Multivariate assessment of anticancer oleanane triterpenoids lipophilicity

Journal of Chromatography A 1656 (2021) 462552

Contents lists available at ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

Multivariate assessment of anticancer oleanane triterpenoids
lipophilicity
Monika Pastewska a, Barbara Bednarczyk-Cwynar b, Strahinja Kovacˇ evic´ c, Natalia Buławska a,
d
Szymon Ulenberg d, Paweł Georgiev d, Hanna Kapica a, Piotr Kawczak d, Tomasz Baczek
˛
,
a
a,∗
Wiesław Sawicki , Krzesimir Ciura
a

´ sk, Al. Gen. Hallera 107, 80-416 Gdan
´ sk, Poland
Department of Physical Chemistry, Medical University of Gdan
´ , Poland
Department of Organic Chemistry, Poznan University of Medical Science, Grunwaldzka 6, 60-780 Poznan
University of Novi Sad, Faculty of Technology Novi Sad, Department of Applied and Engineering Chemistry, Bulevar cara Lazara 1, 21000 Novi Sad, Serbia
d
´ sk, Al. Gen. Hallera 107, 80-416 Gdan
´ sk, Poland
Department of Pharmaceutical Chemistry, Medical University of Gdan
b
c

a r t i c l e

i n f o

Article history:
Received 8 June 2021
Revised 9 September 2021
Accepted 10 September 2021
Available online 15 September 2021
Keywords:
Oleanane triterpenoids
Lipophilicity
Chemometrics
RP-HPLC
IAM

a b s t r a c t
Naturally occurring molecules are excellent sources of lead compounds. A series of oleanolic acid (OA)
derivatives previously synthesized in our laboratory, which show promising antitumor activity, have been
analyzed in terms of lipophilicity evaluation applying chromatographic and computational approaches.
Retention data obtained on three reversed-phase liquid chromatography stationary phases (RP-HPLC) and
immobilized artificial membrane chromatography (IAM-HPLC) were compared with computational methods using chemometric tools such as cluster analysis, principal component analysis and sum of ranking
differences. To investigate the molecular mechanism of retention quantitive structure retention relationship analysis was performed, based on the genetic algorithm coupled with multiple linear regression
(GA-MLR). The obtained results suggested that the ionization potential of studied molecules significantly
affects their retention in classical RP-HPLC. In IAM-HPLC additionally, polarizability-related descriptors
also play an essential role in that process. The lipophilicity indices comparison shows significant differences between the computational lipophilicity and chromatographically determined ones.
© 2021 The Author(s). Published by Elsevier B.V.
This is an open access article under the CC BY license ( />
1. Introduction

Optimizing lipophilicity is an essential process in drug discovery since it noticeably affects the diffusion of molecules
through a biological membrane. Consequently, lipophilicity determines pharmacokinetic processes, including absorption, distribution, metabolism, excretion (ADME), as well as the toxicity of drug
candidates [1–3]. According to the International Union of Pure and
Applied Chemistry (IUPAC) definition, lipophilicity represents the
affinity of a molecule or a moiety towards a lipophilic environment. Lipophilicity is typically determined by solute distribution in
biphasic liquid-liquid or solid-liquid systems [4]. The first protocol
for lipophilicity assessment was initially proposed by Hansh and
based on the shake-flask procedure, which enables the determination of the partition coefficient of the target compound between
n-octanol and water (logP) [5].

∗

Corresponding author.
E-mail address: (K. Ciura).

Nevertheless, this method has many disadvantages - it is effortful, time-consuming, and requires large amounts of organic solvent
and pure substances. For these reasons, methods based on solidliquid partitioning, such as reversed-phase liquid chromatography
(RP-LC), are currently used for lipophilicity estimation. Both IUPAC and organisation for Economic Co-operation and Development
(OECD) consider RP-LC equivalent to the shake-flask approach. It
is worth emphasizing that other separation techniques, like micellar liquid chromatography (MLC), micellar electrokinetic chromatography (MEKC), or microemulsion electrokinetic chromatography (MEEKC), can also be used for lipophilicity value designation. Separation methods owe their popularity to numerous advantages. First of all, they are reproducible, easy to automate, rapid,
and require small amounts of analytes that don’t need to be absolutely pure because their impurities are readily separated during
the chromatographic process. Consequently, the separation methods have become the primary approach to lipophilicity assessment and easily fit into a high-throughput technique. As a property affecting the biological activity, chromatographically determined lipophilicity can then be processed using the Quantitative

/>0021-9673/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license ( />

M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

Journal of Chromatography A 1656 (2021) 462552

Structure-Retention Relationship (QSRR) approach [6]. QSRR is a

valuable support in predicting the behavior of new molecular entities and can provide insight into molecular mechanisms.
Another way for lipophilicity assessment are the theoretical
methods that provide quick information about the lipophilic properties of target molecules. The calculation approach enables the estimation of lipophilicity during the design of drug candidates before their synthesis. Several programs dedicated to logP calculation are plain, freely available online, and generate results nearly
instantaneously. Mentioned in silico approaches are significantly
faster and cheaper when compared to experimental methods. They
require no need for laboratory experiments, specialized equipment,
and any chemical reagents. Nevertheless, it is worth keeping in
mind that various studies noted significant differences between the
calculated logP and experimental data for some chemicals [2].
Biomimetic chromatography with an immobilized artificial
membrane (IAM) can be used to assess affinity for phospholipids
due to the presence of a stationary phase that contains the phosphatidylcholine head group. As a consequence, IAM-HPLC can emulate the lipid membrane monolayer [4]. This affinity is essential
for the drug molecule to achieve its therapeutic target (e.g., receptor) and to produce therapeutic effects. IAM-HPLC links the advantages of HPLC, such as rapid analysis and automatization, with
more biosimilar nature of stationary phase. Therefore, it is frequently used by both academia and pharmaceutical companies.
Terpenoids are a large and structurally diverse family of natural compounds [7]. Their structure is based on the five-carbon 2methyl-1,3-butadiene group, also called isoprene units. These compounds can be divided into several classes depending on the connection pattern of isoprene units, which is related to the quantity of those moieties [7]. Common classes that we can distinguish
are oleanes, lupanes, and ursanes. In the oleane group the most
important is oleanolic acid (OA), which shows various pharmacological effects [8], such as anti-inflammatory [9], spasmolytic [10],
hepatoprotective [11], antiviral [12], antiallergic [13], hypoglycemic
[14,15], cytoprotective and primarily antitumorigenic potential [7].
On account of simply altering the chemical structure and tremendous potential in activity at various stages of tumor development,
OA is a perfect hit molecule [7,16,17].
A library of OA derivatives synthesized in our laboratory
shows various anticancer activities [16–23]. The target compounds
present a higher level of anticancer activity towards breast cancer MCF-7, cervical cancer KB (HeLa) [19], Cellosaurus CCRF-CEM,
CCRF-VCR10 0 0, CCRF-ADR50 0 0, and human leukemia HL-60, SR
[18,24] cell lines [17].
The main aim of this study was to assess the lipophilicity indices of previously synthesized OA derivatives from the retention
behavior on three different RP-HPLC stationary phases and IAMHPLC and compare the experimental indices with those obtained
through computational methods using a chemometric approach.
Additionally, our investigation has two secondary goals. The first of

them was the selection of the OA structures which should exhibit
optimal lipophilicity, high antitumor activity, and the best pharmacokinetic properties for further studies. In order to realize this purpose, investigated molecules were also characterized using SwissADME software. The second goal focuses on exploring the molecular mechanism of retention in RP-HPLC using the QSRR approach.

dissolved in DMSO (1 mg/ml) except for compound 18, dissolved in
THF (1 mg/ml). Each stock solution of analytes was stored at 2–8
°C. Dilutions of the compounds (100 μg/mL) were made just before
analysis.
2.1.2. The analytical standards
In order to carry out chromatographic hydrophobicity index of IAM (CHIIAM ) determination, the model substances were
used. The analytical standards of octanonophenone, butyrophenone, and acetanilide were provided by Alfa Aesar (Haverhill, MA,
USA); acetaminophen, acetophenone, were purchased from SigmaAldrich (Steinheim, Germany); heptanophenone, hexanophenone,
valerophenone, propiophenone, and acetophenone were bought
from Acros Organic (Massachusetts, United States).
2.1.3. Reagents
Ultrapure water (18.2 M × cm−1 ) used to prepare the mobile phase was purified and deionized in our laboratory via a Millipore Direct-Q 3 UV Water Purification System (Millipore Corporation, Bedford, MA, USA). All the analytical reagents were used
without prior purification. Disodium phosphate (Na2 HPO4 ) and
monosodium phosphate (NaH2 PO4 ) were supplied by POCH (Gliwice, Poland). Dimethyl sulfoxide (DMSO), used as a solvent, was
from Merck (Darmstadt, Germany). Acetonitrile (LiChrosolv®) and
tetrahydrofuran (gradient grade for liquid chromatography) were
purchased from Sigma-Aldrich (Steinheim, Germany).
Metabolic stability was performed using the following reagents:
pooled human liver microsomes and sodium salt of NADPH were
obtained from Sigma (Sigma-Aldrich, Saint-Louis, MO, USA), while
monopotassium phosphate and dipotassium phosphate were from
POCH (POCH, Gliwice, Poland).
2.2. Chromatographic analysis
All HPLC analyses were performed on Prominence-1 LC-2030C
3D HPLC system (Shimadzu, Japan) equipped with DAD detector and controlled by LabSolution system (version 5.90 Shimadzu,
Japan). In all cases, the concentrations of the investigated analytes
were 100 μg/mL (in DMSO or THF), and the injected volume was

20 μL. In each system phase A was 10 mM phosphate buffer, and
phase B was acetonitrile (ACN). Hence, retention times (tR ) of investigated triterpenoids were collected, and their detection in all
systems was performed at 200 nm. The temperature, ACN concentration, pH of the buffer, columns suppliers, and dead times, together with the flow for each chromatographic system used in the
study, are given below.
•

•

•

2. Materials and methods
2.1. Materials
•

2.1.1. Oleanolic acid derivatives
The 2D structures and SMILES notation of the target OA derivatives are presented in Table S1, whereas their synthesis and characterization were described in the literature [18]. All samples were
2

CN chromatography: 10 mM phosphate buffer at pH 7.4; the
flow rate 1.5 mL/min. The chromatography was carried out at
40 °C on Agilent SB-CN column (4.6 mm x 150 mm x 3,5 μm;
Zorbax; USA, dead time = 1.22 min) with a linear gradient
phase B 50–100%.
C18 chromatography: 10 mM phosphate buffer at pH 7.4; the
flow rate 1.5 mL/min. The chromatography was carried out at
40 °C on Symmetry C18 column (3.9 × 150 mm x 5 μm; Waters;
USA, dead time = 0.847 min) with a linear gradient phase B
70–100%.
Phenyl (Ph) chromatography: 10 mM phosphate buffer at pH
7.4; the flow rate 0.2 mL/min. The chromatography was carried

out at 30 °C on Unison UK-Phenyl column (2 mm x 150 mm x
3 μm; Imtakt; USA dead time = 2.461 min) with a linear gradient phase B 30–100%.
IAM chromatography: 10 mM phosphate buffer at pH 7.4; the
flow rate 1.5 mL/min. The chromatography was carried out
at 30 °C on IAM.PC.DD2 column (10 × 4.6 mm x 10.0 μm;
Regis Technologies; USA), additionally equipped with IAM guard
column


M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

Journal of Chromatography A 1656 (2021) 462552

In CN, C18 , and Ph chromatography, two gradient runs were applied during experiments differing in gradient time (tG equal to
20 min and 40 min). According to the assumption proposed by
Snyder and co-workers [25,26], appropriate logkw values (i.e., the
retention factor logk extrapolated to 0% organic modifier, as an
alternative to logP) were calculated. DryLab 6.0 software (Molnar
Institute, Berlin, Germany) was used to make this computation.
Dwell volume for these HPLC systems was measured at 0.780 mL.
The IAM-HPLC analyses were carried out in one gradient run and
the analysis time was 6.5 min. The studied compounds’ CHIIAM indices were obtained using a calibration set of reference substances
using the protocol proposed by Valko and co-workers [27]. Each
HPLC analysis was run in triplicate. The obtained retention times
for the target OA derivatives are summarized in Table S2, S3 and
S4.

2.4.2. Sum of ranking differences (SRD) analysis
In the present study, the SRD approach was carried out in order to rank and select the best lipophilicity measures of oleanane
triterpenoids obtained by both in silico and experimental (chromatographic) methods. The benchmark was defined as row-average

(consensus approach) calculated on the basis of all data in one row.
The variables used in the SRD analysis were normalized by minmax normalization method and scaled in the range between 0.01
and 0.99. The SRD results were validated by comparison of rank by
random numbers (CRRN) and 7-fold cross-validation method [54].
The cross-validation was carried out by omitting approximately 1/7
of objects and by performing the ranking on the remaining objects
[55]. The SRD values were normalized (SRD%) in order to easily
compare the results of different SRD analyses.
2.4.3. QSRR calculation
QSRR analysis was done using QSARINS 2.2.4 version software
developed by Gramatica et al. [27,28]. Descriptors selection was
supported by a genetic algorithm (GA), whereas multiple regression was employed as a regression method. The set of parameters
applied to control GA was the size of the population—100 and the
mutation rate—25%. The analyzed OA derivatives were divided into
two groups, the training group (n = 25, ≈ 76%) and the testing
group (n = 8, ≈24%, compounds no: 12, 19, 20, 22, 23, 24, 27,
31). The split was random prior to QSRR analysis. The model fitting, robustness, and predictive abilities were assessed by the coefficient of determination (R2 ), predictive squared correlation coefficient (Q2 ), and root-mean-squared error of cross-validation (RMSECV) coming from the leave-one-out cross-validation technique.
Furthermore, root-mean-square error in prediction (RMSEP) deriving from external validation was calculated.

2.3. In-silico calculation
2.3.1. Theoretical physicochemical descriptors
Models were prepared as .mol files using ACD Chemsketch (Advanced Chemistry Development, Inc., Toronto, Canada),
and afterward converted into Gamess [25] input files using Open Babel software (The Open Babel Package, version
2.3.1 ; accessed Oct 2011) [26]. Geometrical
optimization of 2D models was performed on DFT level of theory,
using B3LYP/6–311 G(d) parameters. Optimized compounds were
afterward subjected to Dragon 7.0 (Talete, Milano, Italy) software
to calculate physicochemical descriptors used later to develop presented models.
2.3.2. In-silico calculation of lipophilicity
Several software packages were used for lipophilicity calculations, whereas each is based on different algorithms. Four different

logP values (ILOGP, WLOGP, Silicos-IT LogP, Consensus LogP) were
calculated using a virtual logP calculator available online: http:
//www.swissadme.ch (developed and maintained by the Molecular
Modeling Group of the Swiss Institute of Bioinformatic). KOWWIN
logP values were obtained using KOWWIN v. 1.68 software (EPI
Suite package v.4.2, U.S. EPA). XLOGP3, MLOGP, AlogP were attained
with Virtual Computational Chemistry Laboratory (VCCLAB, http://
www.vcclab.org/). ACD/ChemSketch logP was derived from ChemSketch software (version 12.1.0.31258; ACD/Labs, Toronto, Canada).

3. Results and discussion
3.1. Lipophilicity estimation by computational methods
Nowadays, computer-aided drug design (CADD) plays an essential role in the drug discovery and development process. Among
huge numbers of theoretical descriptors, lipophilicity remains the
most important one. As a consequence, many computer programs
can estimate the logP values based on various algorithms. Generally, the computational approaches for lipophilicity determination have several advantages over the experimental procedures.
First, the calculated logP indices may be obtained before synthesis
and supported the design process of new derivatives with desired
lipophilicity. Another benefit is that computational methods can
save time and chemical reagents, making it very attractive from
the economic and environmental points of view.
Computational logP values studied compounds estimated by
means of 9 different software are summarized in Table 1, whereas
the classification of the investigated software based on algorithm
type is presented in Table 2. The 2D structures and SMILES codes
of target solutes are listed in Table S1.
The logP values calculated by iLogP are significantly lower for
each tested substance. It is worth emphasizing that the computed
value of logP is almost twice lower than that of other programs.
iLogP platform is based on calculating solvation free energy, which
is a relatively new approach for lipophilicity calculation [29]. On

the contrary, the highest values are obtained for ACD/LogP. This
software implemented a classic algorithm that is based on the
principle of isolating carbons. It should be highlighted that the
same chemical structure features completely different logP parameters. For example, obtained value iLOGP for molecule 4 is
equal 5.09, whereas attained ACD/LogP quantity is 10.42, while
the lipophilicity for this compound, calculated using programs
based on hybrid and fragmentary algorithms, is about 6–7. Among

2.3.1. . In-silico calculation of biological properties
SwissADME data were obtained using a web-based application
available online: (developed and maintained by the Molecular Modeling Group of the Swiss Institute of
Bioinformatic). Input structures for both calculations were generated on the basis of optimized structures as a set of mol files. The
calculated pharmacokinetic properties are summarized in Table S5.
2.4. Data analysis
2.4.1. Cluster analysis and principal component analysis
Cluster Analysis (CA) and Principal Component Analysis (PCA)
were applied for databases which included the obtained chromatographic lipophilicity data and the calculated logP using STATISTICA
13.3 (StatSoft, Tulsa, Oklahoma, USA). Before analysis, data were
standardized to eliminate the impact of different scales by using
the Z-score scaling algorithm (V = mean of V/δ , where V is the
value of variables and δ is the standard deviation). CA has been
carried out using Ward’s amalgamation rule and the Euclidian distance measure. For the purpose of clustering lipophilicity measures
based on the sum of ranking differences, the double dendrograms
in clustered heat maps were obtained using NCSS 2021 Statistical
Software (NCSS, LLC. Kaysville, Utah, USA, ncss.com/software/ncss.)
applying Ward’s algorithm and Euclidean distances.
3


M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

Journal of Chromatography A 1656 (2021) 462552

Table 1
The calculated logP values of the OA derivatives with respect to the computational model.
No.

iLogP

XLogP3

WLogP

MLogP

Silicos-IT LogP

Consensus LogP

ALogP

LogP KOWWIN

ACD/ChemSketch logP

1
2
3
4
5

6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33

3.94

3.96
4.67
5.09
4.57
4.52
4.38
4.76
4.54
4.44
4.21
4.20
4.04
4.46
4.20
3.77
3.79
3.63
3.44
4.48
4.32
4.19
4.28
4.03
4.38
4.20
5.06
4.89
4.66
4.73
4.54

5.02
4.91

7.49
8.06
7.81
8.39
7.5
7.75
6.57
7.17
8.07
7.87
7.56
7.81
6.63
7.23
8.13
7.17
7.43
6.24
7.74
7.32
6.75
6.43
6.69
5.50
6.10
7.00
7.62

7.05
6.73
6.99
5.80
6.40
7.30

7.23
7.80
7.32
7.89
7.53
7.79
6.45
7.25
8.02
7.28
7.49
7.75
6.41
7.21
7.98
7.44
7.70
6.36
7.93
7.07
6.50
6.71
6.97

5.63
6.43
7.20
7.20
6.63
6.84
7.10
5.75
6.55
7.33

5.82
6.06
6.01
6.25
5.92
5.88
5.47
5.92
5.82
6.21
6.11
6.07
5.66
6.11
6.01
5.73
5.69
5.28
5.63

5.33
5.06
4.97
4.96
4.55
4.97
4.89
5.56
5.31
5.22
5.20
4.79
5.22
5.12

5.85
6.36
6.41
6.92
6.99
6.70
6.27
7.72
7.32
6.58
7.17
6.87
6.44
7.90
7.50

6.44
6.14
5.71
6.76
6.63
6.10
6.69
6.39
5.96
7.42
7.03
6.76
6.24
6.83
6.53
6.10
7.55
7.18

6.07
6.45
6.44
6.91
6.50
6.53
5.83
6.56
6.76
6.48
6.51

6.54
5.84
6.58
6.76
6.11
6.15
5.44
6.30
6.17
5.75
5.80
5.86
5.14
5.86
6.06
6.44
6.02
6.05
6.11
5.40
6.15
6.37

6.42
6.80
6.67
7.05
6.63
6.72
5.77

6.67
7.17
6.65
6.61
6.70
5.75
6.65
7.15
6.38
6.47
5.52
6.92
6.12
5.74
5.70
5.79
4.84
5.74
6.24
6.29
5.92
5.88
5.97
5.02
5.92
6.42

7.95
8.96
8.24

9.25
7.73
7.52
7.27
7.15
9.14
8.27
8.10
7.54
7.64
7.52
8.48
7.44
7.23
6.98
8.85
8.19
7.18
6.67
6.46
6.21
6.10
8.08
8.00
7.00
6.48
6.27
6.02
5.91
7.89

9.06
9.96
9.52
10.42
8.94
9.30
7.86
8.73
9.45
7.94
7.37
7.73
6.28
7.15
7.87
8.48
8.84
7.39
8.98
8.14
7.25
7.01
7.37
5.92
6.79
7.40
9.92
9.02
8.44

8.80
7.36
8.23
8.95

Table 2
List of software used with information regarding algorithms and suppliers.
No

softwere

Algorhitms

Supplier

1

iLogP



2
3
4

XLogP3
WLogP
MLogP

5

6
7
8
9

Silicos-IT LogP
Consensus LogP
ALogP
KOWWINlogP
ACD/ChemSketch logP

Physics-based method relying on free energies of solvation in n-octanol and water
calculated by the Generalized-Born and solvent accessible surface area (GB/SA) model
An atomistic method including corrective factors and knowledge-based library
Purely atomistic method based on the fragmental system of Wildman and Crippen
An archetype of topological method relying on a linear relationship with 13 molecular
descriptors
An hybrid method relying on 27 fragments and 7 topological descriptors
Arithmetic mean of the values predicted by the five proposed methods
Ghose-Crippen octanol-water partition coeff. (logP); based on molecular properties
Atom-based approach and fragmental contribution
Based on the principle of isolating carbons

tested derivatives 4 and 32 have the greatest theoretical lipophilicity, especially when compared to parent OA structures. Generally,
the newly synthesized structures demonstrate similar lipophilicity properties as OA. Slightly more hydrophilic are molecules 16,
17, 18, 25, and 31. All tested substances represented relatively high
lipophilicity. Referring to the Lipinski’s rule of five [30], it should
be noted that the given compounds do not exceed the critical value
of lipophilicity (logP<5) only in the case of iLogP software.

/>





/>
Fig. 1, the BOILED-Egg scheme of target molecules is presented.
The BOILED-Egg scheme refers to passive penetration through the
gastro-intestinal wall and drug brain permeability. Although conceptually, BOILED-Egg model is very simple, it is based only on two
physicochemical descriptors (WLOGP and TPSA) Daina and Zoete
proved their excellent classification skills [32].
When we analyze pharmacokinetic parameters determined by
computational methods, some conclusions can be drawn. According to the scheme presented in Fig. 1 it can be concluded that target molecules have not reached the central nervous system (CNS),
which may reduce their side effects. On the other hand, this data
excludes the possibility of using these derivatives to treat tumors
located in the CNS. Another essential point in an early stage of
drug development is gastro-intestinal absorption. Due to the low
cost and simple manufacturing technology, oral administration is
the first choice of drug administration. Additionally, drugs administrated orally are widely accepted by patients. Even though most
investigated OA derivatives will not be absorbed orally, nine of the
studied derivatives can be considered medicaments for oral admin-

3.2. Pharmacokinetics properties estimation by SwissADME
The next step of our investigation focused on calculating the
pharmacokinetic properties of target solutes using SwissADME
software. Among the available tools for CADD modeling SwissADME is one of the newest programs [31]. This software uses support vector machine (SVM) or Bayesian techniques in its calculation algorithms. Other advantages of this tool are the user-friendly
submission system and easy analysis of the results [31]. The obtained SwissADME data are summarized in Table S5, whereas in
4

M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

Journal of Chromatography A 1656 (2021) 462552

Fig. 1. BOILED-Egg scheme of target molecules which refer to passive penetration through gastro-intestinal wall, BBB and actively effluxed by P-gp (PGP+ red; PGP−, blue).

istration. If we analyze the metabolic stability of tested OA derivatives according to SwissADME calculation, almost every derivative
does not exhibit inhibition against Cytochrome P450. Considering
that metabolic stability plays a crucial role in drug candidate’s
safety and efficacy [4], such calculations are vital for the proper
selection of the best drug candidates. In addition, prediction of being substrate or non-substrate of the permeability glycoprotein (Pgp) has been performed. P-gp is the essential member among ATPbinding cassette transporters and plays a critical role in active efflux through biological membranes, including the gastro-intestinal
wall to the BBB or from the BBB [33]. Only two derivatives, no 23
and 24, can be recognized as P-gp substrates among tested compounds. Looking for other calculated properties of OA derivatives,
some indices are intercorrelated with calculated lipophilicity. Still,
we must remember that several of these processes depends on
molecules lipophilicity, like skin permeation. The poor water solubility of these derivatives should be recognized as a point of concern, and it might reduce their possible exploitation as therapeutic agents. Nevertheless, numerous poor water solubility or practically insolubility substances are used as medicines, including antitumoral cyclosporine [34]. This example indicated that poor water
solubility could be overcome by selecting the appropriate technological form for the finished product.

ods are currently mainly applied. In RP-HPLC, which is one of the
most popular modes of liquid chromatography, lipophilicity governs the retention of molecules [2,5,35]. Automation, rapid analysis, and lower costs compared to the shake-flask method are the
main advantages of RP-HPLC.
Furthermore, in the case of the investigated group, its lipophilic
character limited the use of the shake-flask method. According to
the OECD guide (test no 107), the logP should range between −2
and 4, whereas preliminary experiments and computational logP
showed a more lipophilic character.
Generally, chromatographic sorbent activity depends on various
parameters, including specific surface area, the density of the free
active centers per unit of sorbent surface area, the energy of intermolecular interactions between solutes and a used type of sorbent active centers, as well as chemical structures of the sorbent

[36]. Although RP-HPLC is a well-recognized tool for lipophilicity
assessment, there is no universal protocol effective for every chemical group of compounds. For this reason, three different reversedphase stationary phases (C18 , Phenyl, Cyanopropyl) and IAM column were investigated during this study. The obtained lipophilicity indices are presented in Table 3, whereas representative chromatograms are shown in Fig. 2. Traditionally, the lipophilicity index
measured by RP-HPLC is calculated based on the logk determined
by several isocratic experiments and express as extrapolated value
to pure water (logkw ) according to Soczewinski –Wachtmeister
equation [37].
Nevertheless, nowadays, it is possible to reduce the number of
chromatographic experiments and base the estimation on two gradient measurements using the DryLab software, which determines
the logkw indices according to the assumption proposed by Sny-

3.3. Chromatographically determined lipophilicity
In parallel to computational methodologies, indirect approaches
(mostly chromatographic) have been used for lipophilicity assessments. Although the traditional way to lipophilicity assessment is
based on the shake-flask procedure, the chromatographic meth5


M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

Journal of Chromatography A 1656 (2021) 462552

Fig. 2. Representative chromatograms for molecule 23 achieved in investigated HPLC systems: A) C18 B) CN C) IAM D) Ph. The detailed experimental conditions were reported
in Section 2.3.

der and co-workers [40,41]. Retention indices of IAM-HPLC can also
be determined in various ways. In our study, we decided to apply
protocols proposed by Valko et al., which are based on chromatographic hydrophobicity indices of IAM (CHIIAM ) since this procedure was validated for a thousand molecules in the Computational,
Analytical, and Structural Sciences department of GlaxoSmithKline
[38,39].
All chromatographic experiments were performed under 7.4 pH,
since compounds 1, 2, 16,17,18, and 19 are fully ionized due to

COOH groups at the physiological conditions. Analysis of the presented results indicates that the most lipophilic compound within
the studied derivatives is number 4. Hence, this structure exhibited
the highest logkw in C18 , Phenyl, and CN chromatography. These
molecules also presented a very high affinity to phospholipids, and
only compounds 3 and 6 have higher CHIIAM . Analysis carried out
on C18 modifier column indicated that the less lipophilic is parent OA. Whereas in the case of Phenyl columns, some compounds
(molecules 11, 13, 16, 18, 19, 21, 22, 23, 24), primarily derivatives
of methyl 11-oxooleanolate showed lower lipophilicity than OA.
Retention observed on the CN column is very similar to the retention of OA derivatives in C18 , which was indicated by the correlation matrix presented in Table S6 and as plots in Figure S1 in Supplementary materials. Generally, the obtained indexes are highly
intercorrelated (r > 0.79), suggesting that similar intermolecular
interactions govern retention in investigated RP-HPLC systems.
Considering the limits of this traditional RP-HPLC stationary
phase to mimic the electrostatic interactions between molecule
and phospholipid membranes, IAM chromatography experiments
(which combines membrane simulation with rapid measurements)
were performed. The first thing that draws attention is the lowest
correlation coefficient (range between 0.60–0.68) that we observe

in the obtained correlation matrix, comparing all tested chromatographic systems. This probably occurred because the ionization for
some OA derivatives significantly affects its retention under IAM
whereas neutral compounds showed similar retention on RP-HPLC
and IAM [40–48]. Generally, the IAM stationary phase surface is
mainly zwitterionic at pH 7.4, and interact with the negatively
and positively charged molecules [49–51]. According to the theory
proposed by Avdeef and co-workers, choline moieties (positively
charged at pH 7.4) are located in the outer part of the IAM layer. In
contrast, the phosphate groups are negatively charged at the same
pH and present in the phase’s inner part. Consequntly, the bases
are more retained than acids with the same log P value because
they interact more in-depth [50-51]. Whereas the C18 , Phe, and CN

phases are neutral, excluding negative and undesirable influences
of free-silanol groups. These findings suggest that the membrane
interaction of target OA derivatives has a significantly different nature than lipophilic/hydrophobic interactions.
Nevertheless, the less lipophilic structures (molecule no 18),
also showed the weakest affinity to phospholipids membranes.
CHIIAM of investigated OA derivatives ranges from 38.51 to 72.22,
which seems to be lower values than might be expected given
for consideration the highly lipophilic nature of the compounds. In
the case of OA’s initial structure, it exhibits a moderate affinity to
phospholipids from among the tested compounds, CHIIAM = 56.03
(whereas mean and median are 56.03 and 55.50, respectively),
even though both calculated and chromatographically determined
lipophilicity values were the lowest among tested molecules. This
suggests that although the introduced modifications to the structure of the OA usually increased lipophilicity, it did not translate
unequivocally into a change in the affinity for phospholipids.

6


M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

Journal of Chromatography A 1656 (2021) 462552

Table 3
The summarized values of logkw and CHIIAM indices for
OA derivatives.
No

logkw C18

logkw Ph

logkw CN

CHIIAM

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

25
26
27
28
29
30
31
32
33

2.16
3.75
4.21
5.64
4.69
4.37
3.35
4.19
4.84
3.56
3.94
3.69
2.98
3.62
4.18
2.93
3.07
3.41
3.16
4.23

3.33
3.53
3.32
3.06
3.43
3.96
4.26
3.26
3.60
3.45
3.25
3.72
3.27

3.66
3.67
3.98
4.38
4.05
4.07
3.66
4.04
4.33
3.87
3.57
3.92
3.59
3.81
4.07
3.34

3.69
3.36
3.52
3.84
3.65
3.53
3.51
3.56
3.79
3.83
4.06
3.77
3.77
3.85
3.89
4.04
3.70

4.22
4.87
4.87
5.39
4.99
5.03
4.47
4.95
5.36
4.67
4.86
5.03

4.36
4.77
5.11
4.21
4.28
4.09
4.43
5.08
4.67
4.71
4.66
4.25
4.56
5.00
5.10
4.63
4.72
4.72
4.68
4.97
4.21

56.03
58.12
72.22
69.40
64.52
71.58
58.55
60.76

60.90
61.69
56.40
61.49
49.83
53.49
55.01
49.46
56.98
38.51
47.27
55.39
54.15
51.58
54.66
45.86
49.19
51.83
60.14
60.06
55.50
60.37
48.02
52.05
53.34

mation of lipophilicity database. The scree plot (Fig. 4) confirms
that the three principal components have the most crucial information.
Presented on figure PCA loadings indicated that PC2 differentiates chromatographic and computational lipophilicity indexes. The
situation with the CHIIAM parameter, which lies between the two

approaches, is interesting. Taking into account the significant differences between IAM-HPLC and other tested lipophilicity indices,
the present observations suggest that in the case of OA derivatives,
the IAM-HPLC provides new information about compounds that go
beyond classical lipophilicity.
Weighing the advantages and limitations of unsupervised
chemometric tools such as PCA and CA, it should be highlighted
that both methods do not allow for the selection of the best
and the worst approaches for lipophilicity measurement. What is
more, they do not include information about statistical figures of
performed analysis. For this reason, the SRD analysis, introduced
by Héberger [53,54], was applied for the ranking and selection
of lipophilicity indices [53,55,56]. The sum of ranking differences
(SRD) as a non-parametric and robust method in the last decade
has become a widely used tool in ranking compounds, mathematical models, samples, objects, analytical techniques, etc. [53–57].
The application of the SRD approach in molecular science is reflected in the possibility to rank the molecules in terms of their
molecular features, such as lipophilicity [55–57]. The SRD method
is based on the calculation of absolute differences of ranks between defined reference ranking (known as a benchmark) and each
variable which describes every object [53,54]. Those absolute differences are eventually summed into SRD values. A smaller SRD
value means that the particular variable is closer to the benchmark; in other words: the smaller SRD, the better variable.
The results of conducted SRD-CRRN analysis are presented in
Fig. 5. As it can be seen from the graph the Consensus LogP
lipophilicity descriptor is the closest to the reference ranking and
can be considered to be the best lipophilicity measure of oleanane
triterpenoids derivatives or “real” consensus logP descriptor. It is
followed by ALOGP and XLOGP3 as quite suitable lipophilicity descriptors of the analyzed series of compounds. It must be emphasized that according to the Wilcoxon’s matched pairs test, these
lipophilicity descriptors are statistically different in terms of their
SRD% values obtained in 7-fold cross-validation procedure (Fig. 6).
The following measures, including logkw Ph, WLOGP, logkw C18,
logkw CN, CHIIAM , MLOGP, KOWWINlogP, and ACD/ChemSketchlogP
descriptors are grouped close to each other in terms of similar SRD

values in Figs. 5 and in 6 as well. This is confirmed by Wilcoxon’s
matched pairs test as well. The descriptor Silicos-IT LogP is separated from this group significantly. The iLOGP descriptor is placed
furthest from the reference ranking and aforementioned group of
descriptors, but it is placed quite close to random number distribution described by Gaussian curve. Therefore, iLOGP can be considered to be the worst parameter for the description of lipophilicity
of the analyzed group of oleanane triterpenoids.
Chromatographically determined lipophilicity parameters including logkw Ph, logkw C18, logkw CN and CHIIAM do not overmatch the computationally estimated lipophilicity measures – Consensus LogP, ALOGP and XLOGP3. Nevertheless, the parameters
logkw Ph, logkw C18, logkw CN and CHIIAM are placed relatively close
to the reference ranking and far from the random number distribution. CHIIAM , as an experimentally determined phospholipophilicity
measure, is placed very close to MLOGP in silico lipophilicity descriptor. This can be observed in the SRD graph in Fig. 5 and in the
Box-Whisker plot in Fig. 6. Generally, the SRD-CRRN analysis indicated that all the experimentally determined lipophilicity indices
are statistically very similar to WLOGP, MLOGP, KOWWINlogP and
ACD/ChemSketchlogP in silico lipophilicity descriptors, as it can be
noticed in Fig. 6.

3.4. Comparison of chromatographic and computational lipophilicity
indexes
In order to investigate similarities and dissimilarities between
chromatographic and computational lipophilicity measurements of
the investigated molecules, principal component analysis (PCA),
cluster analysis (CA), and the sum of ranking differences (SRD)
were performed. The above chemometric methods are complementary to each other and provide insight into the data structure.
Among the agglomerative clustering methods, Ward’s method was
selected due to its unique properties. It is based on a classical sumof-squares criterion which produces groups and minimizes dispersion within-group at each binary fusion [52]. The results of CA
are presented in Fig. 3. Two main clusters can be separated. The
first cluster included logkw obtained from C18 , CN and Ph chromatography together with two calculated logP: iLOGP and SilicosIT logP. These software calculations were based on solvation free
energy and topological descriptors corrected by fragmental information, respectively. In the second cluster, two subclusters can be
distinguished. In the first subcluster, IIa following descriptors are
located: calculated logP, XLOGP3, WLOGP, ALOGP, Consensus logP,
KOWWINlogP and MLOGP. The second subcluster IIB comprises
of CHIIAM indices and logP calculated by ACD/ChemSketch. This

CA analysis indicated significant differences between phospholipids
affinity determined chromatographically and others lipophilicity
indices.
PCA analysis confirmed the conclusions of the CA. PCA is one of
the basic multivariate techniques which provides an insight into a
data structure, similarities and dissimilarities of variables, disposition of objects, tendencies for their grouping, and outlying effects.
Summary of PCA analysis is noticed in supplementary information
as Table S7. The first three-component included 86% of the infor7


M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

Journal of Chromatography A 1656 (2021) 462552

Fig. 3. The results of CA.

The probabilities that lipophilicity measures are derived randomly are provided in Supplementary materials in Table S8. For
Consensus LogP descriptor this probability is so small (less than
3.95E-10) so it can be neglected. The highest likelihood of random
character is assigned to iLOGP descriptor (between 1.51 and 1.73).
The clustering in the form of double dendrogram in clustered
heat maps of the lipophilicity measures of oleanane triterpenoids
was carried out based on their SRD% values. The double dendrogram is presented in Fig. 7 in which two main clusters of the
lipophilicity measures can be observed. Silicos-IT LogP and iLOGP
descriptors are placed into the separate cluster having the highest
SRD% values in all 7 steps of 7-fold cross-validation. All the other
descriptors are collected in other main cluster consisting of two
subclusters. The dendrogram indicates the grouping of Consensus
LogP, ALOGP and XLOGP3 having the lowest SRD% values. These
results can be excepted, since Consensus logP the average of in silico logP values calculated by SwissADME software (iLOGP, XLOGP3,

WLOGP, MLOGP and Silicos-IT LogP). The rest of the descriptors are
placed into the other subcluster according to similar SRD% values.
This is the confirmation of the grouping of the descriptors assumed
in Figs. 5 and 6.
3.4. QSRR analysis
Fig. 4. The PCA scree plot.

Insights into the molecular mechanism of chromatographically
determined lipophilicity indices were performed using the QSRR
approach, accepting the assumption that all determined chromatographic indices are the expression of the lipophilic character of
the investigated analytes. Generally, QSRR methodology, introduced
by Kaliszan, linked the relationship between retention and analyte structures mathematically [58]. Obtained QSRR models allow
for the prediction and explanation of the nature of interaction involved in the retention mechanism which takes place between the
compounds and the employed stationary phases.
The selection of theoretical descriptors, which affected the
logkw , was performed by using GA-MLR method. As a result, four

Despite the fact that CA and SRD methods have quite different
computational basics, the comparison between the gruping of the
lipophilicity measures obtained by CA (Fig. 3) and SRD indicates
some similarities. For example, the Consensus LogP descriptor is
placed in the same subcluster as ALOGP and XLOGP3 descriptors,
which are quite close to each other in the SRD graph as well. Also,
iLOGP and Silicos-IT LogP descriptors are separated from other in
silico descriptors as in SRD graph.
8


M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

Journal of Chromatography A 1656 (2021) 462552

Fig. 5. The ranking of normalized lipophilicity measures of oleanane triterpenoids by sum of ranking differences and comparison of ranks by random numbers with row
average as a reference ranking. The statistical characteristics of Gaussian fit are the following: first icosaile (5%), XX1 = 294; first quartile, Q1 = 334; median, Mediana
(Med) = 360; last quartile, Q3 = 388; last icosaile (95%), XX19 = 428.

Fig. 6. The Box-Whisker plot representing the normalized sum of ranking differences values (SRD%) obtained by 7-fold cross-validation of each lipopilicity parameter. The
vertical lines separate statistically different lipophilicity parameters by the means of Wilcoxon’s matched pairs test.

models, each describing retention in investigated chromatographic
systems, have been calculated and listed in Table 4 together with
the statistical figures. The values of theoretical descriptors and
their description were listed in tables S9 and S10, respectively.
Considering the number of OA derivatives in the tested group, a
maximum of five molecular descriptors were added into calculated
QSRR equations since at most five analytes should be used for one
independent variable [59]. The descriptors in the equations were
introduced from most important to less significant, taking the pvalue into account.
The statistical features of obtained models show an excellent
fitting of the data and indicate a good prediction ability of the

models. The small RMSEEXT confirmed that the model is not only
well-fitted for the training but also predicts correctly. Each obtained model dedicated for classical RP-HPLC the GATS7i descriptor
was included. These results suggested that the ionization potential
of molecules significantly affected their retention. Additionally, the
obtained models suggested that 3D-MoRSE descriptors can be very
useful for predicting physicochemical properties of this class of
compounds since they appear very often in calculated models. On
the other hand, the polarizability (H7p) and ionization (MATS8i)
related descriptors are present and significantly affect the CHIIAM

values.

9


logkPh = −7.853(±1.224)GATS7i – 65.9445(±27.345)VE2sign_G/D + 0.303(±0.115)Mor20v + 0.095(±0.0249)Mor29s + 0.0225(±0.0036)ALOGP2 + 11.389(±1.250)
logkCN = 154.291(±22.817)VE2sign_D – 11.227(±2.177)GATS7i – 0.701(±0.179)Eig13_EA(ri) – 2.662(±0.896)E2e – 17.392(±1.989)
logkC18 = −23.978(±7.382)GATS7i + 38.491(±7.664)SpMin1_Bh(v) + 0.829(±0.120)Mor05p + 0.684(±0.158)Mor26i – 31.912(±4.753)Du – 34.418(±16.549)
CHIIAM = −145.279(±14.572)MATS8i −100.391(±11.688)SpMin4_Bh(m) + 12.463(±4.551)H7p + 11.636(±0.838)SssssC – 0.373(±0.138)G(N. .S) + 233.858(±21.322)
RMSEext
Q2
RMSEcv
R2
0.888
0.812
0.109
0.208
0.874
0.819
0.148
0.108
0.912
0.858
0.263
0.281
0.947
0.906
2.236
2.171
1

2
3
4

Journal of Chromatography A 1656 (2021) 462552

Fig. 7. The clustered heat map (double dendrogram) of the lipophilicity measures
of oleanane triterpenoids based on SRD% values obtained by 7-fold cross-validation
of each lipophilicity parameter.

5. Conclusions
Nowadays, concern about the development of new anticancer
drugs is still increasing on account of enlarged susceptibility for
neoplastic disease among society. A series of OA derivatives previously synthesized in our laboratory show auspicious various antitumor activity. Among tested solutes promising anticancer-activity,
IC50 2.8 and 1.6 μg/ml for HeLa and MCF-7 cell line and relatively
low lipophilicity among tested substances showed molecule no 24.
SwissADME calculation suggested that these molecules should be
absorbed orally, which is highly desirable and should not exceed
the BBB, limiting side effects related to CNS. QSRR analysis indicated that the ionization potential of studied molecules significantly affects their retention in classical RP-HPLC, whereas on IAMHPLC, polarizability-related descriptors also play an essential role.
The chemometric analysis presented differences between calculated logP and lipophilicity obtained chromatographically. This fact
can explain the nature of computed lipophilicity parameters, which
refer to octanol/water liquid/liquid static partition equilibrium. At
the same time, chromatographically measure parameters refer to
the dynamic equilibrium constant on a large surface of the stationary phases. According to SRD, the best lipophilicity indices of studied derivatives are consensus LogP descriptors followed by ALOGP
and XLOGP3, while chromatographically determined lipophilicity
parameters do not overmatch them.

1
2
3

4

Equation
Model

Table 4
The four models which describe retention in investigated chromatographic systems.

M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

10


M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

Journal of Chromatography A 1656 (2021) 462552

Declaration of Competing Interest

[15] X.-Y. Zeng, Y.P. Wang, J. Cantley, T.J. Iseli, J.C. Molero, B.D. Hegarty, E.W. Kraegen, Y. Ye, J.-M. Ye, Oleanolic acid reduces hyperglycemia beyond treatment
period with Akt/FoxO1- induced suppression of hepatic gluconeogenesis in
type-2 diabetic mice, PLoS ONE 7 (2012).
[16] B. Bednarczyk-Cwynar, P. Ruszkowski, T. Jarosz, K. Krukiewicz, Enhancing anticancer activity through the combination of bioreducing agents and triterpenes,
Future Med. Chem. 10 (2018) 511–525, doi:10.4155/fmc- 2017- 0154.
[17] B. Bednarczyk-Cwynar, P. Ruszkowski, D. Atamanyuk, R. Lesyk, L. Zaprutko, Hybrids of oleanolic acid with norbornene-2,3-dicarboximide-n-carboxylic acids
as potential anticancer agents, Acta Pol. Pharm. - Drug Res. 74 (2017) 827–835.
[18] D. Kaminskyy, B. Bednarczyk-Cwynar, O. Vasylenko, O. Kazakova, B. Zimenkovsky, L. Zaprutko, R. Lesyk, Synthesis of new potential anticancer agents
based on 4-thiazolidinone and oleanane scaffolds, Med. Chem. Res 21 (2012)
3568–3580, doi:10.10 07/s0 0 044- 011- 9893- 9.
´ Anticancer ef[19] B. Bednarczyk-Cwynar, L. Zaprutko, P. Ruszkowski, B. Hładon,

fect of A-ring or/and C-ring modified oleanolic acid derivatives on KB, MCF7 and HeLa cell lines, Org. Biomol. Chem. 10 (2012) 2201–2205, doi:10.1039/
c2ob06923g.
[20] A. Paszel-Jaworska, B. Rubis´ , B. Bednarczyk-Cwynar, L. Zaprutko, M. Ry´
bczynska,
Proapoptotic activity and ABCC1-related multidrug resistance reduction ability of semisynthetic oleanolic acid derivatives DIOXOL and HIMOXOL
in human acute promyelocytic leukemia cells, Chem. Biol. Interact. 242 (2015)
1–12, doi:10.1016/j.cbi.2015.07.011.
´ B. Rubis´ , M. Pakuła, B. Bednarczyk[21] N. Lisiak, A. Paszel-Jaworska, E. Toton,
´
Cwynar, L. Zaprutko, M. Rybczynska,
Semisynthetic oleanane triterpenoids inhibit migration and invasion of human breast cancer cells through downregulated expression of the ITGB1/PTK2/PXN pathway, Chem. Biol. Interact. 268
(2017) 136–147, doi:10.1016/j.cbi.2017.03.008.
´
˙
[22] V. Krajka-Kuzniak,
B. Bednarczyk-Cwynar, M. Narozna,
H. Szaefer, W. BaerDubowska, Morpholide derivative of the novel oleanolic oxime and succinic
acid conjugate diminish the expression and activity of NF-κ B and STATs in human hepatocellular carcinoma cells, Chem. Biol. Interact. (2019) 311, doi:10.
1016/j.cbi.2019.108786.
[23] K. Krukiewicz, M. Cichy, P. Ruszkowski, R. Turczyn, T. Jarosz, J.K. Zak,
M. Lapkowski, B. Bednarczyk-Cwynar, Betulin-loaded PEDOT films for regional
chemotherapy, Mater. Sci. Eng. C. 73 (2017) 611–615, doi:10.1016/j.msec.2016.
12.115.
[24] M.K.A. Paszel, B. Rubis´ , B. Bednarczyk-Cwynar, L. Zaprutko, M.R.J. Hofmann,
Oleanolic acid derivative metyl 3,11-dioxoolean-12-en28-olate targets multidrug resistance related to ABCB1, Pharm. Rep. 3 (2011) 1500–1517.
[25] G.M.J. Barca, C. Bertoni, L. Carrington, D. Datta, N. De Silva, J.E. Deustua,
D.G. Fedorov, J.R. Gour, A.O. Gunina, E. Guidez, T. Harville, S. Irle, J. Ivanic,
K. Kowalski, S.S. Leang, H. Li, W. Li, J.J. Lutz, I. Magoulas, J. Mato, V. Mironov,
H. Nakata, B.Q. Pham, P. Piecuch, D. Poole, S.R. Pruitt, A.P. Rendell, L.B. Roskop,
K. Ruedenberg, T. Sattasathuchana, M.W. Schmidt, J. Shen, L. Slipchenko,

M. Sosonkina, V. Sundriyal, A. Tiwari, J.L. Galvez Vallejo, B. Westheimer,
M. Włoch, P. Xu, F. Zahariev, M.S. Gordon, Recent developments in the general atomic and molecular electronic structure system, J. Chem. Phys. (2020),
doi:10.1063/5.0 0 05188.
[26] N.M. O’Boyle, M. Banck, C.A. James, C. Morley, T. Vandermeersch, G.R. Hutchison, Open Babel: an open chemical toolbox - 1758-2946-3-33.pdf, J. Cheminform. 3 (2011) 1–14.
[27] P. Gramatica, N. Chirico, E. Papa, S. Cassani, S. Kovarich, QSARINS: a new software for the development, analysis, and validation of QSAR MLR models, J.
Comput. Chem. 34 (2013) 2121–2132, doi:10.1002/jcc.23361.
[28] P. Gramatica, S. Cassani, N. Chirico, QSARINS-chem: insubria datasets and new
QSAR/QSPR models for environmental pollutants in QSARINS, J. Comput. Chem.
35 (2014) 1036–1044, doi:10.1002/jcc.23576.
[29] A. Daina, O. Michielin, V. Zoete, ILOGP: a simple, robust, and efficient description of n-octanol/water partition coefficient for drug design using the GB/SA
approach, J. Chem. Inf. Model. (2014), doi:10.1021/ci500467k.
[30] C.A. Lipinski, F. Lombardo, B.W. Dominy, P.J. Feeney, Experimental and computational approaches to estimate solubility and permeability in drug discovery
and development settings, Adv. Drug Deliv. Rev. 46 (2001) 3–26, doi:10.1016/
s0169-409x(0 0)0 0129-0.
[31] A. Daina, O. Michielin, V. Zoete, SwissADME: a free web tool to evaluate
pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small
molecules, Sci. Rep. (2017), doi:10.1038/srep42717.
[32] A. Daina, V. Zoete, A BOILED-egg to predict gastrointestinal absorption and
brain penetration of small molecules, ChemMedChem 11 (2016) 1117–1121,
doi:10.10 02/cmdc.20160 0182.
[33] E. Choong, M. Dobrinas, P.-.A. Carrupt, C.B. Eap, The permeability Pglycoprotein: a focus on enantioselectivity and brain distribution, Expert Opin.
Drug Metab. Toxicol. 6 (2010) 953–965, doi:10.1517/17425251003789394.
[34] P. Berton, M.K. Mishra, H. Choudhary, A.S. Myerson, R.D. Rogers, Solubility
studies of cyclosporine using ionic liquids, ACS Omega 4 (2019) 7938–7943,
doi:10.1021/acsomega.9b00603.
´ H. Kapica, P. Baranowski, W. Saw[35] K. Ciura, J. Fedorowicz, P. Žuvela, M. Lovric,
icki, M.W. Wong, J. Saczewski,
˛
Affinity of antifungal isoxazolo [3,4-b]pyridine3(1H)-Ones to phospholipids in immobilized artificial membrane (IAM) chromatography, Molecules 25 (2020) 4835, doi:10.3390/molecules25204835.
[36] D. Casoni, A. Kot-Wasik, J. Namies´ nik, C. Sârbu, Lipophilicity data for some

preservatives estimated by reversed-phase liquid chromatography and different computation methods, J. Chromatogr. A. (2009), doi:10.1016/j.chroma.2009.
01.029.

None.
CRediT authorship contribution statement
Monika Pastewska: Investigation, Writing – original draft, Conceptualization. Barbara Bednarczyk-Cwynar: Resources, Conceptualization. Strahinja Kovacˇ evic´ : Software, Visualization, Writing –
original draft. Natalia Buławska: Investigation, Software, Conceptualization, Formal analysis. Szymon Ulenberg: Software, Writing
– review & editing. Paweł Georgiev: Investigation. Hanna Kapica:
Investigation. Piotr Kawczak: Software. Tomasz Baczek:
Writing –
˛
review & editing. Wiesław Sawicki: Funding acquisition. Krzesimir
Ciura: Conceptualization, Writing – original draft, Methodology,
Supervision, Project administration, Software, Formal analysis, Investigation.
Acknowledgement
This research was funded by the Ministry of Science and Higher
Education by means of ST3 02-0 0 03/07/518 statutory funds. The
computing part was supported by the computational cluster of
Copernicus Computing rendering farm, located in Włocławska 161,
´ Poland. We also thank Prof. Paola Gramatica for free
87-100 Torun,
academic licenses for the use of QSARINS software.
Supplementary materials
Supplementary material associated with this article can be
found, in the online version, at doi:10.1016/j.chroma.2021.462552.
References
[1] E.H. Di, L. Kerns, Drug-Like Properties, Acad. Press New York, NY, USA, 2008.
´ P. Žuvela, K.E. Greber, P. Baranowski,
[2] K. Ciura, J. Fedorowicz, F. Andric,
P. Kawczak, J. Nowakowska, T. Baỗzek, J. Saỗzewski, Lipophilicity determination of antifungal isoxazolo [3,4-b]pyridin-3(1h)-ones and their n1-substituted

derivatives with chromatographic and computational methods, Molecules
(2019), doi:10.3390/molecules24234311.
[3] M.J. Waring, Lipophilicity in drug discovery, Expert Opin. Drug Discov. 5 (2010)
235–248, doi:10.1517/17460441003605098.
´
[4] D. Kempinska,
T. Chmiel, A. Kot-Wasik, A. Mróz, Z. Mazerska, J. Namies´ nik,
State of the art and prospects of methods for determination of lipophilicity
of chemical compounds, TrAC Trends Anal. Chem. 113 (2019) 54–73, doi:10.
1016/j.trac.2019.01.011.
´ K.E. Greber, A. Gurgielewicz, W. Sawicki,
[5] K. Ciura, J. Fedorowicz, F. Andric,
J. Saczewski,
˛
Lipophilicity determination of quaternary (fluoro)quinolones by
chromatographic and theoretical approaches, Int. J. Mol. Sci. 20 (2019) 1–15,
doi:10.3390/ijms20215288.
´ cˇ ic,
´ M.K. Banjac, N. Miloševic,
´ J. Cur
´ D. Marjanovic,
´ N. Todor[6] S. Kovacˇ evic,
´ J. Krmar, S. Podunavac-Kuzmanovic,
´ N. Banjac, G. Ušcumli
´
´ Comparaovic,
c,
tive chemometric and quantitative structure-retention relationship analysis of
anisotropic lipophilicity of 1-arylsuccinimide derivatives determined in highperformance thin-layer chromatography system with aprotic solvents, J. Chromatogr. A. (2020), doi:10.1016/j.chroma.2020.461439.
[7] A. Loboda, E. Rojczyk-Golebiewska, B. Bednarczyk-Cwynar, L. Zaprutko,

A. Jozkowicz, J. Dulak, Targeting Nrf2-mediated gene transcription by triterpenoids and their derivatives, Biomol. Ther. 20 (2012) 499–505, doi:10.4062/
biomolther.2012.20.6.499.
[9] S. Singh, G.B. Singh, S. Bani, B.D. Gupta, S.K. Banerjee, Anti-inflammatory activity of oleanolic acid in rats and mice, J. Pharm. Pharmacol. 44 (1992) 456–45.
[10] S. Begum, I. Sultana, B.S. Siddiqui, F. Shaheen, A.H. Gilani, Structure and spasmolytic activity of eucalyptanoic acid from Eucalyptus camaldulensis var. obtusa and synthesis of its active derivative from oleanolic acid, J. Nat. Prod. 65
(2002) 1939–1.
[11] J. Liu, Q. Wu, Y.-F. Lu, J. Pi, New insights into generalized hepatoprotective
effects of oleanolic acid: key roles of metallothionein and Nrf2 induction,
Biochem. Pharmacol. 76 (2008) 922–992.
[12] C.M. Mengoni, F. Lichtner, M. Battinelli, L. Marzi, M. Mastroianni, G. Vullo,
V Mazzanti, In vitro anti-HIV activity of oleanolic acid on infected human
mononuclear cells, Planta Med 68 (2002) 111–114.
[13] J.-E. Yuk, M.Y. Lee, O.K. Kwon, X.F. Cai, H.Y. Jang, S.R. Oh, H.K. Lee, K.S. Ahn,
Effects of astilbic acid on airway hyperresponsiveness and inflammation in a
mouse model of allergic asthma, Int. Immunopharmacol. 11 (2011) 266–273.
[14] C.D. Liu, J. Liu, Y. Mao, Q. Klaassen, The effects of 10 triterpenoid compounds
on experimental liver injury in mice, Fundam. Appl. Toxicol. 22 (1994) 34–40.
11


M. Pastewska, B. Bednarczyk-Cwynar, S. Kovacˇ evi´c et al.

Journal of Chromatography A 1656 (2021) 462552

[37] Giaginis Constantinos, Kakoulidou Anna Tsantili, Current State of the Art in
HPLC Methodology for Lipophilicity Assessment of Basic Drugs. A review, Journal of Liquid Chromatography & Related Technologies 31 (1) (2007) 79–96,
doi:10.1080/10826070701665626.
[38] S. Teague, K. Valko, How to identify and eliminate compounds with a risk of
high clinical dose during the early phase of lead optimisation in drug discovery, Eur. J. Pharm. Sci. 110 (2017) 37–50, doi:10.1016/j.ejps.2017.02.017.
[39] K.L. Valko, Application of biomimetic HPLC to estimate in vivo behavior of
early drug discovery compounds, Futur. Drug Discov. 1 (2019) FDD11, doi:10.

4155/fdd- 2019- 0 0 04.
[40] A. Taillardat-Bertschinger, P.-.A. Carrupt, F. Barbato, B. Testa, Immobilized artificial membrane HPLC in drug research, J. Med. Chem. 46 (2003) 655–665,
doi:10.1021/jm020265j.
[41] A. Taillardat-Ertschinger, A. Galland, P.-.A. Carrupt, B. Testa, Immobilized artificial membrane liquid chromatography: proposed guidelines for technical optimization of retention measurements, J. Chromatogr. A. 953 (2002) 39–53,
doi:10.1016/s0 021-9673(02)0 0119-x.
[42] A. Taillardat-Bertschinger, C.A.M. Martinet, P.-.A. Carrupt, M. Reist, G. Caron,
R. Fruttero, B. Testa, Molecular factors influencing retention on immobilized artifical membranes (IAM) compared to partitioning in liposomes and n-octanol,
Pharm. Res. 19 (2002) 729–737, doi:10.1023/a:1016156927420.
[43] M. Chrysanthakopoulos, F. Tsopelas, A. Tsantili-Kakoulidou, Biomimetic chromatography: a useful tool in the drug discovery process, Adv. Chromatogr 51
(2014) 91–125.
[44] F. Tsopelas, T. Vallianatou, A. Tsantili-Kakoulidou, Advances in immobilized
artificial membrane (IAM) chromatography for novel drug discovery, Expert
Opin. Drug Discov. 11 (2016) 473–488, doi:10.1517/17460441.2016.1160886.
[45] L. Grumetto, C. Carpentiero, P. Di Vaio, F. Frecentese, F. Barbato, Lipophilic and
polar interaction forces between acidic drugs and membrane phospholipids
encoded in IAM-HPLC indexes: their role in membrane partition and relationships with BBB permeation data, J. Pharm. Biomed. Anal. 75 (2013) 165–172,
doi:10.1016/j.jpba.2012.11.034.
[46] L. Grumetto, G. Russo, F. Barbato, Immobilized artificial membrane HPLC derived parameters vs PAMPA-BBB data in estimating in situ measured bloodbrain barrier permeation of drugs, Mol. Pharm. 13 (2016) 2808–2816, doi:10.
1021/acs.molpharmaceut.6b00397.
[47] G. Russo, L. Grumetto, F. Barbato, G. Vistoli, A. Pedretti, Prediction and mechanism elucidation of analyte retention on phospholipid stationary phases (IAMHPLC) by in silico calculated physico-chemical descriptors, Eur. J. Pharm. Sci..
99 (2017) 173–184, doi:10.1016/j.ejps.2016.11.026.
[48] G. Russo, L. Grumetto, R. Szucs, F. Barbato, F. Lynen, Determination of in vitro
and in silico indexes for the modeling of blood-brain barrier partitioning of
drugs via micellar and immobilized artificial membrane liquid chromatography, J. Med. Chem. 60 (2017) 3739–3754, doi:10.1021/acs.jmedchem.6b01811.

[49] A. Avdeef, K.J. Box, J.E. Comer, C. Hibbert, K.Y. Tam, pH-metric logP 10. Determination of liposomal membrane-water partition coefficients of ionizable
drugs, Pharm. Res. 15 (1998) 209–215, doi:10.1023/a:1011954332221.
[50] L. Grumetto, C. Carpentiero, F. Barbato, Lipophilic and electrostatic forces encoded in IAM-HPLC indexes of basic drugs: their role in membrane partition
and their relationships with BBB passage data, Eur. J. Pharm. Sci. 45 (2012)
685–692, doi:10.1016/j.ejps.2012.01.008.

[51] L. Grumetto, C. Carpentiero, P. Di Vaio, F. Frecentese, F. Barbato, Lipophilic and
polar interaction forces between acidic drugs and membrane phospholipids
encoded in IAM-HPLC indexes: their role in membrane partition and relationships with BBB permeation data, J. Pharm. Biomed. Anal. 75 (2013) 165–172,
doi:10.1016/j.jpba.2012.11.034.
[52] K. Ciura, M. Belka, P. Kawczak, T. Baczek,
˛
J. Nowakowska, The comparative
study of micellar TLC and RP-TLC as potential tools for lipophilicity assessment based on QSRR approach, J. Pharm. Biomed. Anal. 149 (2018) 70–79,
doi:10.1016/j.jpba.2017.10.034.
[53] K. Héberger, Sum of ranking differences compares methods or models fairly,
TrAC - Trends Anal. Chem. 29 (2010) 101–109, doi:10.1016/j.trac.20 09.09.0 09.
[54] K. Héberger, K. Kollár-Hunek, Sum of ranking differences for method discrimination and its validation: comparison of ranks with random numbers, J.
Chemom. 25 (2011) 151–158, doi:10.1002/cem.1320.
´ K. Héberger, Chromatographic and computational assessment of
[55] F. Andric,
lipophilicity using sum of ranking differences and generalized pair-correlation,
J. Chromatogr. A. 1380 (2015) 130–138, doi:10.1016/j.chroma.2014.12.073.
´ D. Bajusz, A. Rácz, S. Šegan, K. Héberger, Multivariate assessment
[56] F. Andric,
of lipophilicity scales—computational and reversed phase thin-layer chromatographic indices, J. Pharm. Biomed. Anal. 127 (2016) 81–93, doi:10.1016/j.jpba.
2016.04.001.
´ K. Héberger, Towards better understanding of lipophilicity: assess[57] F. Andric,
ment of in silico and chromatographic logP measures for pharmaceutically important compounds by nonparametric rankings, J. Pharm. Biomed. Anal. 115
(2015) 183–191, doi:10.1016/j.jpba.2015.07.006.
[58] R. Kaliszan, H. Foks, The relationship between the RM values and the connectivity indices for pyrazine carbothioamide derivatives, Chromatographia 10
(1977) 346–349, doi:10.1007/BF02274482.
´ M. Natic,
´ Z. Džambaski, R. Markovic,
´ D. Milojkovic-Opsenica,
´

[59] D. Dabic,
´ Quantitative structure–retention relationship of new N-substituted 2Ž. Tešic,
alkylidene-4-oxothiazolidines, J. Sep. Sci. 34 (2011) 2397–2404, doi:10.1002/
jssc.201100266.
[8] Bednarczyk-Cwynar B, Ruszkowski P, Bobkiewicz-Kozlowska T, Zaprutko L.
Oleanolic Acid A-lactams Inhibit the Growth of HeLa, KB, MCF-7 and HepG2 Cancer Cell Lines at Micromolar Concentrations. Anticancer Agents Med
Chem. 2016;16(5):579-92. doi: 10.2174/1871520615666150907095756. PMID:
26343139.

12