Evaluation of antimicrobial activity and retention behavior of newly synthesized vanilidene derivatives of Meldrum’s acids using QSRR approach

Accepted Manuscript Title: Evaluation of antimicrobial activity and retention behavior of newly synthesized vanilidene derivatives of Meldrum’s acids using QSRR approach Authors: Jovana Trifunovi´c Ristovski, Nenad Jankovi´c, Vladan Borˇci´c, Sankalp Jain, Zorica Bugarˇci´c, Momir Mikov PII: DOI: Reference:

S0731-7085(18)30411-4 https://doi.org/10.1016/j.jpba.2018.03.038 PBA 11869

To appear in:

Journal of Pharmaceutical and Biomedical Analysis

Received date: Revised date: Accepted date:

15-2-2018 15-3-2018 16-3-2018

Please cite this article as: Jovana Trifunovi´c Ristovski, Nenad Jankovi´c, Vladan Borˇci´c, Sankalp Jain, Zorica Bugarˇci´c, Momir Mikov, Evaluation of antimicrobial activity and retention behavior of newly synthesized vanilidene derivatives of Meldrum’s acids using QSRR approach, Journal of Pharmaceutical and Biomedical Analysis https://doi.org/10.1016/j.jpba.2018.03.038 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Evaluation of antimicrobial activity and retention behavior of newly synthesized vanilidene derivatives of Meldrum’s acids using QSRR approach Jovana Trifunović Ristovski*1, Nenad Janković2, Vladan Borčić1, Sankalp Jain3, Zorica Bugarčić2 and Momir Mikov1 1

2

Department of Chemistry, Faculty of Science, University of Kragujevac, Serbia

3

SC R

Department of Pharmaceutical Chemistry, University of Vienna, Austria

Corresponding author: Jovana (Trifunović) Ristovski e-mail: [email protected]

U

[email protected]

N

phone: +381 21 522 172 Highlights



A

CC E

Abstract

M



ED



Lipophilicity and antimicrobial activity of 13 vanilidene derivatives of Meldrum´s acid has been evaluated. Clustering of the compounds according to their lipophilicity was performed with the help of principal component analysis (PCA) and hierarchical cluster analysis (HCA). ADME properties of newly synthetized derivatives were analyzed by applying sum of ranking differences (SRD) method. Statistically significant predictive models of antimicrobial activity were obtained using multiple linear regression (MLR) analysis.

PT



IP T

Department of Pharmacology, Toxicology and Clinical Pharmacology, Faculty of Medicine, University of Novi Sad, Serbia

A

Increased antimicrobial resistance together with the lack of new antimicrobial drugs suggest on an urgent need for new therapeutics in this field. Vanilidene derivatives of Meldrum´s acid present one of the possible approaches. In this work lipophilicity of 13 vanilidene derivatives of Meldrum´s acid as well as their predicted antimicrobial activity towards several characteristic species has been evaluated. 10 vanilidene derivatives have been previously synthesized and 3 new compounds are synthetized afterwards following the same procedure. These selected 13 candidates were examined using thin layer chromatography in two different solvent systems. Gained retention parameters were a starting point for further Quantitative Structure Property Relationships (QSRR) studies in which minimum inhibitory concentration (MIC) for Candida albicans, Trichoderma viride, Penicillium italicum, Fuscarium oxysporum, Pseudomonas aeruginosa

and Escherichia coli were determined. Statistically significant QSRR models were established and clustering of the compounds was performed with the help of principal component analysis (PCA) and hierarchical cluster analysis (HCA). Absorption, Distribution, Metabolism, and Excretion (ADME) properties of investigated molecules were subjected to sum of ranking differences (SRD) analysis in order to explore their pharmacokinetic properties. SRD analysis was also performed for the ranking of the established QSRR models.

IP T

It was shown that compounds 6, 8 and 9 possess a significant antimicrobial activity, satisfied ADME properties and these candidates should be further optimized in order to utilize unexplored potential of Meldrum's acid in synthesis of novel antifungal compounds. Keywords: chromatography; Meldrum’s acid; retention factor; QSRR; SRD; HCA; PCA

SC R

1. Introduction

M

A

N

U

Meldrum's acid (2,2-Dimethyl-4,6-dioxo-1,3-dioxane) was discovered by A.N. Meldrum in 1908. It was synthesized using malonic acid and acetone as starting compounds. Considering its high acidity, Meldrum's acid can serve as a reactant in Knoevenagel condensations [1]. Although more than 100 years has passed since Meldrum´s acid was discovered it still represents very attractive compound in organic synthesis [2,3]. Chemical profile allows Meldrum's acid to serve as a scaffold in the synthesis of vast range of compounds with various functional groups. It has been previously shown that various synthetized derivatives often possess different and beneficial pharmacological properties such as antimalarial, antioxidant and antimicrobial activity [4]. These data suggest possible new roles of derivatives of Meldrum´s acid which have not been fully exploited.

CC E

PT

ED

Modern techniques and computer based approach have utilized the use of Quantitative Structure Activity Relationships (QSAR) parameters in the potential development of new drugs. One of the principal QSAR parameter of drugs and xenobiotics is their lipophilicity. Lipophilicity has a significant impact on drug pharmacokinetics and influence on passive diffusion through biological barriers [5]. This feature strongly affects the protein binding, membrane penetration, toxicological properties and adverse effects. Beside important role in achieving pharmacological effects, lipophilicity shows influence on solubility, stability and drug formulations. In rational drug design partition coefficient often represents a measure of lipophilicity which can be further used as a starting point in development of significant molecular models.

A

Mutual combination of analytical techniques such as thin layer chromatography (TLC) together with the QSAR approach present a strong tool in which physicochemical properties of novel compounds are used as predictors for the pharmacological properties which then allows a cost effective change of compounds by in silico modeling [6]. In addition to that TLC presents very useful method for reducing costs and time of analysis; it provides lower usage of solvents and possibility to concurrently handle several samples [7,8]. One other benefit is that TLC can be performed without using robust instrumentation which makes it available for wide application in lipophilicity estimation. Next step is the development of the quantification relation between retention and pharmacological potential. Quantitative Structure Property Relationships (QSRR)

enables comprehension of the role of lipophilicity on pharmacological effects and can serve to help understanding of intramolecular relations between functional groups as well as in describing intermolecular interactions. Gained molecular models could be cross validated and the examined substances could be clustered so that only the compounds with the highest ratio of desired activities remain emphasized. Approach in which lipophilicity was used as a first step towards prediction of antimicrobial activity was already published before [6].

IP T

In this work 13 novel vanilidene derivatives of Meldrum's acid were examined. All of the investigated derivatives were synthesized according to solvent free synthesis in presence of 5 mol% p-toluene sulfonic acid [7]. By using molecular descriptors and QSRR approach this work attempts to utilize chemometric regression tools in prediction of lipophilicity and antimicrobial activity of investigated vanilidene derivatives of Meldrum's acid.

U

SC R

The aim of this QSRR study was to evaluate retention data using multivariate statistic and to establish possible relation between retention and physicochemical parameters of compounds as well as relation between retention factors and (minimum inhibitory concentration) MIC values of examined vanilidene derivatives in order to understand the possible effects of lipophilicity on the antimicrobial activity towards several different species.

N

2. Materials and methods

A

2.1. Investigated vanilidene derivatives

ED

M

In this work sum of total 13 different vanilidene derivatives of Meldrum’s acid were used for the establishment of molecular models. Ten compounds were previously synthetized and their antimicrobial activities against the tested microorganisms are already published [8] while molecules 6, 8 and 10 are synthesized afterwards and their NMR spectra as well as antimicrobial properties are given in supplementary data (S1). All compounds were synthesized according to the procedure of solvent-free synthesis previously published by Jankovic et al. [8]

PT

The chemical structures of 13 examined vanilidene derivatives used in this study are presented in Figure 1.

CC E

2.2. Lipophilicity estimation

A

Chromatography experiments were performed using thin layer chromatography (NP TLC) in two different solvent systems chlorophorm/ethyl acetate and chlorophorm/diethylether. Mobile phases have been chosen based on the previously published data [8]. The mobile phases were prepared by mixing the certain volume of chlorophorm and EtOAc or Et2O in the range from 5 to 25% (v/v) in 5% increments. Experiments were done using 20x20 cm Silica Gel coated on Al Foil Sheets 200 μm (Analtech). The chamber was prior saturated with mobile phase for 30 min. Investigated compounds were dissolved in chloroform (2 mg/cm3) and 1 μl aliquot of each solution was spotted on the plate using a micropipette. After development, plates were dried and spots were detected under the UV light (254 nm). Every TLC analysis was performed in triplicate under the equal conditions of temperature and humidity. No significant statistical differences of the retention factor (Rf) were observed and Rf were calculated as average values. The retention constants (RM) were calculated using following equation:

RM=log(1/Rf-1) (1) Rf represents retention factor and RM values linearly depend on the logarithm of concentration organic modifier in the mobile phase according to Soczewinski [9]: RM=RM0+blogC (2) RM0 represents intercept (Table 1), b is slope (Table 1) of the linear plot and C is the concentration of the organic solvent (in %) in the mobile phase (5-25%).

IP T

Lipophilicity parameter, C0, was calculated using following equation: C0= RM0/b (3)

U

SC R

Descriptors used in this study were selected on the basis of Pearson correlation coefficient of each descriptor with appropriate MIC and RM0 values and on the basis of the distribution of data on the scatter plot. In this work several different softwares were used in order to obtain logP values (www.molinspiration.com, http://preadmet.bmdrc.kr/, ChemDraw Prime 15.1, http://bleoberis.bioc.cam.ac.uk/pkcsm/prediction, https://ilab.acdlabs.com/iLab2/). Calculated logP values are presented in Table 2.

N

2.3. QSRR analysis of studied compounds

A

CC E

PT

ED

M

A

In QSRR analysis several chemometric tools were applied. In order to identify similarities or dissimilarities between compounds following classification methods were used: hierarchical cluster analysis (HCA) and principal component analysis (PCA). HCA represents very useful tool for recognition of clusters of similar objects. By using this tool similar objects are placed in the same cluster. Usually the distance between objects is presented as Euclidian distances [10]. PCA shows some advantages compared to HCA. It can indicate which properties objects share in the variable space. This chemometric method can help in reducing descriptors when correlation is presented among the variables. In order to identify similarities or dissimilarities between vanilidene derivatives PCA analysis based on calculated physicochemical descriptors was conducted. Selection of the suitable molecular descriptors was performed by stepwise selection (SS) procedure. Regression analysis was conducted by using: linear regression (LR), multiple linear regression (MLR) and partial least squares regression (PLS) [11]. Using experimentally obtained MIC values (Candida albicans, Trichoderma viride, Penicillium italicum, Fuscarium oxysporum, Pseudomonas aeruginosa and Escherichia coli), calculated descriptors (partition coefficient, TPSA- total polar surface area and Mw – molecular weight) and experimentally obtained retention values adequate MLR models have been established. MIC values for compounds 6, 8 and 10 are experimentally obtained using previously published procedure [7]. All calculations were performed using OriginPro 2016 and SPSS 2017 software.

3. Results 3.1. Chromatographic behaviour Chromatographic methods represent very reliable techniques for experimental calculations of retention factors which can be further used as starting points for QSAR analysis [9,12].

Different chromatographic conditions are very often used for most precise lipophilicity estimation. The polarity of the mobile phase can be adjusted by mixing different amounts of organic solvents in order to obtain mobile phase with desired characteristic (Figure 2). The eluting power of a solvent system mainly increases as the solvent becomes more polar. Linear dependence of RM0 (intercept) from b (slope) or C0 has been described by adequate functions:

RM0(Ch/Et2O)= 1,151b(Ch/Et2O)+0,879 (r2=0,925; F=148,999; p=9,759x10-8) RM0(Ch/EtOAc)=-1,046C0+0,138 (r2=0,713; F=27,351; p=2,812x10-4)

SC R

RM0(Ch/Et2O)=-0,097C0+0,15 (r2=0,746; F=32,314; p=1,413x10-4)

IP T

RM0(Ch/EtOAc)= 1,449b(Ch/EtOAc)-0,929 (r2=0,971; F=379,725; p=7,063x10-10)

M

A

N

U

Adequate statistical methods could be used in order to obtain suitable mathematical correlation involving small to large number of variables. Various regression and classification based methods can be employed in realization of this purpose. Regression-based methods can be applied when chemical data have numerical values while qualitative values of chemical compounds can be modeled using classification techniques. The purpose of regression analysis is to explain the outcome of one or more independent variables on a dependent variable and that is the reason why it became one of the most commonly used techniques for generating models in order to explain effects of xenobiotics on biological systems. Having in mind all mentioned above in this work two regression techniques (LR and MLR) were used as appropriate chemometric tools.

CC E

PT

ED

LR is the simplest regression technique which can be employed in establishing relations between dependent variable and one independent variable. MLR represents one of the most widely represented regression methods that have been used in QSAR, QSPR and QSRR. This regression method is often applied for quantification of the relations between more than one independent variables and a dependent variable by fitting a linear equation to observed data. In our research, MLR analysis was performed to evaluate the suitability of experimentally obtained RM0 and MIC values [11,13]. Experimentally obtained retention constants of vanilidene derivates were used in different calculations in order to obtain correlation between these constants and standard lipophilicity parameter logP (Table 3). Linear dependence was confirmed which can be observed by strong r2 values presented in the Table 3.

A

The very good correlation coefficients confirm the suitability of the examined mobile phase for estimating the lipophilicity of the vanilidene derivatives. Considering lipophilicity as a complex parameter, there are different other descriptors that are valuable in determination of correlation between chromatography factors and lipophilicity. In the literature has been widespread that theoretically calculated partition coefficients are in correlation with experimentally obtained lipophilicity parameters [14]. Moreover, the correlations between different logP values and slope (b) were calculated and the results are presented in Table 4. Observing the correlations between two parameters RM0 and b with different logP values it can be concluded that values

obtained using TLC can be successfully employed to express lipophilicity of all the examined compounds.

IP T

RM0 values are widely used as alternative to logP constants obtained by shake-flask procedure, because partitioning between 1-octanol and aqueous phase in chromatography often seems to be similar to partitioning in biological membranes. Shake-flask method represents time consuming and inconvenient technique because it requires a very long time for equilibration especially for large series of newly synthesized compounds. To overcome these drawbacks TLC can be considered as a chromatography technique that represents alternative to traditional determination of lipophilicity [15]. It provides sufficiently precise RM0 values which are of special importance when pharmacological activity of molecules represents primary goal.

SC R

3.2. The SRD (standard reference data) analysis of ADME properties of the selected Meldrum’s acid derivatives.

CC E

PT

ED

M

A

N

U

ADME properties are widely used parameters for prediction of the behaviour of molecules of interest. In this study gained ADME properties of selected Meldrum’s acid derivatives were subjected to SRD analysis. Obtained results were meant to enable the specific clusterization of the examined compounds on the basis of their similar ADME characteristics. The SRD modeling is considered to be entirely general implying that a known reference ranking (“golden standard”) is established [16]. Before the SRD analysis was performed, all the available data were arranged in a matrix form und subjected to the scaling between 0 and 100. In the matrix the objects (blood-brain barrier penetration (BBB), oral absorption expressed by Caco2 and Mandin-Darby canine kidney (MDCK) cells permeability, human intestinal absorption (HIA), plasma protein binding (PPB), skin permeability (SP) are placed in the rows and the variables (the compounds to be compared) are listed in the columns. All of the calculated ADME descriptors are presented in the Supplementary data (S1). For the reference ranking the average BBB, CaCo2, MDCK, PPB, SP and HIA row values were chosen. If the average is chosen as the golden standard, the SRD will actually measure the differences from the center and can be considered the method for measuring the similarity or dissimilarity [16, 17]. It is generally considered that the SRD values close to the zero indicate a very strong and accurate model. In a situation where two or more compounds have similar or even the same SRD values they are generally similar and therefore close proximity indicates similitude. All of the written above implies that SRD analysis can give two very important outcomes – one is the clustering of compounds and the other one is the dissimilarity measure between the examined compounds. [16, 17, 18].

A

Results presented on the Figure 3 can lead to a conclusion that compounds 2-11 have very similar ADME characteristics and they are the best positioned compounds on the chart. The similar trend is show in the aspect of antimicrobial properties which is a good indicator suggesting that the most potent compounds also possess best ADME characteristics. Compound 1 stands out from this group and it is on the last position, while the compounds 12 and 13 are located in the middle. 3.3. Application of HCA and PCA methods on the set of vanilidene derivatives

PCA method was applied on molecular descriptors and retention factors to reveal some similarities/dissimilarities among vanilidene derivatives and to recognize suitable descriptors for further analysis. PCA represents a very helpful method for reducing the amount of information when correlation between variables is presented. In the PCA test, different patterns of distribution were gained by plotting scores of the descriptive parameters of vanilidene derivatives related to logP values against each other and the results are presented on Figure 4.

N

U

SC R

IP T

Application of this method on retention parameters can disclose some similarities between molecules that are determined by their structural characteristics and specific interactions in various chromatographic systems. PCA was performed using different partition coefficient descriptors of studied compounds, their retention factors, polar surface area and molecular weight (RM0, TPSA, Mw). Considering significant correlation between variables, two first principal components were sufficient in representing the data variability while only small part of information was neglected [19]. Different partition coefficients have the biggest influence on the positive effects on the PC1. Score plot (Fig .4) show similarities between compounds and their lipophilicity. Calculated logP values are responsible for distribution of compounds on score plot. Two separated groups of examined vanilidene derivatives can be spotted. PCA results suggest that the group which is positioned left on score plot possesses lower values of partition coefficients.

ED

M

A

HCA analysis was applied in order to confirm specific clustering of molecules already observed using the PCA [20]. Dendogram of HCA is shown in the Figure 5. Method for clustering relies on Ward’s linkage and Euclidean distance. Descriptors used for this statistical analysis were the same one used for PCA and the outcome presents cluster of examined vanilidene derivatives which show their similarities and differences. Presented dendogram contains 2 clusters of examined vanilidene derivatives: cluster A consists of molecules which possess halogen atom while cluster B group contains compounds without halogen. It is very well known that halogen atoms inside molecule significantly influence retention characteristics of molecules [21].

CC E

PT

All molecular descriptors were calculated using suitable software for molecular design. Three descriptors with low intercorrelation (VIF values for logP, TPSA, Mw were respectively 1,63; 1,47; 1,38) were chosen for building MLR models.

4. Discussion

A

Lipophilicity is shown to correlate with numerous other descriptors, such as molecular weight, volume, polar surface area, parachor etc. [8, 12, 14]. Calculated lipophilicity descriptors (ClogP, milogP, SKlogP, pklogP and ACDilogP) are applied in prediction of chromatographic behaviour, antimicrobial and physicochemical properties of compounds. Multicollinearity is generally accepted in PCA and PLS modeling. Beside the limitations of MLR, this method could be very valuable in defining corresponding chromatographic phenomena and it is widely applied in QSRR analysis [12, 22, 23]. A complete MLR analysis was performed using milogP, TPSA and Mw as independent variables. The models with best statistical parameters are represented through the equations in Table 5. Chemometric approach offers an alternative procedure in optimization different biological characteristics by considering common

interactions between variables and gives an assessment of combined effect of these variables on the ultimate result. In established MLR analysis (Table 6) for predicting adequate MIC values there is more than one independent variable and chosen variables are not mutually correlated. The intercorrelation needs to be further confirmed in order to increase the significance of the gained models. For that purpose the variance inflation factor (VIF) was calculated. If the VIF is equal or less then 10, the multicollinearity can be rejected (RM0(Ch/EtOAc), TPSA, Mw; RM0(Ch/Et2O), TPSA, Mw respectively 8,46; 7,13; 1,73; 6,24; 4,72; 1,98).

SC R

IP T

The PLS method was applied for the calculation of Variable Importance in the Projection (VIP) and for the establishment of the QSRR models. Importance of each variable has been presented as the VIP parameter. Among all molecular properties that were included in generating the QSRR models, only the variables with the VIP values greater than 0.5 have been considered for the regression.

ED

M

A

N

U

Good predictive ability of the MLR models is confirmed throughout leave-one out (LOO) crossvalidation. This validation is based on exclusion of each sample once thereby constructing new model without this sample. Using LOO cross-validation it becomes possible to predict values of dependent variable for each number of factors. Standard statistical parameters (Table 5 and Table 6) include: coefficient of determination (R2), adjusted determination coefficient (R2adj), Fisher’s test (F) and level of significance (p). Gained molecular models present good predictors of the possible antibacterial and antifungal effects of the examined components. The compound with the strongest antimicrobial activity was compound 6 followed by compounds 8 and 9. From the entire examined pathogen group statistically most significant models were developed for Aspergillus flavus and Candida albicans which suggest further need for investigation of potentially important antifungal properties of investigated compounds. Lack of new antifungal drugs and high ratio of resistance to the modern therapeutics presents a solid field of research on which vanilidene derivatives could present a new direction in the drug development process.

A

CC E

PT

In order to rank the obtained MLR models the base with experimentally observed MIC values of the examined compounds was used as a referent rank. The goal was to notice the best possible correlation between established MLR models and experimentally gained datasets. The ranking was validated by comparison of ranks by random numbers procedure. Gained models were classified by their increasing predicted MIC values. In case that a certain value deviate from the referent ideal rank, the model connected to that value will be positioned from an ideal rank for a certain degree. Based on results shown in Figure 6 and Supporting Information Table 4, MLR4, MLR5, MLR12 and MLR13 are the best ranked. Models MLR2, MLR3 MLR10 and MLR11, also have the same rank and they are farthest from the referent rank than all other models. Since the models MLR4 and MLR5 gave the best data fit, their position that is nearest to the referent rank is justified. SRD method was used to select the best model for prediction of MIC values.

5. Conclusions

SC R

IP T

A series of thirteen vanilidene derivatives have been examined. Thin layer chromatography was employed for determination of retention factors. Different logP values were obtained using several software and linear correlation between retention constants and lipophilicity was established. Incorporation of milogP, TPSA and Mw values have significant effects on model predictability. The results suggest that minimum inhibitory concentrations of vanilidene compounds depend of lipophilicity as well as on TPSA and Mw of the molecule. Considering the results gained in this study, all of the examined compounds show moderate antimicrobial/antifungal activity. The best antimicrobial potential possesses compound 6 which contains carboxylic group. By introducing the ester group biological activity is reduced which is observed by compound 7. Also, compounds 8 and 9 show good activity which might be consequence of bromine presence in the molecule. On the other hand compounds 12 and 13 also contain bromine atom but attached at end of the longer carbon chain (n=3-4). Results gained in this study suggest that these compounds could be further optimized in order to utilize unexplored potential of Meldrum's acid in synthesis of novel antifungal compounds but the wise preselection of candidates needs to be performed.

U

Conflict of interest statement Authors declare no conflict of interest

N

Acknowledgements

M

A

This research was supported by The Ministry of Science and Technological development, Republic of Serbia Project No III 41012 and 172011. J.T.R. would like to thank to Pharmacoinformatics Research Group from University of Vienna for research stay.

ED

References

PT

[1] H.S. Sandhu, S. Sapra, M. Gupta, K. Nepali, R. Gautam, S. Yadav, S. Kumar, S.M. Jachak, M. Chugh, M.K. Gupta, O.P. Suri, K.L. Dhar, Synthesis and biological evaluation of arylidene analogues of Meldrum’s acid as a new class of antimalarial and antioxidant agents, Bioorganic. Med. Chem. 18 (2010) 5626–33.

CC E

[2] J.A. Bull, J.J. Mousseau, G. Pelletier, A.B. Charette, Synthesis of Pyridine and Dihydropyridine Derivatives by Regio- and Stereoselective Addition to N-Activated Pyridines, Chem Rev. 112 (2012) 2642–2713.

A

[3] G.M. Sampaio, A.R. Teixeira, H.M. Coutinho, D.M. Sena, P.C. Freire, R.F. Bento, L.E. Silva, Synthesis and antibacterial activity of a new derivative of the Meldrum acid: 2,2dimethyl-5-(4H-1,2,4-triazol-4-ylaminomethylene)-1,3-dioxane-4,6-dione, Excli. J. 13 (2014) 1022–28. [4] Y. Oikawa, K. Sugano, O. Yonemitsu, Meldrum's acid in organic synthesis. 2. A general and versatile synthesis of beta-keto esters, J. Org. Chem. 43 (1978) 2087–88.

[5] J. Trifunović, V. Borčić, S. Vukmirović, M. Mikov, Assessment of the pharmacokinetic profile of novel s-triazine derivatives and their potential use in treatment of Alzheimer disease, Life Sci. 168 (2017) 1-6. [6] C.G. De Almeida, G.D. Garbois, L.M. Amaral, C.C. Diniz, M. Le Hyaric, Relationship between structure and antibacterial activity of lipophilic N-acyldiamines, Biomed Pharmacother. 64 (2012) 287-290.

IP T

[7] M. Starek, L. Komsta, J. Krzek, Reversed-phase thin-layer chromatography technique for the comparison of the lipophilicity of selected non-steroidal anti-inflammatory drugs, J. Pharm. Biomed. Anal. 85 (2013) 132-137.

SC R

[8] N. Janković, J. Muškinja, Z. Ratković, Z. Bugarčić, B. Ranković, M. Kosani, Solvent-free synthesis of novel vanillidene derivatives of Meldrum's acid: biological evaluation, DNA and BSA binding study Stefanović, RSC Adv. 6 (2016) 39452-459.

[9] E. Soczewinski, C.A. Wachtmeister, The relation between the composition of certain ternary two-phase solvent systems and RM values, J. Chromatogr. 7 (1962) 311–20.

N

U

[10] P. Gaillard, P.A. Carrupt, B. Testa, A. Boudon, Molecular lipophilicity potential, a tool in 3D QSAR: method and applications, J. Comput. Aided Mol. Des. 8 (1994) 83-96.

M

A

[11] L. Jevrić, B. Jovanović, S. Velimirović, A. Tepić, G. Koprivica, N. Mišljenović, Application of lipophilicity parameters in QSRR analysis of newly synthesized s-triazine derivatives – prediction of the retention behaviour, Chem. Ind. 65 (2011) 533–40.

ED

[12] J. Trifunović, V. Borčić, S. Vukmirović, M. Mikov, Structural insights into anticancer activity of D-ring modified estrone derivatives using their lipophilicity in estimation of SAR and molecular docking studies, Drug Test Anal. 10 (2017) 1542-1548.

PT

[13] J.R. Greenwood, D. Calkins, A.P. Sullivan, D.C. Shelley, Towards the comprehensive, rapid, and accurate prediction of the favorable tautomeric states of drug-like molecules in aqueous solution, J. Comput. Aided Mol. Des. 24 (2010) 591-604.

CC E

[14] J. Trifunović, V. Borčić, S. Vukmirović, S.G. Kon, M. Mikov, Retention data of bile acids and their oxo derivatives in characterization of pharmacokinetic properties and in silico ADME modeling, Eur. J. Pharm. Sci. 96 (2016) 194-202.

A

[15] P. Buchwald, N. Bodor, Computer-aided drug design: The role of quantitative structureproperty, structure-activity, and structure-metabolism relationships (QSPR/QSAR/QSMR), Drug Fut. 7 (2002) 577–588. [16] K. Héberger, K. Kollár-Hunek, Sum of ranking differences for method discrimination and its validation: Comparison of ranks with random numbers, J. Chem. 25 (2001) 151–158. [17] K. Héberger, Sum of ranking differences compares methods or models fairly, TRACTrends Anal. Chem. 29 (2010) 101–109.

[18] K. Héberger, B. Škrbić, Ranking and similarity of quantitative structure-retention relationship models in predicting Lee retention indices for polycyclic aromatic hydrocarbons, Anal. Chim. Acta. 716 (2012) 92–100. [19] M. Jalali-Heravi, A. Kyani, Use of computer-assisted methods for the modeling of the retention time of a variety of volatile organic compounds: A PCA-MLR ANN Approach, J Chem. Inf. Comput. Sci. 44 (2014) 1328–35.

IP T

[20] R. Kaliszan, QSPR: Quantitative Structure-(Chromatographic) Retention Relationships, Chem. Rev. 107 (2007) 3212–46. [21] K. Kowalski, On the Predictive Performance of Biased Regression Methods and Multiple Linear Regression, Chemometrics and Intelligent Laboratory Systems, 9 (1990) 177-84.

SC R

[22] K. Ciura, S. Dziomba, S. Nowakowska, M. Markuszewski, Thin layer chromatography in drug discovery process, J. Chrom. A. 1520 (2017) 9-22. Tables

RM0(CHC

RM0(CHCl3/Et2O)

b(CHCl3/EtOAc)

l3/EtOAc)

C0(CHCl3/EtOAc)

C0(CHCl3/Et2O)

-0,5723

-0,34584

-0,42478

b(CHCl3/Et2O)

N

Compound

U

Table 1. Experimentally obtained lipophilicity values used in correlation with different logP values.

0,3439

0,2431

-0,9944

2

0,28

0,1944

-0,7892

-0,5104

-0,35479

-0,38088

3

0,113

0,0765

-0,7541

-0,7254

-0,14985

-0,10546

4

0,2775

0,2064

-0,9398

-0,5853

-0,29528

-0,35264

5

1,35

0,7066

-1,5352

-0,193

-0,87936

-3,66114

6

1,8534

0,883

-1,89

-0,1

-0,98063

-8,83

7

1,2039

0,4361

-1,4421

-0,421

-0,83482

-1,03587

8

0,4291

0,1805

-0,9752

-0,5115

-0,44001

-0,35288

9

0,4042

0,2107

-1,006

-0,5446

-0,40179

-0,38689

10

0,5663

0,3048

-0,9453

-0,4725

-0,59907

-0,64508

11

0,35

0,2701

-0,84

-0,4497

-0,41667

-0,60062

12

0,15

0,0603

-0,6982

-0,7123

-0,21484

-0,08466

13

-0,33

-0,2021

-0,3501

-1,03

0,942588

0,196214

M

ED

PT

CC E A

A

1

IP T SC R U N

ClogP

milogP

SKlogP

pklogP

ACDilogP

1

3,6194

3,51

3,00197

2,7

2,38

2

3,3994

3,37

2,85168

2,7019

2,2

3

4,1484

4,07

3,45594

3,0936

2,66

4

3,8644

3,53

3,27052

2,8696

2,53

1,9914

2,13

1,82773

1,84

1,9

1,85

1,88

1,89859

1,3781

0,82

7

2,7086

2,87

2,25908

1,8566

1,8

8

3,2934

3,37

2,9494

2,6884

2,49

9

3,6224

3,64

3,17265

3,0785

2,67

10

3,3354

3,27

2,85111

2,4795

2,09

11

3,0803

3,56

2,68407

2,2249

2,18

12

4,0014

3,91

3,5867

3,4686

2,88

13

4,5304

4,41

4,04067

3,8587

3,61

A

ED

CC E

6

PT

5

M

Compound

A

Table 2. Calculated logP values for different vanilidene Meldrum’s acid derivatives.

IP T

SC R

Table 3. Linear dependence of retention constants R M0 from different logP values. F

p

0,898

107,063

5,251x10-7

0,92

139,44

1,372x10-7

0,935

172,697

4,547x10-8

0,957

267,1

4,606x10-9

0,856

72,762

3,531x10-6

0,895

103,818

6,128x10-7

0,838

62,993

7,043x10-6

RM0=-0,378pklogP+1,27

0,87

81,443

2,040x10-6

Ch/EtOAc

RM0=-0,832ACDlogP+2,471

0,831

60,143

8,766x10-6

Ch/Et2O

RM0=-0,4ACDlogP-0,08

0,872

82,754

1,887x10-6

RM0=-0,338ClogP+1,406

Ch/EtOAc

RM0=-0,805milogP+3,232

Ch/Et2O

RM0=-0,383milogP+1,557

Ch/EtOAc

RM0=-0,856SKlogP+3,032

Ch/Et2O

RM0=-0,411SKlogP+1,472

Ch/EtOAc

RM0=-0,789pKlogP+2,617

Ch/Et2O

PT

CC E A

N

Ch/Et2O

A

RM0=-0,7111ClogP+2,914

M

Ch/EtOAc

ED

Equation

U

r2adj.

Modifier

Table 4. Linear regression models of b (slope) as function of different logP values.

Equation

r2adj.

F

P

Ch/EtOAc

b=-2,582ClogP+0,47

0,8417

64,836

6,141x10-6

Ch/Et2O

b=-0,282ClogP+0,418

0,91

122,931

2,611x10-7

Ch/EtOAc

b=0,539milogP-2,828

0,904

114,099

3,81x10-7

Ch/Et2O

b=-0,311milogP+0,516

0,893

101,691

6,798x10-7

Ch/EtOAc

b=0,571SklogP-2,675

0,82

55,659

1,26x10-5

Ch/Et2O

b=-0,348SklogP+0,49

Ch/EtOAc

b=0,529pklogP-2,404

Ch/Et2O

b=-0,318pklogP+0,314

Ch/EtOAc

b=0,558ACDlogP-2,31

Ch/Et2O

b=-0,335CDlogP+0,253

SC R 0,918

135,307

1,6x10-7

0,81

51,815

1,756x10-5

0,88

89,411

1,289x10-6

0,808

51,47

1,811x10-5

0,869

80,586

2,15x10-6

U N A M

ED PT CC E A

IP T

Modifier

Table 5. Established MLR models for prediction of retention parameters of vanilidene derivatives.

RM 0

Equation

R2

R2adj.

F

p

RM0(CHCl3/EtOAc)

-0,711ClogP-0,001TPSA-5,846x10-4 Mw+3,265

0,91

0,88

30,478

4,806x10-5

RM0(CHCl3/EtOAc)

-0,319ClogP-2,587x10-4TPSA0,001Mw+1,735 -0,814milogP+0,004TPSA+9,432x10-4 Mw+2,62

0,947

0,93

54,611

0,952

0,937

60,426

0,96

IP T

RM0(ChCl3/Et2O)

4,247x10-6

-0,36milogP+0,002TPSA-4,078x10-4 Mw+1,438

0,97

RM0(CHCl3/EtOAc)

-0,959SKlogP-0,006TPSA+0,001Mw+3,2

0,88

RM0(ChCl3/Et2O)

-0,426SKlogP-0,002TPSA-7,313x105 Mw+1,697

0,91

RM0(CHCl3/EtOAc)

-1,101pklogP-0,019TPSA-0,004Mw+3,438

0,925

RM0(CHCl3/EtOAc)

-1,048ACDilogP0,0102TPSA+0,003Mw+2,457

M

ED

-0,458ACDilogP-0,003TPSA+7,419x10-4 Mw+1,361

A

CC E

PT

RM0(ChCl3/Et2O)

U

N

-0,48pklogP-0,007TPSA+9,392x104 Mw+1,789

A

RM0(ChCl3/Et2O)

97,595

SC R

RM0(ChCl3/Et2O)

2,76x10-6

3,49x10-7

22,108

1,73x10-4

0,88

30,883

4,555x10-5

0,901

37,413

2,074x10-5

0,84

1,226x10-5

0,934

0,912

42,457

0,882

0,843

22,49

1,618x10-4 8,594x10-5

0,897

0,863

26,385

Table 6. Best MLR models for prediction of adequate MIC values.

MIC

Equation

R2

R2adj.

F

p

Candida

0,595milogP+0,015TPSA0,008Mw+0,104

0,76

0,68

0,498

3,77x10-3

0,669milogP+0,212TPSA-0,008Mw0,373

0,749

0,665

8,941

4,6x10-3

0,671

0,562

6,14

0,0147

0,793

0,725

11,525

1,95x10-3

MLR1

albicans

MLR2

Trichoderma

MLR3

viride

-1,685RM0(CHCl3/Et2O)

+0,021TPSA-

-1,286RM0(CHCl3/EtOAc)+0,029TPSA0,006Mw+1,5

E. coli

MLR5

-2,323RM0(CHCl3/Et2O) 0,007Mw-2,671 MLR6 MLR7

Fuscarium oxysporum

+0,017TPSA-

-0,688RM0(CHCl3/EtOAc)+0,015TPSA0,006Mw+1,989 -1,24RM0(CHCl3/Et2O)

MLR10 MLR11

Penicillium italicum

ED

-0,812RM0(CHCl3/Et2O) 0,005Mw+1,965

PT

Aspergillus flavus

CC E

MLR12

A

MLR13

0,625

0,5

5,002

0,026

0,62

0,5

4,884

0,027

0,683

0,58

6,489

0,0125

0,771

0,695

10,122

3,05x10-3

0,7

0,6

7,01

9,93x10-3

0,705

0,607

7,189

9,19x10-3

0,78

0,706

10,636

2,57x10-3

0,74

0,65

8,481

5,46x10-3

-0,015TPSA-

-1,503RM0(CHCl3/EtOAc)+0,012TPSA0,005Mw+3,852 -2,35RM0(CHCl3/Et2O) 0,005Mw+5,261

2,27x10-3

+0,005TPSA-

-0,097RM0(CHCl3/EtOAc)-0,017TPSA0,006Mw+4,323 -0,315RM0(CHCl3/Et2O) 0,007Mw+4,397

11,037

N

A

-0,34RM0(CHCl3/EtOAc)+0,004TPSA0,004Mw+1,676

M

MLR9

Pseudomonas aeruginosa

0,715

+0,008TPSA-

0,007Mw-2,671 MLR8

0,786

U

MLR4

SC R

0,008Mw-2,297

IP T

MLR

-0,008TPSA-

M

A

N

U

SC R

IP T

[23] T. Roemer, D.J. Krysan, Antifungal drug development: challenges, unmet clinical needs, and new approaches, Cold Spring Harb Perspect Med. 1 (2014) 4(5).

A

CC E

PT

ED

Figure 1. Structures of investigated vanilidene derivatives of Meldrum’s acid.

Figure 2. Linear relation of RM0(Ch/EtOAc and Ch/Et2O) and concentration (log C) of modifier (EtOAc and Et2O).

IP T SC R

PT

ED

M

A

N

U

Figure 3. The results of SRD analysis of the scaled ADME values of the studied compounds

A

CC E

Figure 4. Score values of the investigated compounds for the first two principal components

IP T

PT

ED

M

A

N

U

SC R

Figure 5. Dendogram of 13 examined vanilidene derivatives in the space of seven selected molecular descriptors.

A

CC E

Figure 6. SRD-CRRN results for obtained MLR models.