89urn:lsid:arphahub.com:pub:CCBBD2E1-BDAF-5702-AE4B-0E4910A4C7D6PharmaciaPHAR0428-02962603-557XBulgarian Pharmaceutical Scientific Society10.3897/pharmacia.73.e200443200443Research ArticlePharmaceuticsInvestigation of lipophilicity of new stilbene derivatives with chromatographic and computational approachesGeorgievaStelaohsg@chem.uni-sofia.bghttps://orcid.org/0000-0003-0007-25091VoyslavovTsvetomilhttps://orcid.org/0000-0002-4395-99541ChanevChristohttps://orcid.org/0000-0003-4705-29511PetrovOgnyanhttps://orcid.org/0000-0002-5541-541X1Faculty of Chemistry and Pharmacy, Sofia University “St. Kliment Ohridski”, 1 James Bouchier Blvd, 1164, Sofia, BulgariaFaculty of Chemistry and Pharmacy, Sofia University “St. Kliment Ohridski”SofiaBulgariahttps://ror.org/02jv3k292
20262109202673e20044384480ED9-77F2-503E-A33C-2453794036F52005202601092026Stela Georgieva, Tsvetomil Voyslavov, Christo Chanev, Ognyan PetrovThis is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
The octanol-water partition coefficient, log Po/w, serves as the standard parameter for hydrophobicity, a property that is correlated with transmembrane distribution, biological activity, and the pharmacokinetic behavior of bioactive compounds. In our investigation, the lipophilic characteristics of thirty-one novel combretastatin A-4 derivatives were evaluated using five chromatographic systems across three reversed-phase high performance liquid chromatography stationary phases. The derived retention factors, log kw, were subsequently correlated with computationally predicted lipophilicity values, clog P. This comparative analysis assessed the congruence between the experimental and theoretical datasets for the analyzed compounds. The relationship between the chemical structure of examined derivatives and their lipophilic characteristics were described. Linear pattern recognition techniques were applied for chemometric analysis.
Drug discovery and development is an expensive and time-consuming process. A large part of testing compounds drops out owing to low efficacy and high toxicity. The earlier prediction of potential of the compounds of Absorption, Distribution, Metabolism, and Excretion, ADME, could be helpful in rejecting of the inappropriate candidates. As a result, many companies are focusing on compound properties early in preclinical development (Matthews et al. 2016; Lobo 2020; Simoens and Huys 2021). In order to forecast the characteristics of the substances in pharmacokinetic processes it is necessary to be determined their solubility, lipophilicity, stability, and acid-based character.
Lipophilicity expressed as the logarithm of the partition coefficient between n-octanol and water, log Po/w, is a fundamental predictor of the biological activity, the ability of molecules to cross cell membranes, and bind to proteins. Consequently, it significantly influences the pharmacokinetics, pharmacodynamics, and toxicological effects (Waring 2010; Lobo 2020). According to Lipinski’s rule of five, small molecule drug candidates intended for oral administration should ideally possess a log Po/w between 0 and 5 (Lipinski et al. 2001). Comparisons made between marketed oral drugs and candidates in earlier stages of development indicate that values exceeding this criterion are frequently associated with unfavorable drug features, such as poor aqueous solubility, high plasma protein binding, and tissue accumulation. Furthermore, excessive lipophilicity increases the risk of in vitro receptor promiscuity and in vivo toxicity. However, the insufficient lipophilicity results in poor ADME properties (Leeson and St-Gallay 2011; Lobo 2020).
The value of log Po/w of potential bioactive substances can be calculated (clog P) or measured empirically. The Organisation for Economic Cooperation and Development (OECD) has outlined the standard protocols for experimental determination, namely the shake flask method (OECD 1995) and RP-HPLC method (OECD 2022). The advantages of RP-HPLC over the liquid-liquid extraction method are rapidity, accuracy, very little substance is required, and there is no objection to the use of impure substances. A significant benefit of HPLC-based determination is the opportunity use of various types of stationary phases, such as C8, C18, or IAM, and this allows for a more assessment of lipophilic deportment (Cimpan et al. 1998; Nasal et al. 2003; Poole and Poole 2003; Valkó 2004). The number and the nature of residual silanol groups, the length of the hydrocarbon chain, the nature of the bonding with the silica gel, and the manufacturing technology, which includes pore size and specific surface area, have a significant impact on the final capabilities of the stationary phase (Cimpan et al. 1998; Valkó 2004; Godinho et al. 2020).
The chromatographic techniques offer greater potential to determine the partition properties of compounds based on their chromatographic characteristic of retention (tr) which directly reflects the analyte’s distribution between the stationary and the mobile phase. Different approaches exist to express the chromatographic lipophilicity index (Nasal et al. 2003; Valkó 2004). The retention factor with pure aqueous mobile phase (log kw) is commonly used lipophilic descriptor, as this chromatographic system most closely resembles the classical shake flask method. However, direct measurement of log kw is difficult due to prohibitively long retention times of organic solutes in pure water as mobile phase (Cimpan et al. 2000; Nasal et al. 2003; Valkó 2004). To overcome this, a linear regression model is employed to describe the relationship between the volume fraction (φ) of the organic modifier in mobile phase and the measured retention factor (log k) obtained for different eluent compositions. The extrapolation to 0% organic modifier is log kw (Elmansi et al. 2019; Redka et al. 2023; Kovačević et al. 2025).
In silico prediction is an alternative approach for assessing various pharmacokinetic ADMET parameters. These computational methods calculate lipophilicity using various approaches, including molecular fragmentation, atomic contributions, and whole molecule property methods. Such approaches are widely used in drug discovery, because they provide a cost-effective and rapid process even before the compounds are synthezed. The major challenge for computational methods lies in their predictability and reliability (Daina et al. 2017; Silakari and Sindh 2021; Gumieniczek et al. 2025). Consequently, many researchers have advocate for RPLC as an appropriate alternative. This is achieved by using correlations between experimental chromatographic parameters and corresponding log Po/w values from the shake flask experiments or theoretical calculations. It is important to note that the resulting correlation coefficients are quite different and depending on the specific stationary phase and the structure of the compounds under investigation (Casoni et al. 2009; Apostolov et al. 2025; Jeleń et al. 2025; Kovačević et al. 2025).
Cancer is one of the most common and difficult diseases to treat, ranking as the second leading cause of mortality worldwide (OECD 2025). To address this, a new class of compounds has been developed. It combines the cis-configured double bond bridge of Combretastatin A4 (CA-4) with a benzoxazolone heterocycle (Fig. 1). While CA-4 is vigorous microtubule-destabilizing agent that inhibits tumor growth by binding to the colchicine site of tubulin, its therapeutic utility is limited by low aqueous solubility and poor bioavailability in biological media (Wolter and Stein 2002; Gerova et al. 2016; Bukhari et al. 2017; Hamze et al. 2020). The derivatives of 2-benzoxazolone exhibit diverse pharmacological activities – antitumor, antiviral, antidepressant, antioxidant, and anti-inflammatory properties (Imramovsky et al. 2014; Kamal et al. 2020). Therefore, the investigation of these novel CA-4/benzoxazolones hybrids is of significant interest. Following earlier reports on their synthesis, in vitro cytotoxicity, and lipophilicity (log P with two chromatographic systems) (Gerova et al. 2016; Georgieva et al. 2025), we now investigate the properties of the new combretastatin A-4 analogues.
General formula of styrylbenzoxazolones Ri – H or H3CO.
https://binary.pensoft.net/fig/1780251
The current study is a follow-up paper to the previous article (Georgieva et al. 2025). In this publication, the focus was placed on experimental characterization of HPLC chromatographic behavior of 31 new synthesized cis- and trans-styrylbenzoxazolones derivatives of CA-4 on five RP-HPLC systems. We discussed the relationship between chromatographic lipophilicity parameter (log kw) and the chemical structure of analyzed compounds. The in silico lipophilicity of these compounds was also evaluated. The experimental and calculated values of parameters of lipophilicity were correlated as well as chemometric analyzed by HCA and PCA. In our study, we used code names for the newly synthesized benzoxazolones. These names indicate the type of stereoisomers – Z and E, the position of styryl fragment in ring B in the benzoxasolone fragment (4, 5, 6 or 7) as well as the number and position of the methoxy groups in ring A (1/2, 3/4, 5/6, 7/8), (Fig. 1, Table 2).
Materials and methodsChemicals and materials
HPLC-grade methanol and acetonitrile were supplied by Merck KGaA (Darmstadt, Germany). Deionized water used in this study was purified using a Mili-Q system (Milipore, Bedford, MA, USA).
Stock solutions of test benzoxazolones were prepared in acetonitrile-methanol (1:1, v/v) because of the low solubility of some compounds. The injection concentration was 0.2 mg/ml in the working organic modifier (acetonitrile or methanol).
Methods
Instrumental and chromatographic conditions
The experimental data were acquired using a HPLC system (Konik Tech 560) equipped with a thermostat (Tcol 35 °C). A Diod Array UV detector (Konik DAD 560) was employed to measure the retention of test components. DAD was operated in full acquisition mode over the 200–600 nm interval, with chromatograms simultaneously recorded at 254, 274, 292 and 305 nm analytical wavelengths. The absorption maximum of Z-isomers is between 274 to 300 nm and of E-isomers is between 302 to 336 nm.
Retention studies were conducted using three C18 chromatographic columns, Table 1. Isocratic compositions of the mobile phase were used for the experiments. The chromatographic procedure was carried through isocratic conditions with mixtures of methanol or acetonitrile as organic modifiers and water, starting with 95% to 60% organic solvent (φ) – 95, 90, 80, 70, 60% (v/v). The flow rate was 0.75 ml/min in all cases. The injection volume was 10 μL. Each measurement was performed in triplicate and the mean was used in further calculations. The solutes were stable from day to day. The void volume was measured by injecting a solute of the unretained compound potassium nitrate, KNO3.
Characteristics of chromatographic systems - column and mobile phase.
Column
Length, mm
Particle size, µm
Total surface, m2/g
Pore size, Å
Carbon load, %
Matrix
Feature
Manufacture
Mobile phase
System
LiChrosorb
200
5
300
100
16.2
silica
no
Merck Hewlett Packard
CH3OH-H2O
A
CH3CN-H2O
B
Vydac
250
5
70-90
300
8
polymer
no
Vydac
CH3OH-H2O
C
CH3CN-H2O
D
SupelcoSil
150
3
170
120
11
silica
endcapped
Merck-Supelco
CH3OH-H2O
E
Retention characteristics
The most frequently used parameter of retention in HPLC is the capacity factor, k. The k value is determined from the retention time, calculated according to the expression:
k = (tR – t0) / t0 Eq. 1
where tR is the retention time of the analyzed compound;
t0 is the void time of the column used.
In RP-HPLC there is a linear correlation (Snyder-Soczeminski equation) between the retention of the analyte expressed as log k and the percentage of the organic modifier in the mobile phase (φ) for binary water-organic eluents:
log k = log kw – Sφ Eq. 2
where k is the retention factor of the analyzed compound;
kw is a value of k of the analyte extrapolated to the mobile phase containing 100% water, which represents the commonly employed chromatographic hydrophobicity parameter;
S is a solute-dependent solvent strength parameter. It is considered to be related to the specific hydrophobic surface area of the solute (Poole and Poole 2003; Valkó 2004; Han et al. 2012; Coutinho et al. 2023).
A better correlation between log k and φ can be further described by the quadratic relationship (Baçzek et al. 2000; Poole and Poole 2003; Valkó 2004; Tache et al. 2012):
log k = log kw + a1φ + a2φ2 Eq. 3
where k is the analyte retention factor;
a1 and a2 are constants for a given analyte and a reversed phase system;
φ is the volume fraction of the stronger solvent in the mobile phase;
kw is the value of k extrapolated to 100% water as mobile phase (φ = 0).
The quadratic term is often explained through the perturbation of the retention behavior of the analytes on a modified hydrophobic stationary phase when using a water-rich mobile phase. For values of φ greater than 30% this model often gives a more appropriate fitting than the linear model, especially with a mobile phase comprised of acetonitrile as the organic modifier (Valkó 2004).
Log kw is widely accepted as the most reproducible and reliable lipophilicity descriptor derived from chromatographic retention data. The values are most often used as measures of hydrophobicity from RP-HPLC. Log kw values can predict log Po/w only for nonhydrogen bonding or weak hydrogen-accepting substances (Baçzek et al. 2000).
Theoretical calculation of clog P
SwissADME (http://www.swissadme.ch/index.php) gives access to five freely available predictive models, i.e. XLOGP3, an atomistic method including corrective factors and knowledge-based library; WLOGP, own implementation of a purely atomistic method based on the fragmental system of Wildman and Crippen; MLOGP, an archetype of topological method relying on a linear relationship with 13 molecular descriptors; SILICOS-IT, an hybrid method relying on 27 fragments and 7 topological descriptors, and finally iLOGP, in-house 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. iLOGP was benchmarked on two drug or drug-like external sets and performed equally as or better than six well-established predictors. The consensus log Po/w (clog P) is the arithmetic mean of the values predicted by the five proposed methods. (Daina et al. 2017; Silakari and Sindh 2021; Gumieniczek et al. 2025).
Statistics
Mandel’s test compares the residual standard deviations of nonlinear and checks linear models by using a Fisher’s F-test and check the best fit for the calibration curve (linear or quadratic) of each analyte. All regression analyses and statistical evaluations were performed via Microsoft Excel on a custom-made calculator. The statistical data for the p-value were calculated for a 95% confidence limit (Danzer and Currie 1998; Andrade and Gomez-Carracedo 2013; ISO 8466-1:2021 (2021)).
The statistical methods used throughout the study were Principal Component Analysis (PCA) and Hierarchical Clustering Analysis (HCA). Data treatment, statistical analysis, and visualizations were performed using the R language, version 4.4.3 (2025-02-28) (R Core Team 2025).
Results and discussionLipophilicity determination
Chromatographic lipophilicity, log k w
The RPLC is a widely used method to determine the lipophilicity. The C18 stationary phase contains long hydrophobic chains and nonpolar surface. The major intramolecular forces that arise between the stationary phase and organic compounds with different structure or functional groups are hydrophobic. More hydrophobic analytes retain stronger to the stationary phase. The determined parameters of retention (log kw) of tested compounds in all of five chromatographic systems are listed in Table 2.
The chromatographic lipophilicity indices (log kw) and the type of regression model of the investigated compounds.
Comp.
Posa
Substituents in ring A
System A
System B
System C
System D
System E
R1
R2
R3
log kw
RMb
log kw
RMb
log kw
RMb
log kw
RMb
log kw
RMb
S41-Z
4
OCH3
OCH3
OCH3
4.61
non
2.37
non
5.34
non
2.53
non
4.08
non
S42-E
4
OCH3
OCH3
OCH3
5.42
non
2.80
non
4.96
non
1.74
non
5.65
non
S43-Z
4
OCH3
H
OCH3
6.69
non
2.99
non
4.56
non
2.53
non
5.56
non
S44-E
4
OCH3
H
OCH3
6.97
non
3.36
non
2.29
lin
3.17
non
6.49
non
S45-Z
4
OCH3
OCH3
H
4.72
non
2.58
non
3.73
non
1.97
non
4.72
non
S46-E
4
OCH3
OCH3
H
5.68
non
2.89
non
5.57
non
2.06
non
5.57
non
S47-Z
4
H
OCH3
H
5.75
non
3.12
non
5.01
non
1.65
non
5.53
non
S48-E
4
H
OCH3
H
6.49
non
3.33
non
5.84
non
0.80
lin
6.29
non
S51-Z
5
OCH3
OCH3
OCH3
5.09
non
2.88
non
4.78
non
1.77
non
5.45
non
S52-E
5
OCH3
OCH3
OCH3
5.76
non
2.96
non
5.34
non
2.55
non
5.95
non
S53-Z
5
OCH3
H
OCH3
6.11
non
3.35
non
5.84
non
2.14
non
5.83
non
S54-E
5
OCH3
H
OCH3
6.36
non
3.43
non
6.45
non
2.43
non
6.31
non
S55-Z
5
OCH3
OCH3
H
2.05
lin
2.75
non
4.57
non
1.87
non
5.56
non
S56-E
5
OCH3
OCH3
H
5.43
non
2.94
non
5.34
non
1.96
non
5.85
non
S57-Z
5
H
OCH3
H
6.09
non
3.32
non
5.98
non
2.37
non
6.04
non
S58-E
5
H
OCH3
H
6.59
non
3.13
non
6.20
non
2.50
non
5.73
non
S61-Z
6
OCH3
OCH3
OCH3
5.24
non
2.91
non
4.02
non
1.84
non
6.20
non
S62-E
6
OCH3
OCH3
OCH3
5.42
non
2.82
non
4.99
non
1.84
non
5.87
non
S63-Z
6
OCH3
H
OCH3
6.06
non
3.27
non
5.82
non
2.21
non
6.44
non
S64-E
6
OCH3
H
OCH3
6.23
non
3.52
non
6.32
non
2.45
non
6.97
non
S65-Z
6
OCH3
OCH3
H
5.19
non
2.77
non
5.09
non
1.94
non
5.81
non
S66-E
6
OCH3
OCH3
H
5.91
non
2.81
non
5.83
non
2.01
non
6.06
non
S67-Z
6
H
OCH3
H
5.85
non
1.39
lin
5.83
non
2.26
non
6.68
non
S71-Z
7
OCH3
OCH3
OCH3
4.92
non
2.52
non
3.84
non
2.51
non
5.53
non
S72-E
7
OCH3
OCH3
OCH3
5.75
non
3.13
non
4.91
non
2.24
non
6.11
non
S73-Z
7
OCH3
H
OCH3
5.15
non
3.18
non
4.68
non
2.54
non
6.16
non
S74-E
7
OCH3
H
OCH3
6.78
non
3.45
non
5.46
non
3.15
non
6.68
non
S75-Z
7
OCH3
OCH3
H
4.61
non
2.68
non
3.56
non
2.34
non
5.38
non
S76-E
7
OCH3
OCH3
H
6.23
non
2.89
non
4.24
non
2.32
non
6.09
non
S77-Z
7
H
OCH3
H
5.43
non
3.18
non
4.42
non
2.32
non
6.38
non
S78-E
7
H
OCH3
H
5.79
non
3.38
non
5.07
non
2.74
non
6.38
non
a Pos: position of the styryl fragment in ring B. b RM: regression model; lin—linear, non—nonlinear (quadratic).
The solutions of each styrylbenzoxazolones derivatives of CA-4 were analyzed with systems A, B, C, D, and E. The capacity factors (k) of the tested compound were calculated according to Eq. 1. The regression function log k = f (φ) have been extrapolated to 0% organic modifier (CH3OH or CH3CN) to determine the capacity factor represented as log kw. The log kw values were computed through linear and quadratic regression models. It’s known that the log k versus φ is linear just in a short interval of mobile phase (Valkó 2004). However, the values of coefficient a2 (Eq. 3) were too small in all the nonlinear equations: System A from 0.0003 to 0.0007, System B from 0.0001 to 0.0003, System C from 0.0004 to 0. 0007, System D from 0.00001 to 0.0004 and System E from 0.0004 to 0.0006. This introduces great uncertainty regarding the choice of an appropriate regression model. IUPAC recommends Mandel’s test to check the best fit for the calibration curve (linear or quadratic) of each analyte (Danzer and Currie 1998; Andrade and Gomez-Carracedo 2013; ISO 8466-1:2021 (2021)). In our previous publication (Georgieva et al. 2025) we reported a linear correlation for S55-Z and S56-E with system A and S41-Z, S42-E, S46-E, S62-E, S67-Z with system B. We have expanded the work interval of mobile phase with 65% and 75% of organic solvents and recalculate the results. The Mandel’s test indicated that a linear model is better than a quadratic function for S55-Z in System A, S67-Z in System B, S44-E in System C, and S48-E in System D, (Table 2). Therefore, we used the log kw values confirmed from Mandel’s test in further calculations. There is no styrylbenzoxazolones with linear function in System E. The stationary phase is endcapped (Table 1) all undesired reactions are reduced. All analyzed stilbenes react to the stationary phase in this system similarly.
Structure - retention analysis
Methanol is a protonic solvent and can donate H-bonds. Due to its mentioned property, the interaction between the analyte and C18-chains in stationary phase are enhanced. This leads to longer retention. Acetonitrile is aprotic solvent, it has low polarity and led to faster elution to the substances because of stronger hydrophobic interactions with mobile phase (Ilijaš et al. 2013; Kovačević et al. 2025). The obtained values of log kw indicate that all the compounds exhibited higher retention in the system with CH3OH-H2O – System A, System C, and System E, (Fig. 2). The exceptions to this are S55-Z in System A and S44-E in System C. We have also commented on a different retention model for them, linear. Considerable correlations between these three systems indicate similar mechanisms of retention between the compounds and the chromatographic systems. The behavior of analytes in System C show lower retention. As shown in Table 1, the stationary phase in System C features a wider surface pore size. Larger pores reduce the obstruction of the analytes’ movement thereby maintaining the analytes ability to diffuse efficiently. This structure facilitated easy desorption of the compounds. The result is weaker retention. Another significant distinction in System C is its polymer-based matrix and this results in different types of interplay mechanisms.
Profile of lipophilicity index log kw of the investigated stilbene derivatives determined with five chromatographic systems.
https://binary.pensoft.net/fig/1780252
Mobile phase containing acetonitrile has higher eluting power, resulting in faster elution and redused retention, System B and System D. This explains why lipophilic indexes in systems with methanol have higher values than acetonitrile. The differences in the structure on the surface of the column in System D, compared to System B is shown in Table 1. The wider pore size and the polymeric matrix influence the weak molecule interactions and lower retention (log kw) in System D.
The passive diffusion is the mean rough the drug enter the cell and higher lipophilicity is connecting with easier passage cross the cell membrane. The retention in RP-HPLC is related to the penetration of the molecules between the non-polar chains of the stationary phase. This process is greatly facilitated if the molecules have more planar, more elongated structure or nonpolar fragments. Greater retention and higher values for log kw were expected for the E-isomers of the tested styrylbenzoxazolones in all chromatographic systems. All E-isomers showed higher retention than their Z-isomers in System A. Only a few irregularities have been noticed in the other chromatographic systems: S57-Z / S58-E and S61-Z / S62-E for System B, S41-Z / S42-E and S43-Z / S44-E for System C, S41-Z / S42-E, S47-Z / S48-E, S71-Z / S72-E, and S75-Z / S76-E for System D, and S57-Z / S58-E and S61-Z / S62-E for System E. Identical retention characteristics were observed for the Z - and E-isomers in two pairs – S61-Z / S62-E in System D and S77-Z / S78-E in System E. The retentions of the analytes with various composition of the mobile phase are different but the extrapolation led to equal log kw.
Theoretical calculation of lipophilicity, clog P
The in silico lipophilicity parameters of the analyzed styrylbenzoxazolones are provided in Table 3. All the models calculated different values of lipophilicity. There is a distinction of calculated clog P between the cis- and trans-isomers, the number of different substituents in ring A and the position of the styryl radical in cycle B just with iLOGP. The other predictive models – XLOGP3, WLOGP, MLOGP, SILICOS-IT, do not distinguish between the substituent position in the benzoxazolonic ring B, the position of two substituents (3,5-dimethoxy (S_3-Z/S_4-E) and 3,4-dimethoxy (S_5-Z and S_6-E)) in aromatic cycle A or E- and Z-isomers. But they contribute to the consensus log Po/w because they use dissimilar algorithms that take various factors into account. In the following comments we use the consensus log Po/w (clog P) combines all calculated values. The clog P values of the E-isomers are higher than Z-isomers except the pairs S41-Z / S42-E and S43-Z / S44-E. Taking into account the Lipinski’s rule, the studied derivatives are considered to be optimal lipophilic. The values of the clog P are from 3.10 to 3.24. The lipophilicity was the lowest value for compounds with more methoxy groups. The compounds with three H3CO-groups in ring A - S_1 / S_2, were the most polar structures. They showed the lowest clog P in each group of derivatives except S41-Z / S42-E. But the analytes with one methoxy group, S_7-Z / S_8-E, were the most lipophilic. They are more unpolar and they will distribute in the organic media easily. All E-isomers exhibited higher lipophilicity compared to Z-isomer. However, the isomers in S41-Z / S42-E and S43-Z / S44-E have the same values. The most significant difference between the E- and Z-isomers in a pair was found at the position 7 of the styryl residue.
In silico lipophilicity parameters of the new stilbene derivatives.
Comp.
log Po/w (iLOGP)
log Po/w (XLOGP3)
log Po/w (WLOGP)
log Po/w (MLOGP)
log Po/w (SILICOS-IT)
Consensus log Po/w
S41-Z
3.39
3.50
3.11
2.12
3.66
3.16
S42-E
3.41
3.50
3.11
2.12
3.66
3.16
S43-Z
3.04
3.53
3.10
2.43
3.60
3.14
S44-E
3.05
3.53
3.10
2.43
3.60
3.14
S45-Z
3.00
3.53
3.10
2.43
3.60
3.13
S46-E
3.21
3.53
3.10
2.43
3.60
3.17
S47-Z
3.06
3.56
3.09
2.75
3.55
3.20
S48-E
3.14
3.56
3.09
2.75
3.55
3.22
S51-Z
3.39
3.50
3.11
2.12
3.66
3.16
S52-E
3.58
3.50
3.11
2.12
3.66
3.19
S53-Z
3.36
3.53
3.10
2.43
3.60
3.20
S54-E
3.48
3.53
3.10
2.43
3.60
3.23
S55-Z
3.14
3.53
3.10
2.43
3.60
3.16
S56-E
3.45
3.53
3.10
2.43
3.60
3.22
S57-Z
3.16
3.56
3.09
2.75
3.55
3.22
S58-E
3.21
3.56
3.09
2.75
3.55
3.23
S61-Z
3.42
3.50
3.11
2.12
3.66
3.16
S62-E
3.58
3.50
3.11
2.12
3.66
3.19
S63-Z
3.46
3.53
3.10
2.43
3.60
3.22
S64-E
3.48
3.53
3.10
2.43
3.60
3.23
S65-Z
3.25
3.53
3.10
2.43
3.60
3.18
S66-E
3.39
3.53
3.10
2.43
3.60
3.21
S67-Z
3.16
3.56
3.09
2.75
3.55
3.22
S71-Z
3.11
3.50
3.11
2.12
3.66
3.10
S72-E
3.63
3.50
3.11
2.12
3.66
3.20
S73-Z
3.13
3.53
3.10
2.43
3.60
3.16
S74-E
3.45
3.53
3.10
2.43
3.60
3.22
S75-Z
3.04
3.53
3.10
2.43
3.60
3.14
S76-E
3.37
3.53
3.10
2.43
3.60
3.20
S77-Z
3.00
3.56
3.09
2.75
3.55
3.19
S78-E
3.22
3.56
3.09
2.75
3.55
3.24
The retention parameters determined with five chromatographic systems were further correlated with in silico lipophilicity. The established linear relationships are presented in Fig. 3.
Correlation of in silico lipophilicity with hydrophobic parameter (log kw) of the analyzed stilbene derivatives.
https://binary.pensoft.net/fig/1780253
The obtained results indicate significant correlation between clog P and chromatographic lipophilicity for System A, C, and E, all using methanol in the mobile phase. The correlation express with p-value is very good – p < .05. Therefore, we accept these three systems can provide better prediction of the lipophilicity for the examined compounds. The correlation of clog P with chromatographic systems using acetonitrile in the mobile phase is moderate (System B, p ≈ .05) to poor (System D, p > .05).
Chemometrics of the chromatographic lipophilicity
Chemometric analysis uses mathematical, statistical, and machine learning tools to identify, classify, and interpret complex patterns within chemical datasets. HCA was applied to gain insight into the propinquity and dissimilarity between the chromatographic lipophilicity of the analyzed derivatives of stilbene (Fig. 4). The Hierarchical clustering method sets up clusters using a bottom-up iterative algorithm. First, a matrix of similarity (or dissimilarity) measures is created for each pair of objects. Then, individual objects are merged into clusters, and clusters are subsequently merged into superclusters. The final merge brings all objects into a single cluster (Ebeling et al. 2013). The statistically significant formation of distinct clusters was demonstrated by the Sneath-Sokal dissimilarity index (Sneath and Sokal 1973). The hierarchical structure of the data is visualized with a dendrogram. The height at which clusters are joined indicates their similarity; shorter distances mean higher similarity.
Hierarchical dendrogram based on the PCA of retention of 31 CA-4 derivatives with five chromatographic systems.
https://binary.pensoft.net/fig/1780254
There are two main clusters observable in the dendrogram. This dendrogram indicates a general, but not strict, separation between the substituents in ring B. Cluster I (slightly purple) contains the majority of compounds with styryl fragment in position 7 and 5 and some of the 4 and 6 derivatives. However, all compounds with 3,5-dimethoxy groups in ring A, S_3-Z and S_4-E, were classified in cluster I but 3,4-dimethoxy derivatives, S_5-Z and S_6-E, are in cluster II (yellowish brown) except S76-E. The analytes with a single methoxy group (S_7-Z / S_8-E) are distributed across both clusters. Most of the Z-isomers are in cluster II, which contains all Z-derivatives with three methoxy groups (S_1-Z) and 3,4-dimethoxy groups (S_5-Z). However, cluster I contains compounds with higher lipophilicity. They showed higher log kw values in most of the chromatographic systems. The analytes with lower lipophilicity are in cluster II.
PCA was applied in order to further explore the resemblance and dissimilarities between the lipophilicity results. PCA searches for a few uncorrelated linear combinations of the original variables. The graphical interpretation of the multidimensional space of the data set used by the PCA transformation is obtained with a smaller number of new dimensions of space.
One of the main criteria for choosing the optimal number of PCs is the Kaiser criterion, where the principal component has an eigenvalue greater than one (Kaiser 1960). Each principal component with an eigenvalue higher than one explains the larger part of the total variance, meaning it contributes to explaining the system better than a single original variable.
Unlike traditional PCA, which often uses arbitrary loading threshold greater than 0.50 or 0.75, we introduce a more robust approach by evaluating the significance of each loading based on its p-value at a 95% confidence level (p < .05). Given that factor loadings represent the correlation coefficients between original variables and the principal components. This method ensures that only statistically validated relationships inform the interpretation rather than subjective cut-off values.
Two principal components were extracted using criterion of Kaiser (Zagorska et al. 2015). They accounted for 41.98% and 23.69% of the total variation. The first principal component (PC1) is determined by all systems. PC1 can be characterized as a “silica” factor. System C and System D have a very low contribution to this component. The second principal component, PC2, is characterized as a “polymeric”. System C and System D are the chromatographic systems with wide pores and polymeric core of the stationary phase and they defined the PC2 component (Table 4). The score plot of the principal components combination with cluster separation are shown in Fig. 5. They provide information on the relation between the different data sets. The closer together the points are, the more similar the datasets. Two groups of compounds can be clearly distinguished. The HCA results strongly correlate with the PCA findings. The PCA exposes that compounds S44-E and S74-E can be distinguished from other derivatives. They have higher retention in all five systems. However, S44-E has linear retention model in System C, the new variable shows high lipophilicity.
Factor score biplot of principal components combined with HCA – cluster I (slightly purple) and cluster II (yellowish brown).
https://binary.pensoft.net/fig/1780255
Factor loadings, explained variance, and total variance for two principal components. Significant loadings (p-values < .05) are bold.
Chromatographic system
PC1
PC2
loading
p-value
loading
p-value
System A
0.826
1.03E-08
0.022
9.08E-01
System B
0.704
9.94E-06
0.011
9.52E-01
System C
0.395
2.79E-02
-0.766
5.13E-07
System D
0.389
3.06E-02
0.772
3.59E-07
System E
0.783
1.90E-07
-0.03
8.72E-01
Explained variance
41.98%
23.69%
Total variance
41.98%
65.67%
Conclusion
The present study characterizes the n-octanol / water distribution of 31 novel stilbene analogues using five chromatographic systems (log kw) and computational approaches (clog P). The strongest correlation between experimental and calculated data was observed in systems with silica-based stationary phase and a binary methanol-water mobile phase, System A and System E. Faster elution and reduced retention were observed in both acetonitrile – water systems, System B and System D. Notably, System D – a polymer matrix stationary phase with wider surface pore size, proved the most unsuitable for the analysis of the investigated derivatives. The obtained log kw values were evaluated through linear pattern recognition techniques (HCA and PCA), and consistent clustering patterns among the molecules were discussed. Comparative analysis indicated that most compounds with styryl fragment at position 7 or 5, as well as those with 3,5-dimethoxy groups in ring A, possess higher lipophilicity. In contrast, most Z-isomers – all derivatives with three methoxy groups (S_1-Z), and 3,4-dimethoxy groups (S_5-Z), exhibited lower lipophilicity. The results characterize the lipophilicity profiles of the studied compounds and offer a robust framework for future research into other physicochemical characteristics and their ADME properties.
Acknowledgements
The authors are grateful to the Distributed Research Infrastructure INFRAMAT, part of the Bulgarian National Roadmap for Research Infrastructures, supported by the Bulgarian Ministry of Education and Science, for providing the research equipment used in this investigation.
ReferencesAndrade J, Gomez-Carracedo M (2013) Notes on the use of Mandel’s test to check for nonlinearity in laboratory calibrations. Analitical Methods 5: 1145–1149. https://doi.org/10.1039/c2ay26400eApostolov S, Mekić D, Mitrović M, Petrović S, Vastag G (2025) Profiling of disubstituted chloroacetamides’ potential biological activity by liquid chromatography. Organics 6(3): 35–52. https://doi.org/10.3390/org6030035Baçzek T, Markuszewski M, Kaliszan R, van Straten M, Claessens H (2000) Linear and quadratic relationships between retention and organic modifier content in eluent in reversed phase high-performance liquid chromatography: a systematic comparative statistical study. Journal of High Resolution Chromatography 23(12): 667–676. https://doi.org/10.1002/1521-4168(20001201)23:12%3C667::AID-JHRC667%3E3.0.CO;2-PBukhari S, Kumar G, Revankar H, Qin HL (2017) Development of combretastatins as potent tubulin polymerization inhibitors. Bioorganic Chemistry 72: 130–147. https://doi.org/10.1016/j.bioorg.2017.04.007Casoni D, Kot-Wasik A, Namieśnik J, Sârbu C (2009) Lipophilicity data for some preservatives estimated by reversed-phase liquid chromatography and different computation methods. Journal of Chromatography A 1216(12): 2456–2465. https://doi.org/10.1016/j.chroma.2009.01.029Cimpan G, Irimie F, Gocan S, Claessens H (1998) Role of stationary phase and eluent composition on the determination of log P values of N-hydroxyethylamide of aryloxyalkylen and pyridine carboxylic acids by reversed-phase high-performance liquid chromatography. Journal of Chromatography B 714: 247 – 241. https://doi.org/10.1016/S0378-4347(98)00228-XCimpan G, Hadaruga M, Miclaus V (2000) Lipophilicity characterization by reversed-phase liquid chromatography of some furan derivatives. Journal of Chromatography A 869: 49–55. https://doi.org/10.1016/S0021-9673(99)01202-9Coutinho A, Cristofoletti R, Wu F, Al Shoyaib A, Dressman J, Polli J (2023) A robust, viable, and resource sparing HPLC-based log P method applied to common drugs. International Journal of Pharmaceutics 644: 123325. https://doi.org/10.1016/j.ijpharm.2023.123325Daina A, Michielin O, Zoete V (2017) SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Scientific Reports 7: 42717–42730. https://doi.org/10.1038/srep42717Danzer K, Currie L (1998) Guidelines for calibration in analytical chemistry Part 1. Fundamentals and single component calibration. Pure & Applied Chemistry 70(4): 993–1014. https://doi.org/10.1351/pac199870040993Ebeling B, Vargas C, Hubo S (2013) Combined cluster analysis and principal component analysis to reduce data complexity for exhaust air purification. Open Foods Science Journal 7(1): 8–22. https://doi.org/10.2174/1874256401307010008Elmansi H, Nasr J, Rageh H, El-Awady M, Hassan G, Abdel-Aziz H, Belal F (2019) Assessment of lipophilicity of newly synthesized celecoxib analogues using reversed-phase HPLC. BMC Chemistry 13: 84–90. https://doi.org/10.1186/s13065-019-0607-6Georgieva S, Voyslavov T, Chanev C, Petrov O (2025) High-performance liquid chromatography-based evaluation of lipophilicity of a series of novel stilbene derivatives. Pharmacia 72: 1–8. https://doi.org/10.3897/pharmacia.72.e141364Gerova M, Stateva S, Radonova E, Kalenderska R, Rusew R, Nikolova R, Chanev Ch, Shivachev B, Apostolova M, Petrov O (2016) Combretastatin A-4 analogues with benzoxazolone scaffold: Synthesis, structure and biological activity. European Journal of Medicinal Chemistry 120: 121–133. https://doi.org/10.1016/j.ejmech.2016.05.012Godinho J, Naese J, Toler A, Boyes B, Henry R, DeStefano J, Grinias J (2020) Importance of particle pore size in determining retention and selectivity in reversed phase liquid chromatography. Journal of Chromatography A 1634: 461678. https://doi.org/10.1016/j.chroma.2020.461678Gumieniczek A, Berecka-Rycerz A, Dul M, Pryjda A (2025) Standardized chromatographic and computational approaches for lipophilicity analysis of five gliflozin antidiabetic drugs in relation to their biological activity. Molecules 30(1): 1194–1216. https://doi.org/10.3390/biom15081194Hamze A, Alami M, Provot O (2020) Developments of iso Combretastatin A-4 derivatives as highly cytotoxic agents. European Journal of Medicinal Chemistry 190: 112110. https://doi.org/10.1016/j.ejmech.2020.112110Han SY, Liang C, Qiao JQ, Lian HZ, Ge X, Chen HY (2012) A novel evaluation method for extrapolated retention factor in determination of n-octanol/water partition coefficient of halogenated organic pollutants by reversed-phase high performance liquid chromatography. Analytica Chimica Acta 713: 130–135. https://doi.org/10.1016/j.aca.2011.11.020Ilijaš M, Malnar I, Marković V, Stepanić V (2013) Study of lipophilicity and membrane partition of 4-hydroxycoumarins by HPLC and PCA. Journal of Pharmaceutical and Biomedical Analysis 76: 104-111. https://doi.org/10.1016/j.jpba.2012.11.043Imramovsky A, Kozic J, Pesko M, Stolarikova J, Vinsova J, Kralova K, Jampilek J (2014) Synthesis and antimycobacterial and photosynthesis-inhibiting evaluation of 2-[(E)-2 substituted-ethenyl]-1,3-benzoxazoles. Scientific World Journal 2014: 705973. https://doi.org/10.1155/2014/705973ISO 8466-1:2021 (2021) Water quality – Calibration and evaluation of analytical methods – Part 1: Linear calibration function. https://www.iso.org/obp/ui/en/#iso:std:iso:8466:-1:ed-2:v1:enJeleń M, Morak-Młodawska B, Dołowy M, Konefał A (2025) Study of the lipophilicity of tetracyclic anticancer azaphenothiazines. Biomolecules 15(8): 1194–1216. https://doi.org/10.3390/biom15081194Kaiser H (1960) The application of electronic computers to factor analysis. Educational and Psychological Measurement 20(1): 141–151. https://doi.org/10.1177/001316446002000116Kamal U, Javed N, Arun K (2020) Biological potential of benzoxazole derivatives: an updated review. Asian Journal of Pharmaceutical and Clinical Research 13(8): 28–41. https://doi.org/10.22159/ajpcr.2020.v13i8.37958Kovačević S, Karadžić Banjac M, Podunavac-Kuzmanović S, Anojčić J, Ajduković J (2025) Comprehensive chemometric and chromatographic investigation of lipophilicity of biologically active androstane-3-oxime derivatives across diverse UHPLC systems. Pharmaceuticals 18(12): 1778–1795. https://doi.org/10.3390/ph18121778Leeson P, St-Gallay S (2011) The influence of the ‘organizational factor’ on compound quality in drug discovery. Nature Reviews Drug Discovery 10: 749–765. https://doi.org/10.1038/nrd3552Lipinski C, Lombardo F, Dominy B, Feeney P (2001) Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Advanced Drug Delivery Reviews 46: 3–26. https://doi.org/10.1016/S0169-409X(00)00129-0Lobo S (2020) Is there enough focus on lipophilicity in drug discovery? Expert Opinion on Drug Discovery 15(3): 261–263. https://doi.org/10.1080/17460441.2020.1691995Matthews H, Hanison J, Nirmalan N (2016) “Omics”-informed drug and biomarker discovery: opportunities, challenges and future perspectives. Proteomes 4(3): 28–40. https://doi.org/10.3390/proteomes4030028Nasal A, Siluk D, Kaliszan R (2003) Chromatographic retention parameters in medicinal chemistry and molecular pharmacology. Current Medicinal Chemistry 10: 381–426. https://doi.org/10.2174/0929867033368268OECD (1995) Test No. 107: Partition Coefficient (n-octanol/water): Shake Flask Method, OECD Guidelines for the Testing of Chemicals, Section 1, OECD Publishing, Paris. https://doi.org/10.1787/9789264069626-enOECD (2022) Test № 117: Partition coefficient (n-octanol/water), HPLC Method, OECD Guideline for the Testing of Chemicals section 1, OECD Publishing, Paris. https://doi.org/10.1787/9789264069824-enOECD (2025) World health statistics 2025: monitoring health for the SDGs, Sustainable Development Goals. Geneva: World Health Organization; 2025. Licence: CC BY-NC-SA 3.0 IGO. http://www.wipo.int/amc/en/mediation/rules/Poole S, Poole C (2003) Separation methods for estimating octanol–water partition coefficients. Journal of Chromatography B 797: 3–19. https://doi.org/10.1016/j.jchromb.2003.08.032R Core Team (2025) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/Redka M, Baumgart S, Kupczyk D, Kosmalski T, Studzińska R (2023) Lipophilic studies and in silico ADME profiling of biologically active 2-aminothiazol-4(5H)-one derivatives. International Journal of Molecular Sciences 24(15): 12230–12246. https://doi.org/10.3390/ijms241512230Silakari O, Sindh P (2021) ADMET tools: Prediction and assessment of chemical ADMET properties of NCEs. In: Concepts and Experimental Protocols of Modelling and Informatics in Drug Design. Academic Press, 299–320. https://doi.org/10.1016/B978-0-12-820546-4.00014-3Simoens S, Huys I (2021) R&D Costs of New Medicines: A Landscape Analysis. Frontiers in Medicine 8: 760762. https://doi.org/10.3389/fmed.2021.760762Sneath A, Sokal R (1973) Numerical Taxonomy. The Principles and Practice of Numerical Classification, 1st Edn. WF Freeman & Co., San Francisco.Tache T, Naşcu-Briciu R, Sârbu C, Micăle F, Medvedovici A (2012) Estimation of the lipophilic character of flavonoids from the retention behavior in reversed phase liquid chromatography on different stationary phases: A comparative study. Journal of Pharmaceutical and Biomedical Analysis 57: 82–93. https://doi.org/10.1016/j.jpba.2011.08.044Valkó K (2004) Application of high-performance liquid chromatography based measurements of lipophilicity to model biological distribution. Journal of Chromatography A 1037: 299–310. https://doi.org/10.1016/j.chroma.2003.10.084Waring M (2010) Lipophilicity in drug discovery. Expert Opinion on Drug Discovery 5(3): 235–248. https://doi.org/10.1517/17460441003605098Wolter F, Stein J (2002) Biological activities of resveratrol and its analogs. Drug of the future 27(10): 949–959. https://doi.org/10.1358/dof.2002.027.10.856987Zagorska A, Czopek A, Pelka K, Bajda M, Stanisz-Wallis K, Pawlowski M (2015) Reversed-phase high-performance liquid chromatography study of lipophilicity of imidazo[2,1-f]theophylline derivatives. Acta Poloniae Pharmaceutica - Drug Research 72(4): 663–669.Additional informationConflict of interest
The authors have declared that no competing interests exist.
Ethical statements
The authors declared that no clinical trials were used in the present study.
The authors declared that no experiments on humans or human tissues were performed for the present study.
The authors declared that no informed consent was obtained from the humans, donors or donors’ representatives participating in the study.
The authors declared that no experiments on animals were performed for the present study.
The authors declared that no commercially available immortalised human and animal cell lines were used in the present study.
Artificial Intelligence (AI) use
The authors accept full responsibility for the content of the manuscript, including the disclosure of any use of AI.
No AI tools were used in the preparation of this manuscript.
Funding
No funding was reported.
Author contributions
All authors have contributed equally.
Author ORCIDs
S. Georgieva https://orcid.org/0000-0003-0007-2509
T. Voyslavov https://orcid.org/0000-0002-4395-9954
C. Chanev https://orcid.org/0000-0003-4705-2951
O. Petrov https://orcid.org/0000-0002-5541-541X
Data availability
All of the data that support the findings of this study are available in the main text.