{In silico studies on smoothened human receptor and its antagonists in search of anticancer effects} J. Serb. Chem. Soc. 85 (3) 335–346 (2020) UDC 612.398:577.354:542.97+519.8:615.277 JSCS–5304 Original scientific paper 335 In silico studies on smoothened human receptor and its antagonists in search of anticancer effects ANA BOROTA*, SORIN AVRAM, RAMONA CURPAN, ALINA BORA, DANIELA VARGA, LILIANA HALIP and LUMINITA CRISAN** “Coriolan Dragulescu” Institute of Chemistry, Romanian Academy, 24 Mihai Viteazul Avenue, RO-300223, Timisoara, Romania (Received 3 April, revised 5 July, accepted 6 August 2019) Abstract: Lately, the cancers related with abnormal hedgehog (Hh) signalling pathway are targeted by smoothened (SMO) receptor inhibitors that are rapidly developing. Still, the problems of known inhibitors such as severe side effects, weak potency against solid tumors or even the acquired resistance need to be overcome by developing new suitable inhibitors. To explore the structural requirements of antagonists needed for SMO receptor inhibition, pharmaco- phore mapping, 3D-QSAR models, database screening and docking studies were performed. The best selected pharmacophore hypothesis based on which statistically significant atom-based 3D-QSAR model was developed (R2 = = 0.856, Q2 = 0.611 and Pearson-R = 0.817), was further subjected to dataset screening in order to evaluate its ability to prioritize active compounds over decoys. The efficiency of one four-points pharmacophore hypothesis (AAHR.524) was observed based on good evaluation metrics such as the area under the curve (0.795), and weighted average precision (0.835), suggesting that the model is trustworthy in predicting novel inhibitors against SMO receptor. Keywords: pharmacophore; 3D-QSAR; docking; SMO inhibitors. INTRODUCTION The smoothened (SMO) receptor is a G-protein-coupled receptor (GPCR)- -like protein and it is one of the relevant components of the hedgehog (Hh) sig- nalling pathway1. GPCR are seven transmembrane receptors which constitute the largest family of human proteins2 that regulate a various multitude of intracel- lular signalling cascades.3 Late breakthroughs in GPCRs biochemistry4,5 have enhanced our knowledge of drug action on these significant targets, which will lead to the design of better therapeutics in the future. *,** Corresponding authors. E-mail: (*)ana_borota@acad-icht.tm.edu.ro; (**)lumi_crisan@acad-icht.tm.edu.ro https://doi.org/10.2298/JSC190403085B ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ 336 BOROTA et al. It is known that ~25 % of cancers present aberrant Hh pathway activation.6 Regarding the implication of Hh pathway in cancer, three models of action have been proposed.7 The first type is ligand independent and refers to mutations in Hh pathway components, that have been associated with medulloblastoma and basal cell carcinomas.8,9 The II (autocrine) and III (paracrine) types of mech- anisms are ligand dependent and include: lung,10 pancreatic,11 colorectal,12 breast,13 gastrointestinal tract,14 prostate,15 glioma tumours9 and hematological malignancies.16 The growth of these types of tumours can be suppressed by dif- ferent pathway inhibitors, such as SMO receptor antagonists.17 Indeed, a number of SMO antagonists (such as: CUR61414, IPI-926, BMS-833923, PF-04449913, LEQ-506 and TAK-441) have demonstrated anticancer activity and have entered the clinical trials. Hitherto, FDA approved two SMO antagonists (LDE225/ /Sonidegib and GDC-0449/Vismodegib) for the treatment of basal cell carci- noma.18 Unfortunately, these two inhibitors have shown severe side effects, along with weak potency against solid tumours, and, moreover, an issue related to acquired resistance has been identified.1,19 To overcome these downsides, many studies have focused on designing new and diverse SMO inhibitors, with desirable selectivity against the target of interest.20 In this regard, a recently syn- thetized and biologically evaluated series of 26 compounds21 as SMO anta- gonists (Table I) was used for computational investigation. Ligand- and structure-based approaches were involved in order to under- stand which are the essential features responsible for the inhibition of SMO rec- eptor. Phase software (Schrödinger)22 was used for the generation of ligand- based pharmacophore models and 3D-QSAR module was employed to validate the models. An external validation of the best pharmacophore hypothesis was achieved based on virtual screening (VS) and statistical evaluation of the results. The importance of stereochemical aspects involved in ligand-receptor interact- ions were highlighted based on the rigid and flexible docking studies performed with Glide23 and Induced Fit24 programs. COMPUTATIONAL METHODS Pharmacophore generation protocol In the current study, a dataset of 26 compounds newly synthesized and biologically tested against SMO receptor, by Lu and collaborators21 (Table S-I of the Supplementary material to this paper) was the subject of computational analysis for pharmacophore gener- ation and docking studies. The preparation of the ligands for in silico studies along with the steps followed for the generation of pharmacophore hypotheses are presented in detailed man- ner in the Supplementary material. Phase22 with the option: “Develop Common Pharmacophore Hypotheses” was used for the generation and validation of the pharmacophore hypotheses by the involvement of the atom-based QSAR module. The following features: hydrogen bond donor/acceptor, hydro- phobic, negative, positive and aromatic rings were taken in account for the generation of the pharmacophore hypotheses. In order to select the best hypothesis, an atom-based 3D-QSAR ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ COMPUTATIONAL STUDIES ON SMO INHIBITORS 337 analysis was carried out using three partial least-squares (PLS) factors. The test set com- pounds that represent roughly 23 % of the dataset were selected to cover the same range of activity as the compounds from the training set.25 The multidimensional descriptor space between the training and test set was checked26 using 2D fingerprints. The well-known 2D Tanimoto and Euclidean distance coefficients, and the median values for important 2D pro- perties of molecules were calculated (Supplementary material). The best selected pharmacophore hypothesis was employed in VS experiments using the “Advanced Pharmacophore Screening” option of Phase software (the inter-site distance matching tolerance was of 2Å for minimum four site points) to test its ability to distinguish active compounds or positives from inactive ones or negatives (assumed to be the decoys). For this purpose, an external validation dataset was assembled, and it was realized following the protocol described in the Supplementary material. The VS results were evaluated using an in-house developed program, “Evaluation” tool In ChemInformatics (ETICI 1.6).27 The following statistical metrics (Eqs. (1)–(5)) were used for the evaluation: receiver operating characteristics (ROC) curve and its corresponding area under the curve (AUC)28 Boltzmann enhanced discrimination of ROC (BEDROC)29, precision (PPV)30, accuracy (Acc), sensitivity (Se), specificity (Sp)31 and the true positive (TP) at 1, 2, 5, 10 % of false positives (FPs):32 + = 1 1 = 1 + TP FN i i AUC FPR TP FN −  (1) TP PPV TP FP = + (2) TP Se TP FN = + (3) TN Sp TN FN = + (4) TP TN Acc TP TN FP FN + = + + + (5) where: TP indicates the correctly predicted actives; FN designates the incorrectly identified inactives; FPR is false positive rate; TN represents the correctly predicted inactives or nega- tives and FP denotes the incorrectly predicted actives or positives. AUC parameter is used to evaluate the models performance in classification, to dis- criminate actives from inactives and it takes values in the range 0 to 1. The closer AUC value is to 1 the better a model discerns true positive from false positive results.28 BEDROC can be considered as a “virtual screening usefulness scale” while ROC assesses “improvement over random scale”. The difference between the two aforementioned evaluation parameters is that ROC relates to a uniform distribution and BEDROC refers to an exponential distribution.29 For imbalance class when the number of actives is lower than a number of inactives, the pre- cision parameter was replaced with weighted average precision.33 Docking simulations protocol The best resolution X-ray structure of SMO receptor was extracted from PDB (PDB ID: 4jkv) and prepared for docking studies with the “Protein Preparation Wizard” module from Schrödinger suite. Briefly, the following steps were employed: the bond orders were assigned; the Chemical Components Dictionary (CCD) database was used; hydrogens were added; ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ 338 BOROTA et al. create zero-order bonds to metals parameter was selected; disulfide bonds were created; waters beyond 5 Å were deleted and missing side-chains were added to the protein. The pro- per ionization state of the co-crystallized ligand (LY2940680), at physiological pH of 7.0–7.4 was generated and the H-bond assignment was performed using PROPKA module, setting pH to 7.0. Finally, the protein was energetically minimized with OPLS_2005 force field until an RMSD of 0.3 Å for heavy atoms was reached. A molecular docking study was carried out with “Glide”23 and “Induced Fit”24 protocol included in Schrödinger suite. “Glide” module23 was used in standard precision (SP) mode with default settings, i.e., scaling factor of 0.8, partial charge cut-off of 0.15 and only the flex- ibility of ligands was considered. “Induced Fit” docking protocol24 accounts also for protein flexibility and it was carried out with the following settings: SP option, receptor and ligand van der Waals scaling of 0.50. RESULTS AND DISCUSSION Analysing the results generated by Phase,22 22 categories of four features common pharmacophore hypotheses (CPH) were identified. CPH include the same features placed together in similar spatial arrangements. All hypotheses received a good survival score (values > 3) and were subjected to QSAR model development. A QSAR model is reliable if a number of statistical parameters meet certain condition as follows: R2 (squared correlation coefficients for the training set) > 0.6; Q2 (squared correlation coefficients for the test set) > 0.5;26 Pearson-R (correlation between the predicted and observed activity for the test set) > 0.5; p (the significance level of variance ratio) < 0.05; SD (standard devi- ation) and RMSE (root mean square error) should have low values, while F (vari- ance ratio) should have high value.22,34 The best hypothesis named AAHR.524 with a survival score of 5.263 has statistical significant results (Table I). TABLE I. The statistical parameters for the AAHR.524 hypothesis obtained PLS factor SD R2 p F RMSE Q2 Pearson-R 1 0.255 0.601 6.00E-05 27.100 0.308 0.470 0.760 2 0.195 0.779 2.72E-06 29.900 0.303 0.486 0.769 3 0.161 0.859 4.90E-07 32.500 0.263 0.611 0.817 For AAHR.524 hypothesis the distances between pharmacophore sites are shown in Fig. 1 and scatter plots of experimental versus predicted bioactivity values for training and test set are presented in Fig. 2. In order to highlight favorable and unfavorable regions for ligand–protein interactions, the atom-based 3D-QSAR model regarding hydrogen bond donor, hydrophobic and electron withdrawing properties was mapped against the phar- macophore model (Fig. 3). Additionally, the AAHR.524 hypothesis was vali- dated in virtual screening (VS) experiments using Phase Advanced Pharmaco- phore Screening (Fig. 4). In Fig. 3 the combined effects for all the features retri- eved in the workspace ligands along with the overlapping of the best fitted and the less active ligands are depicted. The favorable regions for biological activity ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ COMPUTATIONAL STUDIES ON SMO INHIBITORS 339 are represented by blue cubes, while the red cubes indicate detrimental regions. For the AAHR.524 hypothesis, the presence of blue cubes in the region of the –NH– group of compound 17 (Fig. 3b) indicates that hydrogen bond donor pro- perty of this group is favorable for activity. This finding is also supported by Glide docking results which highlight the contribution of this amino group in the formation of a hydrogen bond with residue Tyr 394 of the SMO binding site. Fig. 1. AAHR.524 pharmacophore hypothesis depicted with the best fitted compound 17. The distances between the following features are presented: hydrogen bond acceptor (A1), hydrogen bond acceptor (A3), aromatic ring (R12) and one hydrophobic region (H8). Fig. 2. Correlation plots of the experimental versus Phase predicted activity of training set (circles) and test set (triangles) for AAHR.524 hypothesis. The pyridine ring of compound 17 is engaged in hydrophobic interactions with the binding site, while in the less active ligand (compound 22) this feature is missing (Fig. 3c). This finding was also confirmed by docking studies. Regarding the electron withdrawing property (Fig. 3a), the Glide docking study revealed that the oxygen atom of the carbonyl group of compound 17 can act as hydrogen bond acceptor for Arg400 residue. In order to assess the efficiency of VS experiments, it is necessary to test their ability to distinguish active compounds or positives from inactive ones or negatives (assumed to be the decoys). In this regard, different evaluation oper- ators were employed. Namely, the receiver operation characteristics (ROC) curve28 (Fig. 4) was used for the visualization of the true positive rate (TPR) versus the false positive rate (FPR) and the area under the curve (AUC) was cal- culated. In our experiment, a good scores for: AUC of 0.795 (±0.027), Se (the ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ 340 BOROTA et al. fraction of actives correctly predicted) of 0.587, Sp (the fraction of inactives correctly predicted) of 0.647, Acc (the fraction of correct predictions) of 0.640, WPPV (weighted average precision) of 0.835 and BEDROC (α = 20) of 0.766, were obtained, suggesting that the AAHR.524 pharmacophore hypothesis pro- vides a good class separation. Additionally, good enrichment factors, TP at x of FPs were obtained: 19.047 % at x = 1 %; 30.476 % at x = 2 %; 42.857 % at x = 5 %; 56.190% at x = 10 %. Fig. 3. Superposition of compound 17 (best fitted over AAHR.524 hypothesis) represented by grey carbon atoms with compound 22 (least active ligand) depicted by green carbon atoms: a) electron withdrawing property; b) hydrogen bond donor property; c) hydrophobic property. Fig. 4. The receiver operation characteristics (ROC) curve, involving the AAHR.524 hypothesis against the validation data set. In Fig. 4 is depicted the ROC curve, the diagonal represents the random sampling of compounds (actives and decoys) and it separates two classification areas: the correctly predicted compounds are shown above, while the incorrectly ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ COMPUTATIONAL STUDIES ON SMO INHIBITORS 341 classified compounds occupy the underneath area. TPR represents the fraction of correctly classified positives (actives) from the total of positives, while FPR expresses the fraction of the incorrectly classified negatives (decoys) from the total negatives.28,35 Following ligand-based analysis, the structure-based study, represented by molecular docking, bring additional light on ligand-protein inter- action profile. In order to validate the reasonability of choosing the two docking software for our study (Glide (Fig. 5a) and Induced Fit (Fig. 5b)), we have undertaken a control experiment. The co-crystallized inhibitor LY2940680 was docked to its native X-ray structure, 4jkv, and the best docked pose was superimposed with the native binding pose from the experimental structure 4jkv. a b Fig. 5. Superposition of the pose of the crystal structure 4jkv (with the inhibitor LY2940680), and the pose of the same ligand obtained by docking with Glide SP (a) superposition of the pose of the crystal structure 4jkv (with the inhibitor LY2940680), and the pose of the same ligand obtained by docking with Induced Fit SP (b). The RMSD values were also calculated. According to Gohlke et al.,35 an RMSD value less than 2.0 Å is broadly accepted as “effectively” docked structure. The docking protocol with Glide and Induced Fit have provided very good RMSD values of 0.192 (Fig. 5a), and 0.700, respectively (Fig. 5b). A comparative study made by Wang et al.36 on the binding profile and inter- actions of the ligands: antagonists and agonists at the human SMO receptor, rev- eals some specific features. The binding sites of antagonists, e.g., LY2940680, cyclopamine and AntaXV and the agonist SAG1.5 consist mostly of residues from the extracellular linker domain and loops. In particular, Asp219 (of the extracellular linker domain) it involved in the formation of hydrogen bonds with ligands: LY2940680, AntaXV and SAG1.5.36 Also, an important residue from the helix V is Arg400 (Arg5.39) which forms hydrogen bonds with antagonists: LY2940680 and AntaXV.36,37 According to the latest studies, the SMO receptor ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ 342 BOROTA et al. contains two distinct binding sites. One ligand-binding site is located in hepta- helical transmembrane domain, and the second, occupied by a cholesterol mole- cule, is placed in the extracellular cysteine-rich domain.38 Thus, the SMO inhi- bitors were classified in two categories: 1) antagonists that bind mainly to the extracellular loops and 2) antagonists that enter deeper into the cavity formed by the transmembrane helices domain.20 The large binding site of SMO receptor allows flexibility to the ligands lead- ing to several distinct binding modes.20,36–38 Although, similar disposition of the ligand in situ were retrieved by the two docking methods employed (Fig. 6c), the orientation of the ligand is reversed in the binding site, hence resulting in differ- ent interactions with the amino acids of the protein (Fig. 6a and b). Two inter- action patterns for compound 17 were observed depending on the docking soft- ware used, Glide (Fig. 6a) or Induced Fit (Fig. 6b). Fig. 6. Representation of the interactions map of the best docked pose of compound 17 with the amino acids of SMO human binding site: glide docking (a); induced fit docking (b); their superposition in situs (c). Following the most important features characterizing the interaction of com- pound 17 with the binding site of SMO receptor resulted from docking with Glide are presented: i) the oxygen atom of the carbonyl group acts like hydrogen bond acceptor for Arg400 and Gln 477; ii) the aliphatic –NH– group is hydrogen bond donor to Tyr394; iii) the dimethyl-pyridine ring is involved in π–π stacking ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ COMPUTATIONAL STUDIES ON SMO INHIBITORS 343 interactions with Trp281 and Phe391; iv) the chlorine atom is involved in alkyl interactions with Ile234, Leu522 and Trp281 (Fig. 6a). The following main interactions were found when the compound was docked with Induced Fit protocol: i) the N= atom of dimethylpyridine is acting as hydrogen bond acceptor for Asn219; ii) the –NH= group of chloropyridine is functioning as hydrogen bond donor to Glu518; and iii) the pyridine ring is involved in π–π stacking interactions with Trp281 and His470 residues (Fig. 6b). Also, the ligand rings are involved in multiple hydrophobic interactions with residues of the binding site, e.g., Pro513, Trp281, Val386, Leu325, Met525, His470 and Ser387 (Fig. 6a and b), leading to the stabilization of the ligand into the binding pocket. The amino acids Arg400 and Tyr394 which interact with compound 17, are also known to be important for the interaction with other SMO antagonists. The relevance of these two amino acids in the antagonistic binding is supported by crystallographic studies.36–38 It can be seen from the Fig. 6b that three pharmacophore features of the four of AAHR.524 hypothesis are recapitulated in the best docking pose resulted from Induced Fit, only the interaction with the acceptor (A3) is missing. CONCLUSIONS In order to discover the common features of inhibitors of SMO receptor with potential anticancer benefits, ligand-based pharmacophore models were first gen- erated. The best pharmacophore hypothesis turned out to be AAHR.524, with the following characteristics: two hydrogen bond acceptors (A), one hydrophobic (H) site and one aromatic ring (R). The pharmacophore hypothesis was found to be statistically significant with the correlation coefficient, R2 of 0.859 for the train- ing set and the correlation coefficient, Q2 of 0.611 for the test set. Virtual screen- ing experiments and molecular docking were involved in the evaluation of the models and to gain insights into the ligand-receptor interaction profile. Consider- ing the large size of the binding site, as well as the different binding patterns of inhibitors to the site, our pharmacophore model being built on a congeneric series of inhibitors it is assumed to lead to increased specificity and reduced diversity in virtual screening experiments. However, we can say that the pharmacophore model was validated through a virtual screening experiment with reliable values of the evaluation metrics, e.g., AUC of 0.795, and enrichment factors. In addition, it can be seen that our docking results reproduce the main interactions with key residues (e.g., Asn219, Arg400, Tyr394 and Asp384) found in the X-ray crystal structure (PDB ID: 4jkv). Between the two docking approaches used: Glide and Induced Fit, the last one which accounts also for protein flexibility has led to better results. The validated pharmacophore model obtained suggests that it could ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ 344 BOROTA et al. be reliably used in predicting novel, promising inhibitors against SMO receptor with potential anticancer activity. SUPPLEMENTARY MATERIAL Additional data are available electronically from http://www.shd.org.rs/JSCS/, or from the corresponding author on request. Acknowledgements. This project was financially supported by Project 1.1 of the “Cori- olan Dragulescu” Institute of Chemistry, Romanian Academy. The authors thank to the Chemaxon Ltd. and OpenEye Ltd. for providing an academic license. И З В О Д IN SILICO СТУДИЈЕ НА ЗАГЛАЂЕНОМ (SMOOTHENED) ХУМАНОМ РЕЦЕПТОРУ И ЊЕГОВИМ АНТАГОНИСТИМА У ПОТРАЗИ ЗА АНТИТУМОРСКИМ ЕФЕКТИМА ANA BOROTA, SORIN AVRAM, RAMONA CURPAN, ALINA BORA, DANIELA VARGA, LILIANA HALIP и LUMINITA CRISAN� “Coriolan Dragulescu” Institute of Chemistry, Romanian Academy, 24 Mihai Viteazul Avenue, RO-300223, Timisoara, Romania У новије време, канцери повезани са ненормалним хеџхог (Hh) сигналним путем (hedgehog (Hh) pathway signalling) се испитују преко инхибиторима заглађених рецептора (SMO, Smoothened (SMO) receptor inhibitors) који се брзо развијају. Међутим, проблеми са познатим инибиторима попут озбиљних споредних ефеката, слабе моћи против солид- них тумора или чак стечене отпорности тек треба да се превазиђу развојем нових погод- них инхибитора. Да би истражили структурне захтеве за антагонисте потребне за инхи- бицију SMO рецептора урадили смо: мапирање фармакофоре, 3D-QSAR моделе, претра- живање база података и студије докинга. Најбоља хипотеза за фармакофору је одабрана, на основу ње је развијен статистички значајан атомски 3D-QSAR модел (R 2 = 0,856, Q 2 = = 0,611 и Pearson-R = 0,817), који је даље подвргнут скринингу, база података, како бу се проценила његова способност да издвоји активна једињења од лажно позитивних. Запа- жена је ефикасност четверопараметарске (four-points) хипотезе фармакофоре (AAHR.524) на бази добре метрике процена попут површине испод криве (0,795) и прецизности одмереног просека (0,835) што сугерише да је модел поуздан за предвиђање нових инхи- битора за SMO рецептор. (Примљено 3. априла, ревидирано 5. јула, прихваћено 6. августа 2019) REFERENCES 1. M. Ruat, L. Hoch, in The Smoothened Receptor in Cancer and Regenerative Medicine, M. Ruat, Ed., Springer, Berlin, 2015, pp. 1–11 (ISBN 978-3-319-19755-5) 2. G. Liapakis, A. Cordomí, L. Pardo, Curr. Pharm. Des. 18 (2012) 175 (http://dx.doi.org/10.2174/138161212799040529) 3. D. Hilger, M. Masureel, B. K. Kobilka, Nat. Struct. Mol. Biol. 25 (2018) 4 (http://dx.doi.org/10.1038/s41594-017-0011-7) 4. J. Shonberg, R. C. Kling, P. Gmeiner, S. Löber, Bioorg. Med. Chem. 23 (2015) 3880 (http://dx.doi.org/10.1016/j.bmc.2014.12.034) 5. L. Halip, A. Borota, M. Mracec, R. Curpan, A. Gruia, M. Mracec, Rev. Roum. Chim. 54 (2009) 157 (http://revroum.lew.ro/wp- content/uploads/2009/RRCh_2_2009/Art%2007.pdf) 6. K. K. Chahal, M. Parle, R. Abagyn, Anti-Cancer Drugs 29 (2018) 387 (http://dx.doi.org/10.1097/CAD.0000000000000609) ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ COMPUTATIONAL STUDIES ON SMO INHIBITORS 345 7. L. L. Rubin, F. J. de Sauvage, Nat. Rev. Drug Discov. 5 (2006) 1026 (http://dx.doi.org/10.1038/nrd2086) 8. C. Wicking, I. Smyth, A. Bale, Oncogene 18 (1999) 7844 (http://dx.doi.org/10.1038/sj.onc.1203282) 9. S. J. Scales, F. J. de Sauvage, Trends Pharmacol. Sci. 30 (2009) 303 (http://dx.doi.org/10.1016/j.tips.2009.03.007) 10. D. N. Watkins, D. M. Berman, S. G. Burkholder, B. Wang, P. A. Beachy, S. B. Baylin, Nature 422 (2003) 313 (http://dx.doi.org/10.1038/nature01493) 11. S. P. Thayer, M. P. di Magliano, P. W. Heiser, C. M. Nielsen, D. J. Roberts, G. Y. Lauwers, Y. P. Qi, S. Gysin, C. Fernández-del Castillo, V. Yajnik, B. Antoniu, M. McMahon, A. L. Warshaw, M. Hebrok, Nature 425 (2003) 851 (http://dx.doi.org/10.1038/nature02009) 12. D. Qualtrough, A. Buda, W. Gaffield, A. C. Williams, C. Paraskeva, Int. J. Cancer 110 (2004) 831 (http://dx.doi.org/10.1002/ijc.20227) 13. S. Mukherjee, N. Frolova, A. Sadlonova, Z. Novak, A. Steg, G. P. Page, D. R. Welch, S. M. Lobo-Ruppert, J. M. Ruppert, M. R. Johnson, A. R. Frost, Cancer Biol. Ther. 5 (2006) 674 (https://doi.org/10.4161/cbt.5.6.2906) 14. D. M. Berman, S. S. Karhadkar, A. Maitra, R. Montes de Oca, M. R. Gerstenblith, K. Briggs, A. R. Parker, Y. Shimada, J. R. Eshleman, D. N. Watkins, P. A. Beachy, Nature 425 (2003) 846 (http://dx.doi.org/10.1038/nature01972) 15. S. S. Karhadkar, G. S. Bova, N. Abdallah, S. Dhara, D. Gardner, A. Maitra, J. T. Isaacs, D. M. Berman, P. A. Beachy, Nature 431 (2004) 707 (http://dx.doi.org/10.1038/nature02962) 16. P. Lin, Y. He, G. Chen, H. Ma, J. Zheng, Z. Zhang, B. Cao, H. Zhang, X. Zhang, X. Mao, Anti-Cancer Drugs 29 (2018) 995 (http://dx.doi.org/10.1097/CAD.0000000000000679) 17. J. Jiang, C. C. Hui, Dev. Cell. 15 (2008) 801 (http://dx.doi.org/10.1016/j.devcel.2008.11.010) 18. T. K. Rimkus, R. L. Carpenter, S. Qasem, M. Chan, H. W. Lo, Cancers 8 (2016) 22 (http://dx.doi.org/10.3390/cancers8020022 ) 19. S. V. Mohan, A. L. Chang, Clin. Cancer Res. 21 (2015) 2677 (http://dx.doi.org/10.1158/1078-0432.CCR-14-3180) 20. M. Xin, X. Ji, L. K. de la Cruz, S. Thareja, B. Wang, Med. Res. Rev. 38 (2018) 870 (http://dx.doi.org/10.1002/med.21482) 21. W. Lu, D. Geng, Z. Sun, Z. Yang, H. Ma, J. Zheng, X. Zhang, Bioorg. Med. Chem. Lett. 24 (2014) 2300 (https://doi.org/10.1016/j.bmcl.2014.03.079) 22. S. L. Dixon, A. M. Smondyrev, E. H. Knoll, S. N. Rao, D. E. Shaw, R. A. Friesner, J. Comput.-Aided Mol. Design 20 (2006) 647 (http://dx.doi.org/10.1007/s10822-006-9087-6) 23. R. A. Friesner, J. L. Banks, R. B. Murphy, T. A. Halgren, J. J. Klicic, D. T. Mainz, M. P. Repasky, E. H. Knoll, D. E. Shaw, M. Shelley, J. K. Perry, P. Francis, P. S. Shenkin, J. Med. Chem. 47 (2004) 1739 (https://doi.org/10.1021/jm0306430) 24. W. Sherman, T. Day, M. P. Jacobson, R. A. Friesner, R. Farid, J. Med. Chem. 49 (2006) 534 (https://doi.org/10.1021/jm050540c) 25. A. Golbraikh, A. Tropsha, J. Mol. Graph. Model. 20 (2002) 269 (https://doi.org/10.1016/S1093-3263(01)00123-1) 26. A. Golbraikh, A. Tropsha, J. Comput.-Aided Mol. Des. 16 (2002) 357 (https://doi.org/10.1023/A:1020869118689) 27. S. I. Avram, L. M. Pacureanu, A. Bora, L. Crisan, S. Avram, L. Kurunczi, J. Chem. Inform. Model. 54 (2014) 2360 (http://dx.doi.org/10.1021/ci5002668) ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ 346 BOROTA et al. 28. J. A. Hanley, B. J. McNeil, Radiology 143 (1982) 29 (http://dx.doi.org/10.1148/radiology.143.1.7063747) 29. J. F. Truchon, C. I. J. Bayly, J. Chem. Inform. Model. 47 (2007) 488 (http://dx.doi.org/10.1021/ci600426e) 30. M. Hall, E. Frank, G. Holmes, B. Pfahringer, P. Reutemann, I. H. Witten, SIGKDD Explor. Newsletter 11 (2009) 10 (http://dx.doi.org/10.1145/1656274.1656278) 31. T. Fawcett, Pattern Recog. Lett. 27 (2006) 861 (http://dx.doi.org/10.1016/j.patrec.2005.10.010) 32. A. N. Jain, J. Comput.-Aided Mol. Des. 22 (2008) 201 (http://dx.doi.org/10.1007/s10822- 007-9151-x) 33. E. Kotsampasakou, G. F. Ecker, J. Chem. Inform. Model. 57 (2017) 608 (http://dx.doi.org/10.1021/acs.jcim.6b00518) 34. B. Badhani, R. Kakkar, J. Biomol. Struct. Dyn. 35 (2017)1950 (http://dx.doi.org/10.1080/07391102.2016.1202863) 35. H. Gohlke, M. Hendlich, G. Klebe, J. Mol. Biol. 295 (2000) 337 (http://dx.doi.org/10.1006/jmbi.1999.3371) 36. C. Wang, H. Wu, T. Evron, E. Vardy, G. W. Han, X. P. Huang, S. J. Hufeisen, T. J. Mangano, D. J. Urban, V. Katritch, V. Cherezov, M. G. Caron, B. L. Roth, R. C. Stevens, Nature Commun. 10 (2014) 4355 (http://dx.doi.org/10.1038/ncomms5355) 37. C. Wang, H. Wu, V. Katritch, G. W. Han, X. P. Huang, W. Liu, F. Y. Siu, B. L. Roth, V. Cherezov, R. C. Stevens, Nature 497 (2013) 338 http://dx.doi.org/10.1038/nature12167) 38. E. F. X. Byrne, R. Sircar, P. S. Miller, G. Hedger, G. Luchetti, S. Nachtergaele, M. D. Tully, L. Mydock-McGrane, D. F. Covey, R. P. Rambo, M. S. P. Sansom, S. Newstead, R. Rohatgi, C. Siebold, Nature 535 (2016) 517 (http://dx.doi.org/10.1038/nature18934). ________________________________________________________________________________________________________________________ (CC) 2020 SCS. Available on line at www.shd.org.rs/JSCS/ << /ASCII85EncodePages false /AllowTransparency false /AutoPositionEPSFiles true /AutoRotatePages /None /Binding /Left /CalGrayProfile (Dot Gain 20%) /CalRGBProfile (sRGB IEC61966-2.1) /CalCMYKProfile (U.S. Web Coated \050SWOP\051 v2) /sRGBProfile (sRGB IEC61966-2.1) /CannotEmbedFontPolicy /Error /CompatibilityLevel 1.4 /CompressObjects /Tags /CompressPages true /ConvertImagesToIndexed true /PassThroughJPEGImages true /CreateJobTicket false /DefaultRenderingIntent /Default /DetectBlends true /DetectCurves 0.0000 /ColorConversionStrategy /CMYK /DoThumbnails false /EmbedAllFonts true /EmbedOpenType false /ParseICCProfilesInComments true /EmbedJobOptions true /DSCReportingLevel 0 /EmitDSCWarnings false /EndPage -1 /ImageMemory 1048576 /LockDistillerParams false /MaxSubsetPct 100 /Optimize true /OPM 1 /ParseDSCComments true /ParseDSCCommentsForDocInfo true /PreserveCopyPage true /PreserveDICMYKValues true /PreserveEPSInfo true /PreserveFlatness true /PreserveHalftoneInfo false /PreserveOPIComments true /PreserveOverprintSettings true /StartPage 1 /SubsetFonts true /TransferFunctionInfo /Apply /UCRandBGInfo /Preserve /UsePrologue false /ColorSettingsFile () /AlwaysEmbed [ true ] /NeverEmbed [ true ] /AntiAliasColorImages false /CropColorImages true /ColorImageMinResolution 300 /ColorImageMinResolutionPolicy /OK /DownsampleColorImages true /ColorImageDownsampleType /Bicubic /ColorImageResolution 300 /ColorImageDepth -1 /ColorImageMinDownsampleDepth 1 /ColorImageDownsampleThreshold 1.50000 /EncodeColorImages true /ColorImageFilter /DCTEncode /AutoFilterColorImages true /ColorImageAutoFilterStrategy /JPEG /ColorACSImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /ColorImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /JPEG2000ColorACSImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /JPEG2000ColorImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 300 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /GrayImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /JPEG2000GrayACSImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /JPEG2000GrayImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict << /K -1 >> /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile () /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False /CreateJDFFile false /Description << /ARA /BGR /CHS /CHT /CZE /DAN /DEU /ESP /ETI /FRA /GRE /HEB /HRV (Za stvaranje Adobe PDF dokumenata najpogodnijih za visokokvalitetni ispis prije tiskanja koristite ove postavke. Stvoreni PDF dokumenti mogu se otvoriti Acrobat i Adobe Reader 5.0 i kasnijim verzijama.) /HUN /ITA /JPN /KOR /LTH /LVI /NLD (Gebruik deze instellingen om Adobe PDF-documenten te maken die zijn geoptimaliseerd voor prepress-afdrukken van hoge kwaliteit. De gemaakte PDF-documenten kunnen worden geopend met Acrobat en Adobe Reader 5.0 en hoger.) /NOR /POL /PTB /RUM /RUS /SKY /SLV /SUO /SVE /TUR /UKR /ENU (Use these settings to create Adobe PDF documents best suited for high-quality prepress printing. Created PDF documents can be opened with Acrobat and Adobe Reader 5.0 and later.) >> /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ << /AsReaderSpreads false /CropImagesToFrames true /ErrorControl /WarnAndContinue /FlattenerIgnoreSpreadOverrides false /IncludeGuidesGrids false /IncludeNonPrinting false /IncludeSlug false /Namespace [ (Adobe) (InDesign) (4.0) ] /OmitPlacedBitmaps false /OmitPlacedEPS false /OmitPlacedPDF false /SimulateOverprint /Legacy >> << /AddBleedMarks false /AddColorBars false /AddCropMarks false /AddPageInfo false /AddRegMarks false /ConvertColors /ConvertToCMYK /DestinationProfileName () /DestinationProfileSelector /DocumentCMYK /Downsample16BitImages true /FlattenerPreset << /PresetSelector /MediumResolution >> /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles false /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /DocumentCMYK /PreserveEditing true /UntaggedCMYKHandling /LeaveUntagged /UntaggedRGBHandling /UseDocumentProfile /UseDocumentBleed false >> ] >> setdistillerparams << /HWResolution [2400 2400] /PageSize [612.000 792.000] >> setpagedevice