A Computational Approach in the Search of New Biologically
Active 9,10-Anthraquinone Derivatives
Maryna Stasevych, Viktor Zvarych and Volodymyr Novikov
Lviv Polytechnic National University, S. Bandera Str. 12, Lviv, 79013, Ukraine


                Abstract
                The results of using a computer approach in the search for new potential biologically active
                compounds in a series of 9,10-anthraquinone derivatives using free online programs PASS
                Online, CLC-Pred (Cell Line Cytotoxicity Predictor), Acute Rat Toxicity and determining
                the level of binding of the studied structures of anthraquinones with target proteins using
                the Schrodinger software package are generalized. The directions of experimental primary
                assessment of antimicrobial, antiplatelet, antioxidant, antiviral, anticonvulsant, antitumor
                action for selected objects of research are determined. Molecular docking shows the
                prospects for studies of the mechanisms of anticancer and antiplatelet agents.

                Keywords 1
                9,10-anthraquinone derivatives, in silico prediction, biological action

1. Introduction
    Despite the significant achievements of modern medical chemistry and pharmacology, the search
for new more effective and safer medicinal substances remains an actual problem [1]. The number of
biological activities studied by modern pharmacology is more than three thousand, and the number of
potential molecular targets of drugs is tens of thousands [2]. Experimental verification of tens / hundreds
of millions of chemical substances for thousands of types of pharmacological effect is practically not
implemented. The basis of modern search and development of new drugs is the data analysis of the
mechanisms of disease development, protein targets and compounds with pharmacological activity, the
effect of which allows to remove the pathological process [3].
    The structural formulas of molecules of studied substances using computer prediction allow to find
new biologically active compounds with necessary properties [4]. The most promising substances for
chemical synthesis are selected by researchers on the basis of in silico prediction and determine the
priorities of their experimental testing, which significantly reduces the cost of experimental research
and eliminates unpromising substances in the earliest stages of research.
    Computer investigations are widely used to analyze the relationships "structure - biological effect"
of organic compounds [5]. Search and designing materials with desired properties and optimization of
pharmacodynamic and pharmacokinetic characteristics of the basic structures of new biologically active
compounds is carried out using them. Most computer programs designed for this purpose are distributed
on a commercial basis by specialized firms (Accelrys, Tripos, ACD Labs, ChemSoft, etc.).
    There are a relatively small number of computer programs available for free over the Internet and
predicting pKa (http://vcclab.org/lab/alogps/start.html), some physicochemical properties
(http://www.molecularknowledge.com/Online/Estimation/online1.htm), solubility, lipophilicity, and
some types of biological activity (http://www.organic-chemistry.org/prog/peo/index.html,
http://www.molinspiration.com/cgi-bin/properties). In recent years, foreign web services have appeared
on the Internet that allow predicting the interaction of chemical compounds with target macromolecules

IDDM’ 2020: 3rd International Conference on Informatics & Data-Driven Medicine, November 19-21, 2020, Växjö, Sweden
EMAIL: maryna.v.stasevych@lpnu.ua (M. Stasevych); viktor.i.zvarych@lpnu.ua (V. Zvarych); volodymyr.p.novikov@lpnu.ua (V. Novikov)
ORCID: 0000-0001-5042-4133 (M. Stasevych); 0000-0003-3036-0050 (V. Zvarych); 0000-0002-0485-8720 (V. Novikov)
             ©️ 2020 Copyright for this paper by its authors.
             Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
             CEUR Workshop Proceedings (CEUR-WS.org)
based on structural similarity (http://bioinformatics.charite.de/superpred/, http://cpi.bio-x.cn/drar/). The
information available on these websites does not allow us to assess the quality of the prediction provided
by these web services.
    The computer program PASS Online [6] became the first of the free online services in the world,
which allows you to predict 5066 types of pharmacological effects based on the structure of the
molecule. Its effectiveness in finding new bioactive substances is constantly confirmed by the numerous
works of more than 23,000 researchers from more than 100 countries, and the training sample is updated
as new data on biologically active compounds for each type of biological activity. In recent years, PASS
Online has been added by a number of free online web services on the Way2Drug platform that predict
more than 4,000 types of biological activity, including acute toxicity to rats with four routes of
administration, effects on tumor and non-tumor cell lines interacting with antytargets and etc..
    Molecular docking (molecular modeling) is actively used to solve virtual screening problems [7].
Its essence is to model the relative position of the studied molecule and the target protein. The spatial
structure of the molecular target and the spatial structure of the ligand (studied structure) make it
possible to explain at the molecular level the mechanism of interaction of the ligand with the protein.
The docking program tests the studied structures using a special scoring function (affinity), which
roughly describes the energy of interaction of the molecule with the target protein. It is possible to reject
from further consideration a substance with poor values of the scoring function using the results of
docking. Modeling of ligand-receptor interactions is carried out using a variety of different software
packages (AutoDock, AMBER, eHiTS, Surflex-Dock, Schrödinger, etc.) [8-11], each of which has its
own advantages and disadvantages, including accessibility via the Internet.
    Considering the above, an in silico approach was used in the search for new derivatives of 9,10-
anthraquinone using the latest resources to determine the experimental directions of research on their
pharmacological activity.

2. Relates works
   The computer program PASS Online and a number of free online web services such as CLC-Pred
(Cell Line Cytotoxicity Predictor), Acute Rat Toxicity of the Way2Drug platform [12] are widely used
by the researchers from different countries for the search of new biologically active compounds among
synthetic and natural compounds and predicted results had and have many examples of experimental
verification. In particular, these programs have shown their effectiveness in the search for antimicrobial
substances [13, 14], determination of toxicity level [15], search of new anticancer drugs and evaluation
of their cytotoxicity [16, 17], etc. For 40 known natural anthraquinone derivatives PASS Online was
used for evaluation of antiviral potential against different types of viruses like Herpes, Hepatitis B and
C, Cytomegalovirus, Adenovirus, Hepatitis, HIV, Parainfluenza, Influenza, Picornas-, Pox-, Rhino-,
and Coronavirus (Covid-19) and molecular docking using SWISS-MODEL was carried out in [18]. In
work [19] molecular docking study of anthraquinone compounds extracted from Anethum sowa L. root
was used for estimation of the their pharmacological potential targeted towards anticancer activity.

3. Materials and methods
   In silico evaluation of the biological properties of new functionalized 9,10-anthraquinone derivatives
(Fig. 1) was performed by the online services PASS Online, CLC-Pred, Acute Rat Toxicity of web portal
Way2Drug. As an initial information, the structural formula of the substance in MOL or SDF file format
was used to obtain prediction data in each of the mentioned programs. The connection table containing
data of the molecule valent bonds and table of atoms types of the loaded structure of compound are
basis for generation of the set of multilevel neighborhoods of atoms (MNA) structure descriptors [6].
The Bayesian mathematical approach is used as an algorithm for estimation of results of predicted
biological activity [6]. This algorithm provides stable in the static sense of the dependence "structure-
activity" and, accordingly, the results of the forecast. The work of the online services is based on the
analysis of the "structure-activity" relationship for substances from the training set, which contains more
than 40,000 different biologically active substances (substances of known drugs and physiologically
active compounds), i.e. the result of prediction of biological activity is compared with known
experimental data [6]. A list of possible types of pharmacological effect shows two probabilities - the
presence of Pa activity and the absence of Pi activity as the result of prediction (Fig. 2-4). Pa values
range from 0.000 to 1.000. When Pa>0.7, the compound has a similar effect to the experimental one
and in this case the chance of this compound being an analog of a known pharmacological drug is very
high. If 0.5<Pa<0.7, the compound has a similar effect to the experimental one, but this probability is
less and the compound is not similar to the known pharmacological drug. When Pa<0.5, the compound
does not correspond to the experimental activity; however, the presence of this activity confirmed by
experimental data, can become a new object for investigation.
    CLC-Pred service allow to predict the cytotoxic action for non-transformed and cancer cell lines
based on loaded molecular structure in comparison with the training set obtained from ChEMBLdb
Database (https://www.ebi.ac.uk/chembldb/) [20]. Results of prediction using CLC-Pred are also
presented as a list of Pa/Pi values of probable cancer cell lines (Fig. 2). Acute Rat Toxicity online service
[21] allow in silico estimate probable level of toxicity using a rat model for four different routes of
administration: intravenous, intraperitoneal, oral, subcutaneous (Fig. 3). Results of prediction are
compared with the training sets of known compounds with determined experimental toxicity LD50
values        obtained          from        Toxicity        Database         of        SYMYX           MDL
(http://www.akosgmbh.de/accelrys/databases/symyx_dbs.htm). Output data show probable LD50 value
in log10 (mmol/kg) or mg/kg and class of acute rodent toxicity.
    The molecular modeling of affinity (binding) with the corresponding target protein site was used to
determine probable mechanisms of action using Small-Molecule Drug Discovery Suite of Schrödinger
within the test access [22] and AutoDock Vina [23] was carried out in addition to the above-mentioned
free online access programs for predicting the pharmacological effects of the studied structures. The
conformation of the ligand (studied chemical structure) that best interacts with the protein binding site
is the result of molecular modeling. The structure of biological targets for molecular modeling were
taken from database RCBS Protein DataBank (https://www.rcsb.org). For the molecular docking as
target proteins were selected: receptor proteins-tyrosine kinases cyclooxygenase-1 (COX-1) - 3N8X,
glycoprotein-IIb/IIIa (GPIIb /IIIa) - 2VDM, glycoprotein-VI (GP-VI) - 2G17, purine receptor P2Y12 -
4PXZ, prostacyclin receptor (PG-I2) - 4F8K, protein-activated receptor-1 (PAR-1) - 3VW7,
antithrombin III (ATIII), factor-X (FX), factor-II (F-II), factor -IX (F-IX) and vitamin K-epoxy
reductase (VKOR) - 1NQ9, 1KSN, 5JZY, 1RFN and 3KP9, c-Kit, B-Raf, EGFR (1NQL, 1IVO, 1M17,
2GS6) and PDGF (1T46, AKT1, ERK2), non-receptor tyrosine kinases SRC (1SKJ), nonspecific
tyrosine kinases ABL (3OXZ, 3QRJ, 2ABL). Then, the target protein is prepared using the Protein
Preparation Wizard with the removal of water molecules and the addition of missing hydrogen atoms
in the structure of the target protein and the optimization of the structure of protein using the PROBKA
subprogram. At the next stage, the protein structure is minimized by the OPLS-2005 subprogram.
Preparation of ligand (studied chemical structure) is performed using LigPrep Wizard. Next, the ligand
binding region to the receptor is generated using the Receptor Grid Generation module of the Glide
Maestro and docking is performed in the XP phase. The result of docking is a set of structures in 3D
with a set of numerical indicators describing their affinity Gscore (scoring function) to the active site of
the protein.
                                    Figure 1: Objects in silico study




 Figure 1: Example of prediction results         Figure 2: Example of prediction results of CLC-Pred
 of PASS Online




                     Figure 3: Example of prediction results of Acute Rat Toxicity

4. Results and Discussion
   Analysis of prediction data for all new structures of functionalized 9,10-anthraquinone derivatives
1-12 (Fig. 1) showed that the main, expected, direction of experimental studies is the investigation of
the antitumor effect (Pa in the range 0.3-0.6). Since the studied structures 1-12 are new objects for the
PASS Online training set, the Pa values for many structures are in the range 0.3-0.6. Taking into account
results of anticancer activity prediction, cytotoxicity was evaluated in silico against cancer cell lines
using the web service CLC-Pred [20]. The analysis of the obtained results showed that the most
probable effect of the studied compounds 1-12 (Pa=0.3-0.8) is on cancer cell lines of the lungs,
lymphoid tissue, glandular tissue of the breast, and for some compounds is on also tissues of the cervix,
brain, and colon.
    The introduction of new biophore fragments into the 9,10-anthraquinone molecule allowed to
expand the range of experimental in vitro and in vivo studies, in addition to the study of the antitumor
effect predicted using PASS Online, in the following areas. In particular, antimicrobial and antioxidant
activities were predicted for N-acylamino-9,10-anthraquinones 1, amino acid derivatives of 2-chloro-
N-(9,10-dioxy-9,10-dihydroanthracen-1-yl) acetamide 2 (Pa=0.3-0.6). 1,2,3-Substituted guanidine 3
would be interesting subjects to test for antianginal (Pa=0.3-0.4), anti-ischemic (Pa=0.3-0.35),
cardiotonic (Pa=0.3-0.45) effects. The direction of research on hypoglycemic activity would be
interesting for 5-arylidene derivatives 4 (Pa=0.3-0.35). Iminothiazoles 5, 1,2,4-triazoles 6 and tetrazoles
7 can be studied also for antimicrobial (Pa=0.3-0.4), anti-inflammatory (Pa=0.3), anti-allergic effects
(Pa=0.3-0.4) and effects on vascular processes (Pa=0.3-0.33). An additional area of testing for
pyrrolylantracenediones 8 is to determine the level of antibacterial and antifungal activity (Pa=0.4-0.5).
The antitumor activity of the dithiocarbamate derivatives 9, 10 was supplemented by antimicrobial,
antioxidant, antiplatelet, antiviral, anticonvulsant effects at Pa=0.3-0.55 according to Pa prediction data.
Predicted Pa values (Pa=0.3-0.4) for the dithiocarbamates 9, 10 and hydrazone derivatives 11, 12
showed perspective for their further investigations for antiviral properties.
    The analysis of the prediction results showed that, in the overwhelming majority, the Pa values for
the studied structures are in the range Pa=0.3-0.6. Taking into account the categorization of Pa values,
such low Pa values can be explained by the novelty of molecular structures for the PASS Online training
set. The predicted data became the basis for the experimental study of the above mentioned
antraquinone derivatives 1-12.
    The probable acute toxicity of LD50 using the online resource Acute Rat Toxicity using in silico rat model
with four different routes of administration: intraperitoneal, intravenous, oral and subcutaneous [21] for
promising compounds, selected by the experimental tests, allows to estimate the level probable toxicity of
anthaquinone derivatives 1-12. The results showed that the studied structures of compounds can be classified
as medium-, low- and non-toxic compounds.
    As outcome result of the molecular docking, the values of Gscore scoring functions were determined
and the anthraquinone derivatives with high, medium and low levels of binding to a specific target
protein PDGF (1T46), VKOR (3KP9), c-Kit and B-Raf associated with anticancer and antiplatelet
actions with Gscore in the range from -6.7 to -11.8 were found among the studied molecular structures 1-
12 [24]. Examples of visualization of binding of hit compounds among triazole 6, dithiocarbamate 9,10
and triazinone 7 anthraquinones in the corresponding active zone of the sites of previously mentioned
proteins are given below (Fig. 2-7).




 Figure 2: Visualization of the binding of the hit Figure 3: Visualization of the binding of the hit
 compound triazole anthraquinone in the active compound pyrrolidine dithiocarbamate in the
 zone of the protein 1T46 (Gscore= -10.7)          active zone of the protein 1T46 (Gscore= -10.92)
 Figure 4: Visualization of the binding of the hit         Figure 5: Visualization of the binding of the hit
 compound piperidine dithiocarbamate in the active         compound pyrrolidine dithiocarbamate in the
 zone of the protein 3KP9 (Gscore = -10.39)                active zone of the protein c-Kit (Gscore= -9.1)




 Figure 6: Visualization of the binding of the             Figure 7: Visualization of the binding of the hit
 hit compound anthratriazinone in the active               compound anthratriazinone in the active zone of
 zone of the protein c-Kit (Gscore = -11.8)                the protein B-Raf (Gscore = -11.3)

   It should be noted that the dithiocarbamate derivatives of anthraquinone of type 9,10, which had the
highest values of the scoring function Gscore= -10.92…-11.2 are promising objects for experimental
study of the mechanisms their antioxidant, anticancer and antiplatelet action.
   Therefore, the results of predicted by the online programs PASS Online, CLC-Pred, Acute Rat
Toxicity regarding the probable manifestation of antimicrobial, antiplatelet, antioxidant, antiviral,
anticonvulsant, antitumor actions and predicted values of acute toxicity, as well as molecular docking
data obtained by Small-Molecule Drug Discovery Suite Schrödinger and AutoDock Vina have been
verificated by experimental in vitro and in vivo investigations.

5. Conclusion
    In this article, we present the generalized results of using in silico tools in the search for new potential
pharmacologically active compounds in the series of 9,10-anthraquinone derivatives. The used
computer approach carried out by free online programs PASS Online, CLC-Pred, Acute Rat Toxicity
allowed to determine the directions of the experimental primary assessment of antimicrobial,
antiplatelet, antioxidant, antiviral, anticonvulsant, antitumor action for the selected objects of study.
Molecular docking data on biotargets of pathological processes allowed to define compounds with a
high degree of affinity and showed the prospect of studying the mechanisms of anticancer and
antiplatelet agents from this class of compounds by modifying the ligand with various pharmacophore
fragments. The results of in silico approach allowed experimentally identify promising compounds in
the series of 9,10-anthraquinone.

6. Acknowledgements
   This work (Project number: 0119U002252) is supported by the Ministry of Education and Science
of Ukraine.
7. References
[1]    S. P. Elliott, T. Holdbrook, S. D. Brandt, Prodrugs of New Psychoactive Substances (NPS): A New
       Challenge, J. Forensic Sci. 65 (2020) 913–920. doi:10.1111/1556-4029.14268.
[2]    K. M. Reznikov, Paradigm of modern pharmacology: development and current approaches,
       Research result: pharmacology and clinical pharmacology 2 (2016) 107–114. doi:
       10.18413/2500-235X-2016-2-3-107-114.
[3]    S. Marshall, R. Madabushi, E. Manolis, K. Krudys, A. Staab, K. Dykstra, S. A. G. Visser, Model‐
       Informed Drug Discovery and Development: Current Industry Good Practice and Regulatory
       Expectations and Future Perspectives, CPT: Pharmacometrics & Systems Pharmacology 8
       (2019) 87–96. doi:10.1002/psp4.12372.
[4]    S. M. Ivanov, A. A. Lagunin, D. A. Filimonov, V. V. Poroikov, Computer prediction of adverse
       drug effects on the cardiovascular system, Pharm. Chem. J. 52 (2018) 758–762. doi:
       10.1007/s11094-018-1895-1.
[5]    S. P. Leelananda, S. Lindert, Computational methods in drug discovery, Beilstein J. Org. Chem.
       12 (2016) 2694–718. doi:10.3762/bjoc.12.267.
[6]    PASS Online Home Page, 2020. URL: https://www.way2drug.com/passonline.
[7]    X. Y. Meng, H. X. Zhang, M. Mezei, M. Cui, Molecular docking: a powerful approach for
       structure-based drug discovery, Curr. Comput. Aided Drug Des. 7 (2011) 146–157.
       doi:10.2174/157340911795677602
[8]    G. M. Morris, R. Huey, W. Lindstrom, M. F. Sanner, R. K. Belew, D. S. Goodsell, A. J.
       Olson, Autodock4 and AutoDockTools4: automated docking with selective receptor flexiblity, J.
       Computational Chem. 16 (2009) 2785–2791. doi: 10.1002/jcc.21256.
[9]    R. Salomon-Ferrer, D. A. Case, R. C. Walker, An overview of the Amber biomolecular
       simulation package, WIREs Comput. Mol. Sci. 3 (2012) 198–210. doi: 10.1002/wcms.1121.
[10]   J. L. de Almeida Filho, J. H. Fernandez MDR SurFlexDock: A Semi-automatic Webserver for
       Discrete Receptor-Ensemble Docking, in: L. Kowada, D. de Oliveira (Eds.), Advances in
       Bioinformatics and Computational Biology. BSB 2019. Lecture Notes in Computer Science,
       Springer, Cham, 2020, pp. 3-128.
[11]   R. Spitzer, A.N. Jain, Surflex-Dock: Docking benchmarks and real-world application, J. Comput.
       Aided Mol. Des. 26 (2012) 687–699. doi: 10.1007/s10822-011-9533-y.
[12]   Way2Drug Predictive Service Home Page, 2020. URL: http://www.way2drug.com.
[13]   S. Murugavel, S. Deepa, C. Ravikumar, P. Alagusundaram, Synthesis, structural, spectral and
       antibacterial activity of 3,3a,4,5-tetrahydro-2H-benzo[g]indazole fused carbothioamide
       derivatives as antibacterial agents, J. Mol. Struct., 1222 (2020), 128961.
       doi:10.1016/j.molstruc.2020.128961.
[14]   M. E. Khalifa, Synthesis and Evaluation of New 2‐Mercaptomethyl Benzimidazole Scaffolds as
       Potential Antibacterial, Antioxidant and Cytotoxic Agents, ChemistrySelect, 5 (2020), 10562.
       10.1002/slct.202002822.
[15]   F. Agahi, C. Juan, G. Font, A. Juan-García, In silico methods for metabolomic and toxicity
       prediction of zearalenone, α-zearalenone and β-zearalenon, Food Chem. Toxicol. 146 (2020),
       111818. doi: 10.1016/j.fct.2020.111818.
[16]   O.B. Kazakova, G. V. Giniyatullina, A. G. Mustafin, D. A. Babkov, E. V. Sokolova, A. A.
       Spasov, Evaluation of Cytotoxicity and α-Glucosidase Inhibitory Activity of Amide and
       Polyamino-Derivatives of Lupane Triterpenoids. Molecules, 25 (2020), 48334. doi:
       10.3390/molecules25204833.
[17]   K. Kurosaki, R. Wu, Y. Uesawa, A Toxicity Prediction Tool for Potential Agonist/Antagonist
       Activities in Molecular Initiating Events Based on Chemical Structures, Int. J. Mol. Sci., 21
       (2020), 7853. doi: 10.3390/ijms21217853.
[18]   P. Khanal, B. M. Patil, J. Chand, Y. Naaz, Anthraquinone Derivatives as an Immune Booster and
       their Therapeutic Option Against COVID-19. Nat. Prod. Bioprospect. 10 (2020), 325–335. doi:
       10.1007/s13659-020-00260-2.
[19] M. M. Saleh-E-In, A. Roy, M. A. Al-Mansur, C. Mahmood Hasan, M. M. Rahim, N. Sultana, S.
     Ahmed., M. R. Islam, J. van Staden, Isolation and in silico prediction of potential drug-like
     compounds from Anethum sowa L. root extracts targeted towards cancer therapy, Computational
     Biology and Chemistry, 78 (2019), 242–259. doi: 10.1016/j.compbiolchem.2018.11.025.
[20] A. A. Lagunin, V. I. Dubovskaja, A. V. Rudik, P. V. Pogodin, D. S. Druzhilovskiy, T. A.
     Gloriozova, D. A. Filimonov, N. G. Sastry, V. V. Poroikov, CLC-Pred: A freely available web-
     service for in silico prediction of human cell line cytotoxicity for drug-like compounds, PLOS
     One 13 (2018) e0191838. doi: 10.1371/journal.pone.0191838.
[21] A. Lagunin, A. Zakharov, D. Filimonov, V. Poroikov, QSAR modelling of rat acute toxicity on
     the basis of PASS prediction, Mol. Inf. 30 (2011) 241–250. doi: 10.1002/minf.201000151.
[22] J. Bhachoo, T. Beuming, Investigating Protein–Peptide Interactions Using the Schrödinger
     Computational Suite, in: Schueler-Furman O., London N. (Eds.) Modeling Peptide-Protein
     Interactions. Methods in Molecular Biology, Humana Press, New York, NY, 2017, pp. 3-68.
[23] O. Trott, A. J. Olson, Autodock Vina: improving the speed and accuracy of docking with a new
     scoring function, efficient optimization and multithreading, J. Comput. Chem. 31 (2010), 455–
     461. doi: 10.1002/jcc.21334.
[24] M. Stasevych, V. Zvarych, V. Lunin, T. Halenova, O. Savchuk, O. Dudchak, M. Vovk, V.
     Novikov, Novel anthraquinone-based derivatives as potent inhibitors for receptor tyrosine J.
     kinases, Indian J. Pharm. Sci. 77 (2015) 634–637. doi: 10.4103/0250-474X.169062.