JAC Advance Access originally published online on January 13, 2006
Journal of Antimicrobial Chemotherapy 2006 57(3):489-497; doi:10.1093/jac/dki470
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Identification of new antimalarial drugs by linear discriminant analysis and topological virtual screening
1 INSERM U511, Immuno-biologie Cellulaire et Moléculaire des Infections Parasitaires, CHU Pitié-Salpêtrière, Assistance Publique-Hôpitaux de Paris, Université Pierre et Marie Curie, Paris, France; 2 Laboratoire de Parasitologie-Mycologie (EA3520), and Saint-Louis Hospital, Assistance Publique-Hôpitaux de Paris, 1 Avenue Claude Vellefaux, 75010 Paris, France; 3 Unidad de Investigación de Diseño de Fármacos y Conectividad Molecular, Dep. Química Física, Facultad de Farmacia, Universitat de València, Burjassot, Valencia, Spain; 4 Xarxa de Recerca de Malalties Tropicals, Dep. Biología Celular y Parasitología, Facultat de Farmàcia. Universitat de València, Burjassot, Valencia, Spain
* Corresponding author. Tel: +34-63544291; Fax: +34-63544892; E-mail: ramon.garcia{at}uv.es
Received 16 June 2005; returned 16 October 2005; revised 2 November 2005; accepted 4 December 2005
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
Objectives: A quantitative structure–activity relationship study using a database of 395 compounds previously tested against chloroquine-susceptible strains of the blood stages of Plasmodium falciparum to predict new in vitro antimalarial drugs has been developed.
Methods: Topological indices were used as structural descriptors and were related to antimalarial activity by using linear discriminant analysis (LDA) and multilinear regression (MLR). Two discriminant equations were obtained (FD1 and FD2), which allowed us to carry out successful classification of 90% and 80% of compounds, respectively. The IC50 values of the compounds were introduced to get an MLR equation model suitable to predict their in vitro activities.
Results: Using this model, a set of 27 drugs against a chloroquine-susceptible clone (3D7) of P. falciparum have been selected and evaluated in vitro. Among these drugs are monensin, nigericin, vincristine, vindesine, ethylhydrocupreine and salinomycin with in vitro IC50s at nanomolar concentrations (0.3, 0.4, 2, 6, 26 and 188 nM, respectively). Other compounds such as hycanthone, amsacrine, aphidicolin, bepridil, amiodarone, ranolazine and triclocarban showed in vitro IC50 values below 5 µM in the mathematical model.
Conclusions: These results demonstrate the usefulness of the approach for the selection and design of new lead drugs active against P. falciparum.
Keywords: molecular topology , topological indices , antimalarials , Plasmodium falciparum , QSAR studies
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
Every year, 300–500 million people contract malaria and
3 million die; most of them are children under 5 years of age. Malaria remains an important health problem in the tropics and subtropics.1 The emergence of chloroquine-resistant strains of Plasmodium falciparum and the increasing resistance against established antimalarial drugs emphasizes the demand for new effective drugs with new chemotherapeutic targets.2 In addition to in vitro and in vivo tests, which are time-consuming and complex for Plasmodium, powerful methodologies for drug database screening and selection are now available.3 Equation systems linking structure and activity (QSAR studies) are particularly relevant, and application of the mathematical models obtained to large libraries of computer-generated compounds is known as virtual computational screening.4,5 An important aspect of this method is the use of good structural descriptors that represent the molecular features responsible for the relevant biological activity. In this regard, molecular topology has proved to be a very useful technique for describing molecular structure. It follows a two-dimensional approach taking into account the internal atomic arrangement of compounds. The structure of each molecule is represented by specific subsets of topological indices (TI).6 The indices code for information on molecular size, shape and branching, the most important features of molecular structure.7 The computation of TI is very swift and the TI have the advantage of being true structural invariants, which means that their values are independent of molecular conformations. Their usefulness in the modelling of physical,8 chemical and biological properties such as various therapeutic activities as well as toxicological properties,9 the drug-like character10,11 and the molecular similarity/diversity12 has been firmly established, even within structurally heterogeneous groups of compounds. Thus, they were successfully used in the prediction of drug activities against the Mycobacterium avium complex13,14 and Toxoplasma gondii.15 Furthermore, a first topological approach to the finding of new antimalarial drugs was developed by our research group in 1999 using a small database.16 This reference is relevant since the antimalarial activity modelled there is general.
In contrast, the present article deals with a larger training database, and the objective of this study was to develop QSAR models based on TI, statistical linear discriminant analysis (LDA) and multilinear regression (MLR) analysis to identify new antimalarial drugs against chloroquine-susceptible strains of the blood stages of P. falciparum and estimate their in vitro activities.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
Database
A database of 395 antimalarial drugs has been built up from articles selected in PubMed with the following criteria: (i) the articles published between 1996 and 2003 were selected, (ii) the evaluation of antimalarial activity was performed in vitro by a radioactive isotope assay,17 (iii) the strains used were characterized either as chloroquine-susceptible by authors or by the IC50 values of chloroquine for each strain, (iv) each drug was identified by chemical structure and characterized by IC50 expressed in micromolar concentration, and (v) parasite cultures used for assays were synchronized. The database comprises 36 families of congeneric chemicals. Among these groups we have for instance naphthoquinones, natural alkaloids, amino-alcohols, amino-quinolines, biimidazoles, acyclic and cyclic peroxides, antibiotics, inhibitors of plasmepsin II, cinnamic acid derivatives, terpenoids, histone inhibitors, tebuquine, xanthone, siderophores, aminooxy compounds, oxazine, tetraoxane, sponge extract, tetraoxacycloalkanes and non-peptidic compounds.
Topological descriptors
The molecular structure descriptors (TI) used in the present research are described in Table 1 along with their definitions and references. Each compound was characterized by a set of about 120 TI specific to each molecule. We used the connectivity, electrotopological and charge indices but also other indices. All descriptors were calculated using Desmo1118 and Molconn-Z19 programs. They were computed from the adjacency topological matrix obtained from the hydrogen depleted graph.
|
LDA
Based on the information from the database, the compounds were separated into three groups, namely highly active, active and inactive, according to their IC50s around cut-offs of 0.06 and 5 µM. These cut-offs were chosen because they allowed us to clearly differentiate between highly active, active and inactive drugs.
In practice, it is usually considered that a drug is potentially interesting as an antimalarial if its IC50 is below 10 µM. However, we used a cut-off level of 5 µM to have a preliminary screening of antimalarial activity just to reduce the number of false positives. All compounds with IC50s larger or equal to 5 µM were considered as inactive. We also evaluated in vitro activity up to 5 µM.
The variables used to compute the linear classification functions are chosen stepwise, based on the Fisher–Snedecor parameter F. This means that at each step the variable that makes the larger contribution to the separation of the groups was entered as the discriminant function. Conversely, selected variables which lower the statistical significance of the classification were removed.
The discriminant ability was assessed by the percentage of correct classifications attained for each set. The classification criterion is the minimal Mahalanobis distance (distance of each case to the mean of all the cases in a category). The quality of the discriminate function was evaluated using the Wilks parameter, 
, which was obtained by a multivariate analysis of variance that tests the equality of group means for the variable in the discriminant model. LDA was then applied to the database, except for the molecules reserved as the test group, to obtain a predictive mathematical model linking structural descriptors and activity.
When there are three groups, it is useful to pay attention to the canonical variables. These satisfy two conditions: within-group variances equal one and overall mean equals zero. The first canonical variable is the linear combination of variables that best discriminates among the groups. The second canonical variable is the best linear combination orthogonal to the first one. Thus, the correlation between second and first canonical variables is zero.
The independent variables in this study were the TI and the discriminant property was antimalarial activity.
The software used for the LDA study was the BMDP New System 2 package (Cork Technology Park, Model Farm Rd), module 7M, which randomly chooses the compounds reserved for the test set.20
MLR
Correlation between the calculated TI and observed IC50s of the 246 compounds of the database was obtained by MLR to predict the IC50s of new antimalarial drugs. MLR was performed with the 9R module of the BMDP program, which estimates regression equations for best subsets of predictor variables by the Furnival–Wilson algorithm and provides detailed residual analysis.21 The lower Mallows Cp was used to identify the best subsets. Mallows' Cp = RSS/s2 – n + 2p', where RSS is the residual sum of squares for the best subset being tested, p' is the number of independent variables in the subset (including the constant), n is the number of cases and s2 is the residual mean square based on the regression using all independent variables.
Pharmacological distribution diagrams
Pharmacological distribution diagrams (PDDs) were constructed to determine the intervals of equation in which the probability of finding active compounds increase. PDDs are histogram calculated values of the mathematical functions in which expectancies appear on the ordinate axis.22 For an arbitrary interval of values of a given function, we can define an expectancy of activity as Ea = a/(i + 1), where a is the number of active compounds in the interval divided by the total number of active compounds and i is the number of inactive compounds in the interval divided by the total number of inactive compounds. The expectancy of inactivity is defined in a symmetrical way as Ei = i/(a + 1). This representation provides good visualization of the regions of minimum overlap and selects regions in which the probability of finding improved compounds is maximum.
Topological virtual screening
The selection of potential antimalarial drugs was performed from the functions obtained. Virtual screening seeking potential antimalarial activity against chloroquine-susceptible strains of P. falciparum was performed on a dataset of drugs belonging to several therapeutic categories (antineoplastics, antibacterials, antipsychotics, antiseptics, antiarrhytmics, vitamins, antivirals, antianginals, etc.). A first selection from the Merck Index23 database was established using the mathematical model, and a second selection was performed to ameliorate the prediction power and to limit the number of compounds selected as actually active. This process required the use of three models acting as filters. PDDs allowed us to carry out the assignment of thresholds useful to discriminate active from inactive compounds with the highest probability of success. Only the compounds predicted as active by all the three function values within the predetermined thresholds were identified as potential antimalarial drugs and assayed in vitro against P. falciparum.
Pharmacological tests
In vitro tests were performed on 27 drugs that were selected by the LDA and MLR models. For each compound, purchased from Sigma Aldrich (Paris, France), stock solutions were prepared in DMSO or ethanol or the appropriate solvent as described in the technical card. The solvents used were checked for their non-toxicity towards P. falciparum cultures.
In vitro tests were performed using the chloroquine-susceptible (3D7w) clone of P. falciparum which was maintained in continuous culture according to a modified version of the method of Trager and Jensen as described previously.24 The in vitro activities of the drugs were evaluated by using the method of Desjardins et al.17 modified as follows: 96-well plates were preloaded with nine concentrations (7.62 x 10–4 to 5 µM) of drugs, serial dilutions of chloroquine were placed in positive-control wells, and 200 µL aliquots of ring-stage parasitized erythrocytes (parasitaemia, 0.5%; haematocrit, 1.8%) were added to each well; all drugs were tested in triplicate. After 48 h, [3H]hypoxanthine was added to each well and then the plates were incubated for 24 h. Parasites were harvested, and incorporation of radioactivity was determined by liquid scintillation counting (Beckman LS1701, les Ulis, France). The counts per minute (cpm) were adjusted relative to controls to obtain percentages of inhibition. The IC50 was estimated by linear regression of log10 of concentration of compounds for data points between plateaus, from the highest concentration in the lower plateau to the lowest concentration in the higher plateau. For the assay, the parasites were synchronized by treatment with D-sorbitol.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
Mathematical modelling
The first LDA equation, named DF1, was designed to discriminate compounds with no significant in vitro activity against the chloroquine-susceptible strain of P. falciparum (i.e. IC50 > 10 µM) from those with probable activity (IC50 < 5 µM). This equation comprised eight independent variables:
 |
= 0.352. In the training group (179 compounds), 89 out of 98 were correctly classified as active compounds (90.8% accuracy) and 74 out of 81 were correctly classified as inactive compounds (91.4% accuracy). In the test group, consisting of 104 randomly chosen compounds, 65 out of 66 active compounds (98.5%) and 35 of 38 inactive compounds (92.1%) were correctly classified. Cross-validation (jack-knifed matrix) performed on the training group (179 compounds) shows that 71 out of 81 inactive compounds (87.7%) and 86 out of 98 active compounds (87.8%) were correctly classified. Overall, the rate of correct classification with cross-validation was 87.75%. The PDDs of this model show that for a DF1 value between –2 and 1.5 the classification of the drugs is uncertain, because of marked overlap of DF1 values of several active and inactive drugs (Figure 1). In contrast, the highest activity expectancy occurred at a DF1 value <–2.
|
The second LDA, named DF2, was designed to choose the most promising candidates from the previously selected ones according to the DF1 function. The training group was divided into three subgroups according to IC50. Compounds with IC50s > 5 µM were considered as inactive against parasite development, compounds with IC50s between 0.06 and 5 µM were considered as active, and finally values of IC50 < 0.06 µM imply the drugs to be very active against the development of P. falciparum.
We have obtained three equations with 13 independent variables. Each equation characterizes the compounds as highly active (DF2HA), active (DF2A) or inactive (DF2I):
 |
 |
 |
= 0.254. In the training group (353 compounds), 108 out of 135 inactive compounds (80.0%), 82 out of 110 active group compounds (74.5%), and 91 out of 108 highly active compounds (84.3%) were correctly classified. The cross-validation shows that, in the training group of 353 compounds, 104 out of 135 inactive (77%), 82 out of 110 active (74.5%) and 89 out of 108 highly active compounds (82.4%) were correctly classified. Figure 2 shows the canonical variable 2 versus canonical variable 1 plot with the clustering pattern for the three groups.
|
On the basis of the in vitro results, we used MLR to define a mathematical model which enables us to correlate experimental and calculated IC50s. The predicted property was the log IC50 in µM. Based on a training set of 246 molecules, the best correlation between log IC50 calculated and log IC50 experimental was obtained using the following equation:
 |
|
|
Topological virtual screening
Topological virtual screening was performed from the Merck Index database. Table 3 shows a representative list of compounds selected by the model, with the posterior probabilities for the LDA and the calculated log IC50 for the prediction equation. Of the 2000 structures from the Merck Index screened (drugs with therapeutic properties such as antiseptics, antifungals, antineoplastics, antihypertensives, antibacterials, anthelmintics, antianginals, antitussives, antiemetics, antipsychotics, insecticides, anticholesteremics, antidepressants, antivirals, vitamins, etc.) 3% of drugs have been predicted to be active against P. falciparum by our QSAR model.
|
Pharmacological test
Table 4 shows the results of the in vitro test against the chloroquine-susceptible clone (3D7w) of P. falciparum, for the 22 selected candidates out of the 63 with potential antimalarial activity. The tests were also performed for the five compounds labelled as inactive (at the end of Table 3) to confirm the validity of the model. From the 22 compounds predicted by the model as potentially active, 16 compounds showed an inhibitory effect on parasitic growth with IC50s ranging between 0.327 nM and 4.6 µM. Among these compounds, seven have IC50s below 60 nM (monensin, nigericin, vincristine, vindesine, vinblastine, quinacrine, ethylhydrocupreine) and nine compounds show IC50s below 5 µM (salinomycin, hycanthone, amsacrine, aphidicolin, bepridil, amiodarone, ranolazine, triclocarban, hexetidine). Monensin and nigericin were the most effective compounds, with IC50s of 0.327 and 0.425 nM, respectively. Six compounds had no inhibitory effect on the development of parasite at the concentrations used (i.e. estimated IC50 > 5 µM). For the five inactive-labelled candidates, four compounds were actually inactive and one, clofazimine, was slightly active with an IC50 of 968 nM.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
The main objective of this study has been to build a QSAR model which enables us to identify antimalarial compounds from molecular databases using LDA and MLR.
From the database of 395 antimalarial compounds, we have obtained two LDA equations (FD1 and FD2) capable of separating compounds according to their biological activity (active or inactive) against the chloroquine-susceptible strain of P. falciparum and an MLR equation which establishes a correlation between the IC50 and chemical structure of each drug. The selected discriminant function DF1 is capable of correctly classifying 90% of the training set and 95% of the test set and the cross-validation in the training set shows that 88% of compounds are correctly classified. The overlapping between the active and inactive groups is very small and most of the antimalarial drugs lie in the interval of DF1 between –2 and –9. About 80% of compounds are correctly classified by DF2 in the training set and above 75% in the cross-validation (leave-one-out) set. The overlapping between the groups is small, which guarantees the quality of the selected discriminant function.
The set of compounds which were selected from the Merck Index database as potential antimalarials showed a great structural diversity. These compounds concern various therapeutical areas, such as anthelmintics (hycanthone), antianginals (bepridil), antiarrhythmics (amiodarone), anticoccidials (salinomycin), antihypertensives (althiazide), antimalarials (quinacrine), antineoplastics (vinblastine, vindesine, vincristine and amsacrine), antiseptics (ethyldrocupreine, triclocarban and hexetidine), antivirals (indinavir, ritonavir and saquinavir) and vitamins (ß-tocopherol and 
-tocopherol), among others.
A bibliographic search performed after the in vitro tests with the compounds idientified as active revealed that many of these compounds had shown antimalarial activity in several experimental studies, confirming the reliability of our prediction. These results are summarized in the next paragraphs.
Satayavivad et al.25 selected the calcium blockers chlorpromazine and verapamil to test for their antimalarial activity against P. falciparum in vitro. The results disclosed that the drugs inhibited parasite population growth in the following order of IC50: verapamil 1 µM and chlorpromazine 3.5 µM. Recent studies by Skinner-Adams et al.26 have shown that antiretroviral protease inhibitors may affect outcome in malarial disease. They have investigated the antimalarial activities of six commonly used antiretroviral agents. The results show that saquinavir, ritonavir and indinavir directly inhibit the growth of P. falciparum in vitro at clinically relevant concentrations. This is the first report showing that antiretroviral PIs can directly inhibit the in vitro growth of both drug-susceptible and drug-resistant P. falciparum. These findings indicate that PI drugs may aid in the removal of Plasmodium parasites when administered to individuals co-infected with malaria. Phenothiazine drugs—fluphenazine, chlorpromazine, methotrimeprazine and trifluoperazine—were evaluated as modulating agents against Brazilian chloroquine-resistant fresh isolates of P. falciparum. All the drugs demonstrated intrinsic antiplasmodial activity at concentrations lower than those described in the literature. In addition, IC50 estimates have been shown to be inferior to the usual antipsychotic therapeutic concentrations. Statistical analysis also suggested a predominant antiparasitic effect of phenothiazine over chloroquine when used in combination.27 Usanga et al.28 reported that the mitotic inhibitor vinblastine is highly toxic to the malarial parasite P. falciparum. In cultures, in vitro growth is inhibited by 50% at a vinblastine concentration of 28 nM and totally abolished at a concentration of 100 nM. Results obtained by Neerja et al.29 suggest that a combination of azithromycin with the second-line treatment regimen of fansidar may enhance the therapeutic efficacy of the latter and also provide better prophylaxis against P. falciparum malaria. Adovelande et al.30 showed that monensin and nigericin exhibit in vitro intrinsic antimalarial activities at low concentrations.
All these results confirm the effectiveness of the topological model proposed here for the search and selection of new potential antimalarial drugs. The results obtained suggest that the cut-off of 5 µM is useful to identify new drugs on large databases.
As said, 63 compounds were initially selected for the antimalarial activity in vitro test, but only 22 were assayed because they were commercially available and were not known antimalarials. Out of the 22 compounds selected as theoretically active, 16 showed experimental antimalarial activity with IC50s < 5 µM. Vincristine, vindesine, ethylhydrocupreine, salinomycin and clofazimine stand out with IC50s < 1 µM (2, 6, 26, 188 and 968 nM, respectively). The most potent compounds were monensin and nigericin, with IC50 values of 0.3 and 0.4 nM, respectively, although these compounds have already been tested on asynchronous cultures of parasite in previous works.30 It is interesting that among the compounds selected as theoretically inactive (bithionol, bupropion, clofazimine, dapsone and sotalol) all except clofazimine showed IC50s > 5 µM.
The potential of the method is assessed by the fact that known antimalarials, not previously included in the training set, have been identified. That is, the antimalarial active compounds were selected exclusively on the basis of their corresponding structural similarity with the training set and this is independent of whether or not the molecules are reported in the literature, so that for the model they were new. The identification of pharmacophores could be performed attending to the virtual substitution of radicals and its corresponding activity variation.
Our model has the capability to discriminate between active and inactive compounds. We selected potential antimalarial drugs with the mathematical model, and the cytotoxicity of these compounds was evaluated in vitro. These cytotoxicity assays were performed to verify the discriminant capability to distinguish between specific antimalarial compounds and cytotoxic agents (data not shown in this article). We evaluated the cytotoxicity towards hepatocytes with an MTT assay for the 16 active compounds against P. falciparum and only one compound was found to be toxic (hexetidine). For the other compounds, the toxic concentrations (50%) obtained in vitro were largely superior to their IC50s. This result indicates that the mathematical model established can discriminate specific antimalarial activity and toxicity. In addition, this model could eventually be completed by an equation predicting the toxicity of antimalarial drugs.
These results demonstrate the effectiveness of molecular topology as a powerful tool for the search of new antimalarial drugs. In this work we presented a valid model for the prediction of the antimalarial activity of potential drugs for the erythrocytic stage of chloroquine-susceptible strains.
|
Transparency declarations |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
The authors have no conflicting financial interests.
|
Acknowledgements |
|---|
N. M. was financially supported by the Ministère de l'Education Nationale, de la Recherche et de la Technologie and by the Raoul Follereau Association. We acknowledge the financial support from the following sources: Cooperation Franco-Espagnol en recherche medicale, accord INSERM-CSIC projet conjoints-2002–2005; Projet PAL+ Minestère de la Recherche, Spanish Red de Investigación de Centros de Enfermedades Tropicales, RICET, (C03/04) of the Fondo de Investigación Sanitaria, Ministerio de Sanidad, Spain, grant number BQU2003-07420-C05 of the former ‘Ministerio de Ciencia y Tecnología’ within the Spanish Plan Nacional I+D.
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionTransparency declarationsReferences |
|---|
1. World Health Organization. World malaria situation in 1994: population at risk. Wkly Epidemiol Rec 1997; 72: 269–74.[Medline]
2. Ridley RG. Medical need, scientific opportunity and the drive for antimalarial drugs. Nature 2002; 415: 686–93.[CrossRef][Medline]
3. Kleinberg ML, Wanke LA. New approaches and technologies in drug design and discovery. Am J Health Syst Pharm 1995; 52: 1323–36; 1341–3.
4. de Julián-Ortiz JV, Gálvez J, Muñoz-Collado C et al. Virtual combinatorial syntheses and computational screening of new potential anti-herpes compounds. J Med Chem 1999; 42: 3308–14.[CrossRef][Web of Science][Medline]
5. Lahana R. Chimie combinatoire virtuelle. Pour la Science: Edition francaise de Scientific American 1997; 241: 56–8.
6. Balaban AT. Topological indices based on topological distances in molecular graphs. Pure App Chem 1983; 55: 199–206.
7. Kier LB, Hall LH. Molecular Connectivity in Structure-Activity Analysis. Letchworth, UK: John Wiley & Sons, 1986; 225–46.
8. Devillers J, Balaban AT. Topological Indices and Related Descriptors in QSAR and QSPAR. Amsterdam: Overseas Publishers Association, 1999.
9. Basak SC, Mills DR, Balaban AT et al. Prediction of mutagenicity of aromatic and heteroaromatic amines from structure: a hierarchical QSAR approach. J Chem Inf Comput Sci 2001; 41: 671–8.[CrossRef][Web of Science][Medline]
10. Gálvez J, García-Domenech R, de Julián-Ortiz JV et al. Topological approach to drug design. J Chem Inf Comput Sci 1995; 35: 272–84. Erratum: J Chem Inf Comput Sci 1995; 35: 938.[CrossRef]
11. Gálvez J, de Julián-Ortiz JV, García-Domenech R. General topological patterns of known drugs. J Mol Graph Model 2001; 20: 84–94.[CrossRef][Web of Science][Medline]
12. Ivanciuc O, Klein DJ. Computing Wiener-type indices for virtual combinatorial libraries generated from heteroatom-containing building blocks. J Chem Inf Comput Sci 2002; 42: 8–22.[CrossRef][Web of Science][Medline]
13.
Gozalbes R, Brun-Pascaud M, García-Domenech R et al. Prediction of quinolone activity against Mycobacterium avium by molecular topology and virtual computational screening. Antimicrob Agents Chemother 2000; 44: 2764–70.
14.
García-García A, Gálvez J, de Julián-Ortiz JV et al. New agents active against Mycobacterium avium complex selected by molecular topology: a virtual screening method. J Antimicrob Chemother 2004; 53: 65–73.
15.
Gozalbes R, Brun-Pascaud M, García-Domenech R et al. Anti-toxoplasma activities of 24 quinolones and fluoroquinolones in vitro: prediction of activity by molecular topology and virtual computational techniques. Antimicrob Agents Chemother 2000; 44: 2771–76.
16. Gozalbes R, Gálvez J, Moreno A et al. New antimalarials found by molecular connectivity. J Pharm Pharmacol 1999; 51: 111–7.[CrossRef][Web of Science][Medline]
17.
Desjardins RE, Canfield CJ, Haynes JD et al. Quantitative assessment of antimalarial activity in vitro by a semiautomated microdilution technique. Antimicrob Agents Chemother 1979; 16: 710–8.
18. García-Domenech R. DESMOL11 Software. Unidad de Investigación de Diseño de Fármacos y Conectividad Molecular, Facultad de Farmacia, Universitat de Valencia, Spain.
19. Hall LH. Molconn-Z (version 3.0). EastemNazaree College, Quincy, MA, USA.
20. Dixon WJ, Brown MB, Engelman L et al. BMDP Statistical Software Manual, Vol I. University of California Press, 1990; 3390–4358.
21. Furnival GM, Wilson RW. Regresions by leaps and bounds. Technometrics 1974; 16: 499–511.[CrossRef]
22. Gálvez J, García-Domenech R, de Gregorio-Alapont C et al. Pharmacological distribution diagrams: a tool for de novo drug design. J Mol Graph 1996; 14: 272–6.[CrossRef][Web of Science][Medline]
23. Budavari S, Ed. The Merck Index. An encyclopaedia of chemicals, drugs, and biologicals. 12th edition. MERCK & CO., INC., Whitehouse Station, NJ, 1996.
24.
Trager W, Jensen JB. Human malaria parasites in continuous culture. Science 1976; 193: 673–5.
25. Satayavivad J, Wongsawatkul O, Bunnag D et al. Flunarizine and verapamil inhibit chloroquine-resistant Plasmodium falciparum growth in vitro. Southeast Asian J Trop Med Public Health 1987; 18: 253–8.[Medline]
26. Skinner-Adams TS, McCarthy JS, Gardiner DL et al. Antiretrovirals as antimalarial agents. J Infect Dis 2004; 190: 1998–2000.[CrossRef][Web of Science][Medline]
27. Menezes CM, Kirchgatter K, Di Santi SM et al. In vitro chloroquine resistance modulation study on fresh isolates of Brazilian Plasmodium falciparum: intrinsic antimalarial activity of phenothiazine drugs. Mem Inst Oswaldo Cruz 2002; 97: 1033–9.[Web of Science][Medline]
28. Usanga EA, O'Brien E, Luzzato L. Mitotic inhibitors arrest the growth of Plasmodium falciparum. FEBS Lett 1986; 209: 23–7.[CrossRef][Web of Science][Medline]
29. Neerja J, Puri SK. Plasmodium yoelii: activity of azithromycin in combination with pyrimethamine or sulfadoxine against blood and sporozoite induced infections in Swiss mice. Exp Parasitol 2004; 107: 120–4.[CrossRef][Web of Science][Medline]
30. Adovelande J, Schrevel J. Carboxylic ionophores in malaria chemotherapy: the effects of monensin and nigericin on Plasmodium falciparum in vitro and Plasmodium vinckei petteri in vivo. Life Sci 1996; 59: PL309–15.[CrossRef][Web of Science][Medline]
31.
Kier LB, Murray WJ, Randi
M et al. Molecular connectivity V: connectivity series concept applied to density. J Pharm Sci 1976; 65: 1226–1230.[CrossRef][Web of Science][Medline]
32. Kier LB, Hall LH. General definition of valence delta values for molecular connectivity. J Pharm Sci 1983; 72: 1170–3.[CrossRef][Web of Science][Medline]
33. Hall LH, Kier LB. Electrotopological state indices for atom types: a novel combination of electronic, topological, and valence state information. J Chem Inf Comput Sci 1995; 35: 1039–45.[CrossRef]
34. Shannon CE. A mathemathical theory of communication. Bell System Tech J 1948; 27: 379–423.
35. Gálvez J, García-Domenech R, Salabert MT et al. Charge indexes. New topological descriptors. J Chem Inf Comput Sci 1994; 34: 520–5.[CrossRef]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  N. Mahmoudi, R. Garcia-Domenech, J. Galvez, K. Farhati, J.-F. Franetich, R. Sauerwein, L. Hannoun, F. Derouin, M. Danis, and D. Mazier New Active Drugs against Liver Stages of Plasmodium Predicted by Molecular Topology Antimicrob. Agents Chemother., April 1, 2008; 52(4): 1215 - 1220. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. M. Sanchez, D. Thomas, E. J. Gillespie, R. Damoiseaux, J. Rogers, J. P. Saxe, J. Huang, M. Manchester, and K. A. Bradley Amiodarone and Bepridil Inhibit Anthrax Toxin Entry into Host Cells Antimicrob. Agents Chemother., July 1, 2007; 51(7): 2403 - 2411. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||