NOVEL ANTICANCER AGENTS FOR TREATMENT ...

In: Cancer Research Journal Volume 4, Issue 4, pp. 1–13

ISSN: 1935-2506 © 2011 Nova Science Publishers, Inc.

NOVEL ANTICANCER AGENTS FOR TREATMENT OF BREAST CANCER UTILIZING IN SILICO OPTIMIZATION AND COMPARABILITY TO MELPHALAN Ronald Bartzatt* University of Nebraska, Durham Science Center Chemistry Department, 6001 Dodge Street Omaha, Nebraska 68182 USA Abstract Breast cancer is a type of cancer that begins usually in the inner lining of the milk ducts or lobules. Treatment includes surgery, drugs (hormone therapy and chemotherapy), and radiation. This work presents 24 drug constructs for the treatment of breast cancer that are derived through application of in silico optimized molecular properties and substituent substitution on a molecular scaffold. The parent is the bifunctional nitrogen mustard agent melphalan, which has been utilized previously in clinical treatment of metastatic breast cancer. All 24 constructs are nitrogen mustard agents having two or less violations of the Rule of 5. Various property descriptors were determined and showed a remarkable range in Log P,molar volume, aqueous solubility, and formula weight. The numerical values of Log P range from 0.45 to 10.97, formula weight from 275.05 to 676.34, molar volume from 242.2 A3 to 694.26 A3 , and aqueous solubility from 0.001 mg/Liter to 250.8 mg/Liter. These results reveal the wide range of effects that substituent substitution imputes on molecular properties that are important for pharmaceutical activity. Cluster analysis of molecular properties revealed the three most similar constructs to melphalan. Analysis of similarity (ANOSIM) indicated that generally all 24 constructs were related to melphalan. Sensitive discriminant analysis and K-means cluster analysis determined underlying relationships within the numerical properties. Application of pattern recognition methods such as these reveal efficacious variations in pharmacodynamic activity among the constructs. Overall an assemblage of 24 potentially beneficial anticancer drug agents were identified.

Introduction Breast cancer is a type of cancer that begins usually in the inner lining of the milk ducts or lobules. There are different types of breast cancer that will have different levels of proliferation, aggressiveness, and genetic makeup. The survival rate varies depending on those factors. The 10-year disease-free survival rate varies from 10% to 98%. Treatment includes surgery, drugs (hormone therapy and chemotherapy), and radiation. Worldwide, *

Email: [email protected]

2

Ronald Bartzatt

breast cancer is the second most common type of cancer after lung cancer. Previous studies showed that melphalan could be utilized separately but as second in a sequential ternary drug regimen (paciltaxel, melphalan, cyclosphos-phamide) for metastatic breast cancer treatment, and with high response outcome[1]. In addition, a ternary drug regimen (busulfan, melphalan, thiotepa) included melphalan showed significant benefit in the treatment in stage II and stage III operable tumors of the breast[2,3]. Likewise, a well tolerated and beneficial outcome for treatment of metastatic breast carcinoma was achieved by combination chemotherapy having mitomycin C, melphalan, and methotrexate[4]. Other approaches for clinical treatment of breast cancer that includes melphalan with supplemental care show considerable potential in managing proliferated disease. The single application of intensive combination alkylating agents (high dose cyclophosphamide, cisplatin, melphalan or carmustine) in addition to bone marrow support produced rapid and more frequent complete response compared to conventional chemotherapy[5]. High dose melphalan chemotherapy applied with surgery has demonstrated substantial benefits for cancer patients[6]. Melphalan administered by intravenous injection rather than orally, allows a manageable control of toxicity and greater predictability in blood levels[7]. Use of melphalan for clinical treatment of multiple myeloma and without autologous bone marrow transplant resulted in favorable outcome[8]. Adjuvant treatment of ovarian carcinoma applying melphalan after surgery resulted in a very favorable absence of residual disease[9]. Encouraging evidence resulted after high dose melphalan with noncryopreserved autologous bone marrow treatment was achieved for malignant melanoma and neuroblastoma [10]. Melphalan included in a complex multidrug regimen for treatment of relapsed Hodgkin’s disease was shown to be just as active as other combination chemotherapy regimens currently in use[11]. Studies of tumor response pertaining to melphalan application strongly support the contention that perfusion and drug distribution are major factors in variable tumor response[12]. Novel drug design by substituent substitution and in silico optimization presented here produces analogous constructs of bifunctional alkylating activity that have a broad range of property attributes that pose advantages in cell membrane permeation (indicated by Log P and polar surface area properties0 to maximize tumor response to therapy (ie. enhancement of tumor reduction and elimination). Multidrug resistance to important anticancer drugs such as melphalan in chemotherapy of breast cancer[13, 14, 15], is an alarming development that begs the further elucidation and development of chemotherapeutic agents. This work presents 24 novel drug designs having modified structural features that alter the pharmaceutical properties to present a flexible and effective treatment regimen in the face of tumor to tumor variation in drug perfusion and distribution.

Novel Anticancer Agents for Treatment of Breast Cancer Utilizing in Silico…

3

Experimental Design and Modeling Molecular Modeling and Assembly of Constructs Numerical values of some molecular properties and modeling was accomplished utilizing ACD/ChemSketch modeling v. 10.00 (Advanced Chemistry Development, 110 Yonge Street, Toronto Ontario, M5C 1T4 Canada). Other properties; polar surface area, violations of Rule of 5, molecular volume, number of oxygens/nitrogens/amines/hydroxyls, etc were determined using Molinspiration (Molinspiration Chemifor-matics, Nova ulica 61, SK-900 26 Slovensky Grob, Slovak Republic). In silico structure search for substituent replacement was accomplished using Chemical substructure and similarity search with MolCart Chemical Data Base (Molsoft L.L.C. 3366 North Torrey Pines Court, Suite 300, La Jolla, CA 92037 U S A). Visualization of 3-dimensional form where necessary was accomplished utilizing SPARTAN (Wavefunction, 18401 Von Darman Avenue, Irvine CA 92612 USA).

Pattern Recognition and Elucidation To identify underlying associations/patterns within the properties numerical matrix required the use of various pattern recognition techniques. Included in the analysis is hierarchical cluster analysis accomplished by KyPlot v. 2.0 Beta 15 (copyright Koichi Yoshioka 1997-2001). Non-metric multidimensional scaling, discriminant anaysis, ANOSIM (analysis of similarity), and non-hierarchical K-means cluster analysis were performed by PAST v. 1.80 (copyright Oyvind Hammer, D.A.T. Harper 1999-2008).

Numerical Analysis of Multivariate Data Matrix Statistical analysis of all numerical data was performed by Microsoft EXCEL (EXCEL 2003, copyright 1985-2003). Correlation analysis by Pearson r was done for some descriptors and was accomplished by GraphPad Software (GraphPad Instat v. 3.00 for Windows 95 GraphPad Software, San Diego California USA)

Results and Discussion Melphalan (Alkeran) is a bifunctional alkylating nitrogen mustard agent that is utilized either orally or intravenous for the clinical treatment of breast cancer, multiple myeloma[7], and ovarian cancer[9]. The two principal types of breast cancer are the most common form of ductal carcinoma and lobular carcinoma which begins in the lobules or milk producing tissue. Melphalan has been shown previously to be valuable in the treatment of breast carcinoma individually[5], consorted with surgery[6], and as part of a multidrug chemotherapy regimen (11). However issues have been identified associating melphalan to multidrug resistance in cancer treatment[13, 14, 15]. Novel drug constructs derived by substituent substitution and in silico optimization to melphalan are presented

4

Ronald Bartzatt

here to demonstrate the efficacy of modifying molecular properties to enhance pharmacodynamic action.

Figure 1. Novel drug constructs 2 through 18 including the parent scaffold melphalan as parent. model. Note that each novel drug construct retains the N-mustard group covalently bonded to the aromatic ring similar to melphalan. Variation in structure for constructs 2 through 18 occurs at the substituent that is para position to the N-mustard group. Constructs 2 through 18 have zero violations of the Rule of 5.


5

Figure 2. Constructs 19 through 25 likewise retain the N-mustard moiety with structural variations occurring in para position of the aromatic ring. Constructs 19 to 21 show one violation of the Rule of 5. Constructs 22 through 25 show two violations of the Rule of 5. The methodical and mensurated substitution of substituents comprising the scaffolding of melphalan produces novel analogues that have altered pharmaceutical properties which may enhance and/or expand the medicinal usefulness of this anticancer agent. Accomplishing this goal resulted in the 24 analogues presented in Figure 1 and

6

Ronald Bartzatt

Figure 2. All 24 analogues retain a minimum of one aromatic ring and the nitrogen mustard bifunctional alkylating moiety. Other structure components appearing include carboxyl groups (-COOH), hydroxyl (-OH), amine (-NHn ), amide (-CONHR), and aliphatic carbons. Numerical values of ten important molecular properties for melphalan and analogues are presented in Table 1. These structure alterations gave rise to substantial variation in Log P values (standard deviation (sd) = 2.57), as well as formula weight (sd = 110), molar volume (sd = 127 A3 ), polar surface area (sd = 20.5 A2 ), and aqueous solubility (sd = 51.1 mg/Liter). Comparatively smaller variation was also achieved on violations of Rule of 5 (sd = 0.768), number of oxygens & nitrogens (sd = 1.79), and number of –OH and –NHn (sd = 1.04). Various tools for profiling drug likeness have improved the efficiency of identifying potential clinical chemotherapeutic agents in the face of rising costs and complexity of synthesis and pharmacodynamics/pharmacokinetics. The Lipinski’s Rule of 5 contends that orally active drugs have the following attributes[16]: 1) Not more than 5 hydrogen donors (-OH and –NHn ) ; 2) Not more than 10 hydrogen bond acceptors (oxygen and nitrogen); 3) A Log P less than 5; 4) Formula weight less than 500. All these criteria are fulfilled by analogues 2 through 18 (see Figure 1). Melphalan analogues 19, 20, and 21 have one violation (see Figure 2), with analogues 22, 23, 24, and 25 (see Figure 2) showing two violations of Rule of 5. Compounds having violations are believed to have problems in drug absorption. Consequently, analogues 2 through 18 would be anticipated to have favorable absorption and permeation[16]. However further extensions for drug likeness include those of Ghose[17] which have acceptable values for Log P and formula weight different (Log P -0.4 to 5.6, formula weight 160 to 480) than those of Rule of 5. Anticancer activity of these analogues are not limited to breast cancer but could also include carcinoma of the central nervous system using the parameters determined to be maximal for blood-brain barrier penetration[18, 19]: 1) Log P from 1 to 4; 2) Formula weight less than 400; 3) Polar surface area less than 90 Angstroms 2 . Following these criteria then the following analogues to melphalan are determined to have potential as brain cancer chemotherapeutics: 5, 7, 11, 14, 17, and 18. A dual application imputes higher efficacious potential for these particular drug constructs. Criteria for favorable intestinal absorption have also been investigated with focus on polar surface area as the primal guiding descriptor [20]. Applying these criteria it can be determined that up to 50 % of orally administered drug would be absorbed in the intestinal tract for analogues 3, 4, 8, 10, 12, 15, 16, and 20. By polar surface area values then the following analogues would have greater than 50 % intestinal absorption: 2, 5, 6, 7, 9, 11, 13, 14, 17, 18, 19, 21, 22, 23, 24, and 25. Clearly these outcomes strongly support the potential of analogs of melphalan for administration as treatment for breast cancer and carcinoma of the central nervous system in the case of analogues 5, 7, 11, 14, 17, and 18. Clearly, the design of anticancer drugs utilizing melphalan as the parent scaffold will yield many potentially advantageous constructs.

8

Ronald Bartzatt

Determining Inter-Drug Relationships by Pattern Recognition Discriminant function analysis (DA) can classify cases (drug analogs to melphalan here) into a dichotomy[21, 22], therefore testing for separation (determining if two potential groups are distinct) . For this family of anticancer drugs (1 through 25, see Table 1) the application of DA is to classify drugs into dichotomous groups, determine the most parsimonious way to distinguish the two groups, and asses the efficacy of distinguishing the members of each group (ie. Molecular predictors). Application of pattern recognition methods have been shown previously to enhance selection of drug candidates in clinically important sectors such as the development of novel types of antihelmintic compounds[23]. Streptocococcus pneumoniae is a potentially deadly disease, in which the recognition of potentially beneficial 6-fluoroquinolone derivatives has been enhanced by application of discriminant analysis [24]. Outcome of discriminant analysis of properties presented in Table 1 produces two populations of these constructs which are distinguished from each other (ie. Maximum dissimilarity). These two populations are as follows: GROUP 1) Drug 1 (melphalan), 2, 5, 10, 11, 13, 14, 17, 18, 19, 21, 22, and 23; GROUP 2) Drug 3, 4, 6, 7, 8, 9, 12, 15, 16, 20, 24, and 25. Therefore by molecular properties inclusive of polar surface area, n-octonal/water partitioning, molecular weight, etc the constructs most similar to melphalan (2, 5, 10, 11, 13, 14, 17, 18, 19, 21, 22, and 23), an effective tool in the treatment of breast cancer, are identified and would be anticipated to convey analogous pharmaceutical activity. However the intrinsic similarities of these constructs that use melphalan as the parent scaffolding is reinforced by application of Analysis of Similarity (ANOSIM). ANOSIM tests for difference among groups of cases (drugs) within a multivariate data set[21]. This test acquires the distance measurements within the data set and converts these into ranks. In this analysis the R value closest to the numerical value of one suggests substantial dissimilarity among groups. ANOSIM analysis of Table 1 properties produce a result of R equal to 0.07117, a value acknowledging the similarity of scaffolding of constructs 2 to 25 to the parent melphalan (drug 1). Non-hierarchical K-means cluster analysis has the advantage of simplicity and and speed of analysis for the cluster analysis of large data sets[25]. The investigator designates the number of clusters for outcome however similarly to hierarchical cluster analysis the drugs are assigned to clusters to maximize the similarity among cluster members[25]. K-means analysis of Table 1 properties arranged the drugs into four clusters as follows: CLUSTER 1) Drug 3, 6, 9, 10, 19, 20, 21; CLUSTER 2) Drug 1 (melphalan), 5, 7, 14, 17, 18; CLUSTER 3) Drug 2, 4, 8, 11, 12, 13, 15, 16; CLUSTER 4) Drug 22, 23, 24, 25. This analysis achieves a higher resolution of all 25 drugs and their inter-relationships and indicates again the constructs potentially clinically similar to melphalan (ie. drugs 5, 7, 14, 17, 18).


9

Figure 3. Hierarchical cluster analysis of drugs 1 through 25 by molecular properties arrange into clusters (super node A, B, C) in which members have maximum levels of similarity. Dendrogram presentation into Cluster A) Drugs 11, 20, 21, 3, 6, 19, 4, 12, 9, 10, 15, 13, 2, and 16; Cluster B) 5, 14, 17, 1 (melphalan), 7, and 18; Cluster C) 22, 24, 25, 23. Drug 8 is distinct from all other constructs and placed into a distinct cluster. Hierarchical cluster analysis is a pattern recognition method for finding homogeneous clusters of drugs, by utilizing measured descriptors[26]. Outcome of cluster analysis is represented as a 2-way dendrogram that presents the drugs into clusters where each member is determined to be most similar to other members. Outcome for cluster analysis of Table 1 descriptors is shown in Figure 3, utilizing single linkage (nearest neighbor) conditions and Euclidean distance (geometric distance in multidimensional space). Drugs 1 through 25 fall into three cluster super nodes A, B, C, excepting drug 8, in which members have maximum levels of similarity. The 2-way dendrogram representation is into Cluster A) Drugs 11, 20, 21, 3, 6, 19, 4, 12, 9, 10, 15, 13, 2, and 16; Cluster B) 5, 14, 17, 1 (melphalan), 7, and 18; Cluster C) 22, 24, 25, 23. The parent melphalan is determined to be most similar to constructs 5, 14, 17, and 18. Drug 8 is unique in that it has a large amide group having five hydroxyl groups (-OH), which provides this drug with the highest water solubility at 250.8 milligrams/Liter in lieu of greater hydrogen bonding capability.

10

Ronald Bartzatt

Figure 4. Non-metricmultidimensional scaling arranges drugs to maximize similarity by molecular properties in which proximity suggests likeness. Grouping of constructs appears as a 23, 24, 22, 25; followed by super group 21, 20, 19, 6, 3, 9, 10, 15, 4, 12, 13, 2, 16, and 11; followed by super group 18, 7, 17, 14, 1 (melphalan), and 5. Note that construct 8 is outlier and not falling into any super group. There are two types of multidimensional scaling, metric and non-metric, and this pattern recognition method allows investigators to determine similarity or dissimilarity between objects (drugs)[27]. Non metricmultidimensional scaling (NMDS) has fewer restrictions than that of of the metric approach and less rigor[27]. NMDS is ordinal and has the further advantage of allowing the investigator to categorize and examine data where comparisons are useful. Almost any measures (parameters) can be used if utilized as similarity or dissimilarity[27]. These novel drug designs are arranged maximize similarity by molecular properties in which proximity suggests likeness. Grouping of constructs appears as a 23, 24, 22, 25; followed by super group 21, 20, 19, 6, 3, 9, 10, 15, 4, 12, 13, 2, 16, and 11; followed by super group 18, 7, 17, 14, 1 (melphalan), and 5. Note


11

that construct 8 is an outlier and not falling into any super group. Constructs 18, 7, 17, 14, and 5 are determined to be most similar to melphalan (drug 1) by the molecular properties of Table 1, and consequently anticipated to have analogous activity clinically. With these results it is apparent that alteration of the moiety that is in the para position to the N-mustard renders nitrogen mustard anticancer constructs having similar pharmaceutical properties that suggest the likelihood of similar clinical application in the treatment of breast cancer. While some properties vary considerably, it is possible to determine the underlying relationships by applying pattern recognition methods of analysis. For many constructs the pivotal properties of molecular weight, polar surface area, partition coefficient Log P, etc strongly supports the contention that the novel constructs will have favorable bioavailability (by Rule of 5) in addition to delivering the bifunctional cytotoxic nitrogen mustard moiety that is the anticancer activity of this class of anticancer drugs. In this fashion novel drug designs are produced with supporting evidence of their substantial potential in the clinical treatment of breast cancer.

Acknowledgements This work was supported by the College of Arts & Sciences, University of Nebraska, Durham Science Center, Omaha Nebraska 68182.

References [1]

[2]

[3]

[4]

Vahdat, LT; Balmaceda, C; Papadopoulos, K; Frederick, D; Donovan, D; Sharpe, E; Kaufman, E; Savage, D; Tiersten, A; Nichols, G; Haythe, J; Troxel, A; Antman, K; Hesdorffer, CS. Phase II trial of sequential high-dose chemotherapy with paciltaxel, melphalan, and cyclophosphamide, thiotepa, and carboplatin with peripheral blood progenitor support in women with responding metastatic breast cancer. Bone marrow Transplant, 2002 30(3), 149-155. Gutierrez-Delgado, F; Holmberg, LA; Hooper, H; Appelbaum, FR; Livingston, RB; Maziarz, RT; Weiden, P; Rivkin, S; Montgomery, P; Kawahara, K; Bensinger, W. High-dose busulfan, melphalan, and thiotepa as consolidation for non-inflammatory high-risk breast cancer. Bone Marrow Transplant, 2000 26(1), 51-59. Bensinger, WI; Schiffman, KS; Holmberg, L; Appelbaum, FR; Maziarz, R; Montgomery, P; Ellis, E; Rivkin, S; Weiden, P; Lilleby, K; Rowley, S; Petersdorf, S; Klarnet, JP; Nichols, W; Hertler, A; McCroskey, R; Weaver, CH; Buckner, CD. High-dose busulfan, melphalan, thiotepa, and peripheral blood stem cell infusion for the treatment of metastic breast cancer. Bone Marrow Transplantation, 1997 19(12), 1183-1189. Perez, DJ; Powles, TJ; Gazet, JC; Ford, HT; Coombes, RC. Mitomycin C, melphalan, and methotrexate combination chemotherapy for palliation of disseminated breast cancer. Cancer Chemotherapy and Pharmacology, 2004 13(1), 36-38.

12 [5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13] [14]

[15]

[16]

[17]

Ronald Bartzatt Peters, WP; Shpall, EJ; Jones, RB; Olsen, GA; Bast, RC, Gockerman, JP; Moore, JO. High-dose combination alkylating agents with bone marrow support as initial treatment for metastic breast cancer. Journal of Clinical Oncology, 1988 6(9), 1368-1378. Stemmler, HJ; Menzel, H; Salat, C; Lindhofer, H; Kahlert, S; Heinemann, V; Kolb, JH. Lasting remission following multimodal treatment in a patient with metastic breast cancer. Anti-Cancer Drugs, 2005 16(10), 1135-1137. Sarosy, G; Leyland-Jones, B; Soochan, P; Cheson, BD. The systemic administration of intravenous melphalan. American Society of Clinical Oncology, 1988 6(11), 1768-1782. Cunningham, D; Paz-Ares, L; Gore, ME; Malpas, J; Hickish, T; Nicolson, M; Meldrum, M; Viner, C; Milan, S; Selby, PJ. High-dose melphalan for multiple myeloma: long-term follow-up data. Journal of Clinical Oncology, 1994 12, 764768. Spatti, G; Regazzoni, M; Koronel, R; Zecchini, AM; De Palo, G. Adjuvant treatment with melphalan in ovarian carcinoma with no residual disease following surgery. Tumori, 1987 73(2), 157-162. McElwain, TJ; Hedley, DW; Gordon, MY; Jarman, M; Millar, JL; Pritchard, J. High dose melphalan and non-cryopreserved autologous bone marrow treatment of malignant melanoma and neuroblastoma. Experimental Hematology, 1979 7 Suppl 5, 360-371. Straus, DJ; Myers, J; Koziner, B; Lee, BJ; Clarkson, BD. Combination chemotherapy for the treatment of Hodgkin’s disease in relapse. Results with lomustine (CCNU), melphalan (Alkeran), and vindesine (DVA) alone (CAD) and in alternation with MOPP and doxorubicin (Adriamycin), bleomycin, and vinblastine (ABV). Cancer Chemotherapy and Pharmacology, 1983 11(2), 80-85. Simpson-Herren, L; Noker, PE; Wagoner, SD. Variability of tumor response to chemotherapy. II. Contribution of tumor heterogeneity. Cancer Chemotherapy and Pharmacology, 1988 22(2), 131-136. Saeki, T; Tsuruo, T; Sato, W; Nishikawsa, K. Drug resistance in chemotherapy for breast cancer. Cancer Chemother. Pharmacol., 2005 56 Suppl 7, 84-89. Moscow, JA; Johnston, PG; Cole, D; Poplack, DG; Cowan, KH. Characterization of cross-resistance to methotrexate in a human breast cancer cell line selected for resistance to melphalan. Biochem. Pharmacol., 1995 49(8), 1069-1078. Moscow, JA; Swanson, CA; Cowan, KH. Decreased melphalan accumulation in a human breast cancer cell line selected for resistance to melphalan. British Journal of Cancer, 1993 68(4), 732-737. Lipinski, CA; Lombardo, F; Dominy, BW; Feeney, PJ. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Del. Rev., 2001 46, 3-26. Ghose, AK; Viswanadhan, VN; Wendoloski, JJ. A knowledge-based approach in designing combinatorial or medicinal chemistry libraries for drug discovery. J. Combin. Chem., 1999 I, 55-68.


13

[18] van de Waterbeemd, H; Kansy, W. Hydrogen-bonding capacity and brain

penetration. Chimia, 1992 46, 299-303. [19] van de Waterbeemd, H; Camenisch, G; Folkers, G; Chretien, JR; Raevsky, OR.

[20]

[21] [22] [23]

[24]

[25] [26] [27]

Estimation of blood-brain crossing of drugs using molecular size and shape, and Hbonding descriptors. Journal of Drug Targeting, 1998 6, 151-165. Palm, K; Stenberg, P; Luthman, K; Artursson, P. Polar molecular surface properties predict the intestinal absorption of drugs in humans. Pharm. Res., 1997 14(5), 568571. Bow, ST. Pattern Recognition. New York: Marcel Dekker; 1984. McLachlan, GJ. Discriminant Analysis and Statistical Pattern Recognition. New ED edition. New York: Wiley-Interscience; 2004. Marrero-Ponce, Y; Castillo-Garit, J; Olazabal, E; Serrano, H; Morales, A; Castanedo, N; Ibarra-Velarde, F; Huesca-Guillen, A; Jorge, D; del Valle, A; Torrens, F. Tomo-Cardd, a novel approach for computer-aided rational drug design: I. Theoretical and experimental assessment of a promising method for computational screening and in silico design of new anthelmintic compounds. Journal of Computer-Aided Molecular Design, 2004 18(10), 615-634. Perez, J; Garcia, A; Carlos, F; Alvarez, I; Sanz, M. A topological-substructural molecular design (TOPS-MODE) approach to determining pharmacokinetics and pharmacological properties of 6-fluoroquinolone derivatives. European Journal of Pharmaceutics and Biopharmaceutics, 2003 56(2), 197-206. Hartigan, JA; Wong, MA. A K-means clustering algorithm. Applied Statistics, 1979 28(1), 100-108. Anderberg, MR. Cluster Analysis for Applications. San Diego: Academic Press; 1973. Borg, I; Groenen, P. Modern Multidimensional Scaling: Theory and Applications. New York: Springer-Verlag; 1997.