Z, Guise CP, et al. ACS Med Chem Lett. 2017;8:543–548.
2 Alam R, Alam A, Panda AK, Rahisuddin. Med Chem Res. 2017:560–570.
3 Choi MJ, Roh EJ, Hur W, et al. Bioorg Med Chem Lett. 2018;28:3761–3765.
4 Del Bello F, Farande A, Giannella M, et al. Bioorg Med Chem. 2015;23:5725–5733.
5 Kim S, Jung JK, Lee HS, et al. Bioorg Med Chem Lett. 2011;21:3002–3006.
6 Kim J, Hong S, Hong S. Bioorg Med Chem Lett. 2011;21:6977–6981.
7 Fukuda T, Ishiyama T, Katagiri T, et al. Bioorg Med Chem Lett. 2018;28:3333–3337.
8 Sutherlin DP, Bao L, Berry M, et al. J Med Chem. 2011;54:7579–7587.
9 Koroleva EV, Ignatovich ZI, Sinyutich YV, Gusak KN. Russ J Org Chem. 2016;52:
139–177.
36 Keedy DA, Hill ZB, Biel JT, et al. Elife. 2018;7:1–36.
37 Pusch S, Krausert S, Fischer V, et al. Acta Neuropathol. 2017;133:629–644.
38 Lack NA, Axerio-Cilies P, Tavassoli P, et al. J Med Chem. 2011;54:8563–8573.
39 Al-Suwaidan IA, Abdel-Aziz NI, El-Azab AS, et al. J Enzyme Inhib Med Chem. 2015;30:
679–687.
40 Tian Zhao B, Hung Nguyen D, Mi Lee B, et al. Med Chem Res. 2017;26:2870–2878.
41 Huang GH, Du L, Li N, et al. Cancer Lett. 2018;432:93–102.
42 Munuganti RSN, Hassona MDH, Leblanc E, et al. Chem Biol. 2014;21:1476–1485.
¨
43 Sever B, Altıntop MD, Ozdemir A, et al. Molecules. 2020:25.
10 van Staveren DR, Metzler-Nolte N. Chem Rev. 2004;104:5931–5985.
11 Ornelas C, New J. Chem. 2011;35:1973–1985.
44 Khan I, Garikapati KR, Setti A, et al. Eur J Med Chem. 2019;163:636–648.
ˇ
45 Jakovl¸evs A, Vanags A, Gardovskis J, Strumfa I. Proc Latv Acad Sci Sect B Nat Exact
12 Fouda MFR, Abd-EIzaher MM, Abdelsamaia RA, Labib AA. Appl Organomet Chem.
2007;21:613–625.
Appl Sci. 2020;74:318–324.
46 Murray GI, Mody N, Neel BG, Bence KK, Wilson HM, Delibegovic M. Cancer Res.
2018; 78: 75–87.
13 Wang R, Chen H, Yan W, Zheng M, Zhang T, Zhang Y. Eur J Med Chem. 2020;190,
112109.
47 Balius TE, Mukherjee S, Rizzo RC. J Comput Chem. 2011;32:2273–2289.
48 Maia EHB, Medaglia LR, Da Silva AM, Taranto AG. ACS Omega. 2020;5:6628–6640.
49 Nunes RR, Da Fonseca AL, Pinto ACDS, et al. Mem Inst Oswaldo Cruz. 2019;114:1–10.
50 Carregal AP, Maciel FV, Carregal JB, dos Reis SB, da Silva AM, Taranto AG. J Mol
Model. 2017;23:1–9.
14 Fiorina VJ, Dubois RJ, Brynes S. J Med Chem. 1978;21:393–395.
15 Guo Y, Wang SQ, Ding ZQ, Zhou J, Ruan BF. J Organomet Chem. 2017;851:150–159.
´
16 Skiba J, Karpowicz R, Szabo I, Therrien B, Kowalski K. J Organomet Chem. 2015;794:
216–222.
17 Hafez TS, Osman SA, Yosef HAA, et al. Sci Pharm. 2013;81:339–357.
18 Xie YS, Zhao HL, Su H, et al. Eur J Med Chem. 2010;45:210–218.
19 Claisen L, Claparbde A. Eur J Inorg Chem. 1881:2460–2468.
20 Atwal KS, Rovnyak GC, O’Reilly BC, Schwartz J. J Org Chem. 1989;54:5898–5907.
51 Trott O, Olson AJ. J Comput Chem. 2009;31:455–461.
52 Sander T, Freyss J, Von Korff M, Rufener C. J Chem Inf Model. 2015;55:460–473.
53 Fernando Perez, Brian E. Granger. 2018, Computing in Science and Engineering.
8