Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Science. Author manuscript; available in PMC 2012 August 5.
Published in final edited form as:
PMCID: PMC3396183

Chemical genomic profiling for antimalarial therapies, response signatures and molecular targets


Malaria remains a devastating disease largely because of widespread drug resistance. New drugs and a better understanding of the mechanisms of drug action and resistance are essential for fulfilling the promise of eradicating malaria. Using high-throughput chemical screening and genome-wide association analysis, we identified 32 highly active compounds and genetic loci and genes associated with differential chemical phenotypes (DCPs), defined as ≥5-fold differences in half-maximum inhibitor concentration (IC50) between parasite lines. Chromosomal loci associated with 49 DCPs were confirmed by linkage analysis and tests of genetically modified parasites, including three genes that were linked to 96% of the DCPs. Drugs whose responses mapped to wild type or mutant pfcrt alleles were tested in combination in vitro and in vivo, yielding promising new leads for antimalarial treatments.

Keywords: Plasmodium falciparum, high-throughput screening, genetic mapping, chemical genomics, phenotype

The deployment of artemisinin (ART) and its derivatives against Plasmodium falciparum malaria parasites has been effective, and ART-based combination therapies (ACTs) are currently the recommended treatment in most endemic regions (1). The choice of partner drug is critical; an ideal partner drug should have pharmacokinetic and pharmacodynamic properties compatible with ART, employ a mode of action different from that of ART, retain efficacy against existing populations of drug-resistant parasites, and have no adverse pharmacologic interactions or additional toxicity (2). Unfortunately, parasites resistant to ART and its current partner drugs have been reported (3-5). New drugs or combinations are therefore urgently needed. Indeed, some promising leads have recently been identified through large-scale screening (6-12). Combinations of new or existing drugs that are synergistic or act on variant forms of parasite targets may mitigate the emergence of drug resistance. Here we have used quantitative high-throughput screening (qHTS) and genome-wide association and linkage analyses to identify candidate new antimalarial drugs with complementary or distinct response signatures for effective combination therapies. We show that many of the responses to a diverse collection of compounds are determined by a surprisingly limited number of genes, a finding that has broad implications for antimalarial drug development.

Chemical library screens for inhibitors and DCPs

Sixty-one parasite lines (Table S1) were screened against the NIH Chemical Genomics Center Pharmaceutical Collection containing 2816 compounds registered or approved for human or animal use (13). The compounds were tested at eight 5-fold serial dilutions (from 29 μM to 0.5 nM) using a parasite growth inhibition assay (9, 14) to obtain dose-response curves and IC50 determinations for each compound. From 171,776 drug assays that generated ~1.4 million data points, we identified 32 highly active compounds that inhibited the growth of at least 45 parasite lines with IC50 values ≤ 1 μM (Table 1). Among these pan-active compounds, seven (ecteinascidin 743, gramicidin, artenimol, decoquinate, epothilone B, atovaquone, and actinomycin D) yielded mean IC50 values lower than that of ART (IC50 < 10 nM) and, to our knowledge, 10 have not been reported to have antimalarial activity. Pairwise comparison of the IC50 values among 61 parasite lines identified 72,538 DCPs from 689 compounds, including 161 compounds that elicited DCPs between the parents of three P. falciparum laboratory crosses (7G8×GB4, Dd2×HB3, and 3D7×HB3) (Table S2). As most of these compounds have been approved for human use, their ability to inhibit parasite growth at nanomolar levels make them promising candidates for developing new antimalarial drugs or drug combinations. These DCPs can be investigated using genome-wide association studies (GWAS) or linkage analysis to identify genetic determinants of susceptibility and study mechanisms of differential drug sensitivity.

Table 1
Compounds highly active against multiple Plasmodium falciparum isolates

Compounds with correlated responses and response signature groups

Compounds with positively correlated response patterns may act on common pathways or targets within the parasite. To search for compounds with correlated response patterns, we performed pairwise comparisons of 492 compounds (Table S3) that were active (see SOM for active compound definition) against at least one-third of the 61 lines profiled and identified 2082 pairs of compounds with highly correlated responses (correlation coefficient, [CC] > 0.7) among these lines using methods described (21). For example, responses to 52 compounds were highly correlated with that of ART, 40 with that of mefloquine, and 25 with both of ART and mefloquine. The 25 compounds included halofantrine, lumefantrine, dihydroergocristine, ergotamine, rifapentine, and bromocriptine. Correlations of parasite responses to ART, mefloquine, halofantrine, and lumefantrine have been documented in studies of culture-adapted parasites and in vivo patient isolates (15-17). The observations suggest that the compounds with highly correlated responses may share common features of drug action and/or resistance.

To investigate this idea further, we clustered the compounds into groups of common chemical response signatures using K-means and Dunn’s index algorithms (18, 19). The 492 compounds were grouped into 44 clusters, with some having relatively high activity indices (AIs, ranging from 0 to 1 with 1 being 100% correlation) that suggest potential common mechanisms of response (19) (Fig. 1A and Table S4). Although the compound library was screened against 61 parasite lines, the numbers of clusters reached a plateau with just 10 lines (Fig. 1A). This result suggests that parasite responses to the compounds are restricted to a few common pathways. Based on their response patterns, parasite lines generally clustered according to their geographic origins (Fig. 1B). Distinct groups included one from South America, two from Cambodia, and a cluster of Africa/Central America/Thailand lines that could represent a recent parasite expansion from Africa. One of the Cambodian clusters exhibited high IC50 values to most of the compounds, reminiscent of an earlier description of SE Asian parasites with an accelerated resistance to multiple drugs phenotype (20), and might represent an emerging ART-resistant population. These patterns of chemical response may reflect the population separation of Cambodian parasites that has been proposed on the basis of genome-wide microsatellite (MS) and single-nucleotide polymorphism (SNP) data (17, 21) and suggest that antimalarial drugs have played an important role in the recent evolution and population structure of P. falciparum.

Fig. 1
Clusters of chemical response patterns and major compounds in the components separating parasite populations

The compounds within most clusters generally showed strongly correlated responses and, in some cases, had similar structures as measured by a structure index (SI; values ranging from 0 to 1, with higher values having more similar structures) (22). For example, cluster 26 (AI = 0.55 and SI = 0.4) comprises ten compounds, eight of which contain a quinoline core, including chloroquine (CQ), cinchonine, quinacrine, quinidine, and quinocidum (Fig. S1 and Table S4). P. falciparum responses to various quinolines have been associated with the P. falciparum CQ resistance transporter (PfCRT) and multidrug resistance 1 protein (PfMDR1, also named P-glycoprotein homolog 1), both of which are proposed to function as transporters (23-25). Clusters 28 and 40 were characterized by high AI values (0.73) but relatively low SI values (0.15 and 0.2, respectively), suggesting highly correlated responses among compounds of diverse structures. Further investigation of these clusters and their targets may reveal how parasites respond to structurally diverse compounds.

Mutations enabling resistance to one compound may render parasites more sensitive to other compounds; in cases of negative correlation, pairs of such compounds may be good candidates for drug combination therapy, particularly if they are complementary in their actions on wild type and mutant forms of a parasite target. Pairwise comparison of the 492 active compounds identified 1,250 pairs that were negatively correlated, with CC ranging from −0.26 to -0.90 (Table S5). In particular, 43 compounds were negatively correlated with CQ, including ciprofloxacin and clindamycin, each of which has been reported to have antimalarial activity alone (26, 27). The combination of clindamycin and CQ (CC = −0.27) substantially improves patient cure rate as compared with CQ alone (94% compared to 32%) (26), and ciprofloxacin (CC = −0.30) significantly enhances CQ activity in vitro (27). Tylosin, a veterinary drug used to treat bacterial infections, is negatively correlated with CQ as well (CC = -0.43) and synergizes with it to inhibit P. falciparum in vitro (28). Further investigation of the compounds negatively correlated with CQ response could lead to new drug combinations for treating resistant parasites (also see below).

Tocainide and etonogestrel showed marked negative correlations with responses to ART or mefloquine (CC = −0.52 to −0.67) (Table S5), warranting further evaluation as candidates for new classes of ACTs. These compounds were also negatively correlated with quinine, although to a somewhat lesser extent (CC = −0.34 and −0.42, respectively), suggesting potential shared mechanisms in response to quinine, ART, and mefloquine.

Impact of antimalarial drugs on parasite populations

To investigate how antimalarial drugs influence parasite evolution and population structure, we analyzed parasite responses to 134 compounds that produced an IC50 value with a good quality dose-response curve (curve class.1.1, 1.2, and 2.1, see SOM) in each of the 61 parasite lines. Consistent with the geographic clustering, principal component analysis (PCA) separated the American parasites from the African and Asian parasites by component 2 and the Asian parasites from the African parasites by component 5 (Fig. 1C). Compounds contributing most positively to component 5 were predominantly the quinoline antimalarial drugs in cluster 26, while compounds contributing negatively to this component included mibefradil, vinorelbine, and homoharringtonine (Table S6). These results suggest that CQ and the other quinoline drugs have played a substantial role in the evolution of these two parasite populations. Similarly, the compounds contributing most negatively to component 2 that separates the New and Old World parasites included dihydroergotamine, dihydroergocristine, and reserpine; responses to these drugs were mapped to pfmdr1 in the progeny of the P. falciparum crosses (see below). Compounds contributing most positively to this component included docetaxel and clobetasone. Although these compounds have not been used to treat malaria, they could be linked to (mutant) alleles of parasite molecules (possibly pfmdr1) selected by other antimalarial drugs. These data suggest that pfmdr1 played a significant role in shaping the P. falciparum populations in America. The two groups of compounds showing positive or negative contributions to component 2 (eigenvector values ≤ 0.14 and those > 0.14) were themselves positively correlated in response patterns within each group, but were largely negatively correlated between the groups (Fig. 1D). Similar correlations were observed for the compounds in component 5, for example, CQ and mibefradil (Fig. 1E). The compounds with negatively correlated responses provide additional starting points for novel combination therapies.

GWAS of parasite responses with SNPs

To identify genes that may associate with the differences in responses to the active compounds, we performed GWAS using 3,354 SNPs collected previously to search for SNPs associated with differential drug responses (17). First, we investigated whether known resistance determinants could be identified using this method. Indeed, mutations in pfcrt were significantly (corrected P < 0.01) associated with responses to more than 200 compounds, including the expected quinolines such as hydroxychloroquine, CQ, quinine, and quinacrine (Table S7). Similarly, responses to dihydrofolate reductase (DHFR) inhibitors such as trimethoprim, trimetrexate, and triamterene were significantly (corrected P < 0.005) associated with mutations in pfdhfr (Table S8). Compounds associated with mutations in the pfmdr1included dihydroergotamine and lumefantrine, consistent with previous reports (9, 29) (Table S9). Additionally, the major compounds contributing to PCA component 2 separating the South American population from those of Asia and Africa were also associated with polymorphisms of pfmdr1 (Table S9). These results show that known mutations strongly influencing parasite drug responses can be identified using GWAS.

Fifteen genes were significantly (corrected P-values < 0.005) associated with responses to both ART and mefloquine, consistent with the presence of these drugs in cluster 23 (Tables S4, Table S10, and Table S11). The top six genes associated with responses to ART and its derivatives were MAL13P1.268 (Plasmodium conserved protein), PF11_0188 (heat shock protein 90), PFE0565w (conserved Plasmodium protein), PF08_0130 (rRNA processing WD-repeat protein), PFA0655w (SURFIN), and PFI0355c (ATP-dependent heat shock protein). Additionally, the majority of the genes associated with ART response were also associated with responses to derivatives such as artemisininum, artenimol, and artemetero (Table S10). Parasite response to primaquine was also significantly associated with MAL13P1.268 (corrected P-value = 1.74E-04) (Table S12). Since MAL13P1.268 was associated with responses to primaquine, ART, mefloquine, dihydroergotamine, and dihydroergocristine, the role of this gene in drug resistance should be studied further. Association analysis of 26 PCA-corrected parasites from the Thai-Cambodian border linked many SNPs with different drug responses; however, the genes associated with responses to CQ, mefloquine, ART, and antifolate drugs were not evident in this analysis because these parasites were all resistant to these drugs (Table S13). Although some of the associated genes identified in this study could be false positives, these studies provide a starting point for further functional investigations of the genes and their roles in antimalarial drug resistance.

Three genetic loci linked to the majority of DCPs

To identify the genetic determinants that contribute to the DCPs and to further investigate some of the compounds associated with mutations in pfcrt and pfmdr1, we tested 128 DCP compounds on 33 recombinant progeny of the P. falciparum 7G8×GB4 cross and 98 DCP compounds on 34 recombinant progeny of the Dd2×HB3 cross (Table S1, and S14). Using quantitative trait loci (QTL) analysis, we mapped 49 DCPs to 57 chromosomal loci with a logarithm (base 10) of odds (LOD) score ≥ 3.0. Remarkably, 47 of the 49 (96%) DCPs were mapped to three loci containing pfdhfr (chromosome 4), pfmdr1 (chromosome 5), or pfcrt (chromosome 7) (Fig.2 and Table S14). These results indicate that pfdhfr, pfmdr1, and pfcrt dominate the parasite’s differential response to many drugs. Additionally, eight DCPs were mapped to loci on chromosomes 3, 7, 8, 12 and 14, with DNA segments ranging from 55 kb to 193 kb (Table S14 and S15 and SOM text). Further studies are needed to identify the mutations conferring these DCPs.

Fig. 2
Genetic loci linked to differential chemical phenotypes determined using progeny from two genetic crosses

The linkage analyses also confirmed many differential compound sensitivities associated with pfcrt, pfmdr1 or pfdhfr by GWAS. Responses to 11 compounds associated with pfcrt, three antifolate drugs with pfdhfr, and 15 compounds with pfmdr1 were confirmed by linkage analysis and/or testing of parasites with genetically modified pfcrt or pfmdr1 (Tables S7-9 and Table S14) (also see below). These results demonstrate the usefulness of GWAS for detecting mutations mediating drug resistance in malaria parasites.

Confirmation of PfDHFR as the target of trimethoprim and triamterene using parasites with different mutant pfdhfr alleles has been reported (9). To determine whether pfcrt and pfmdr1 are indeed responsible for the DCPs that map to their respective loci, we tested the DCP compounds against parasites with genetically modified pfmdr1 or pfcrt genes (Table S14 and S16). There are five common amino acid polymorphisms in PfMDR1 (30), and changes at positions S1034C, N1042D, and D1246Y have been shown to confer differential sensitivity to some drugs (9). In two progeny of the Dd2×HB3 cross, the CQ-sensitive GC03 parasite and the CQ-resistant 3BA6 parasite, conversion of the SDD pfmdr1 allele (encoding amino acids S1034, D1042, and D1246) to a SND allele decreased sensitivity to some antimalarial agents, while conversion to a CDY allele resulted in increased sensitivity. For the eight DCP responses mapped to pfmdr1 in Dd2×HB3 and 7G8×GB4 crosses, the IC50 values were 3- to 23- fold higher for parasites carrying the SND allele versus those carrying the CDY allele. As expected, control parasites in which the allelic exchange produced no change from pfmdr1 SDD resulted in less than two-fold changes in IC50 values for all DCPs mapped to the pfmdr1 locus (Table S14). Comparison of IC50 ratios from a genetically modified parasite receiving the parental SND pfmdr1 allele over the one receiving the CDY allele showed significantly higher mean ratios (unpaired t-test, P<0.001) for the DCP compounds that mapped to pfmdr1 than those that did not (Table S14). These particular amino acid substitutions therefore mediate the DCPs that map to this locus.

Among the 23 DCPs linked to chromosome 7, two-fold or greater changes in IC50 value were seen for 19 DCPs in the GC03 parasites engineered to carry the mutant pfcrt alleles of Dd2 or 7G8 (GC03Dd2, GC037G8) (Table S14). Similarly, the mean IC50 ratios for compounds mapped to pfcrt were significantly higher than those not mapped to pfcrt (P=0.002) in the recombinant GC03Dd2 parasites. In the genetically modified parasites, IC50 values were elevated for four compounds and decreased for 15, which suggests specific interactions with wild type or mutant pfcrt alleles, respectively (Table S14). Responses to many compounds such as perhexiline, lopinavir, lumefantrine, carbetapentane, memantine, and duloxetine that did not map to any loci were also altered in the pfcrt-or pfmdr1-modified parasites, suggesting that these compounds target or are transported by PfCRT or PfMDR1, both of which reside on the membrane of the intraerythrocytic parasite’s digestive vacuole, the site of hemoglobin degradation (23, 31).

The compounds generating the DCPs linked to pfcrt have amine groups that could become protonated at physiologic or lower pH (Fig. S2). A plot of the parasite response patterns showed two groups of compounds with positive correlation in responses within each group, but negative correlation between groups (Fig. 3A). Except for the presence of a basic amine, however, there was no consistent structural basis for separating the two groups of compounds whose responses correlated positively or negatively with that of CQ (Fig. S2). Although agents with structures similar to some of these compounds have been tested for ‘reversal’ of CQ resistance (32-35), our data provide genetic evidence showing that the mechanism of this reversal is specific to mutant PfCRT, consistent with results from a recent study (36). These compounds may act as blockers of the PfCRT drug transport pore, as they all carry a basic nitrogen group (Fig. S2) that may impede the mutant PfCRT-mediated flux of CQ out of the digestive vacuole (37). As expected, the DCPs linked to pfmdr1 also included compounds with diverse structures, including dihydroergotamine, dihydroergocristine, miconazole, and rifampin in cluster 40; ergotamine and zeaxanthin in cluster 41; and reserpine, loe 908, and sorafenib in cluster 36.

Fig. 3
Correlated responses among the progeny of the Dd2×HB3 cross and changes in parasite sensitivity against compounds mapped to pfcrt in the presence of low-dose chloroquine in vitro and in vivo

Drug combinations targeting PfCRT

CQ is no longer effective in treating P. falciparum infections because of mutations in pfcrt (38). Identification of compounds that interact with either wild type or mutant forms of PfCRT may allow the development of novel combination therapies to help overcome CQ resistance. Mibefradil, NNC55-0396, and other compounds whose response profiles negatively correlated with CQ (Fig. 3A; Tables S5, S6, and S7) could be combined with CQ for treating both CQ-resistant and sensitive parasites. Indeed, when tested with a low concentration of CQ (IC15) in vitro, these combinations proved to be highly potent against the CQ resistant line Dd2 (Fig. 3B and Table S14). Among the 23 compounds that mapped to the pfcrt locus, 17 showed reduced IC50 values (21 to 229 fold) in the presence of a low dose of CQ (IC15), including six with IC50 values that dropped to below 100 nM. Such combinations offer a promising approach for treating CQ-resistant parasites and could prevent new resistance mutations because of structural and functional constraints on PfCRT. Similar strategies can be used to develop combinations targeting PfMDR1.

To further investigate the interactions of compounds negatively associated with CQ, we performed isobologram analysis of the combinations of CQ with either mibefradil or NNC55-0396 against the CQ-resistant line Dd2 and the CQ-sensitive line HB3. The actions of mibefradil and NNC55-0396 (both T-and L-type Ca++ channel blockers) were synergistic with that of CQ against Dd2 (Fig. 3C). Verapamil, a L-type Ca++ channel blocker and a known CQ resistance reversal agent (36), is synergistic with CQ in Dd2, but antagonistic in HB3 (Fig. S3). However, the IC50 of verapamil against Dd2 is 3-4 fold higher than of mibefradil and NNC55-0396 (1.3 μM for verapamil vs. ~200 nM for NNC55-0396 and ~300 nM for mibefradil). These results suggest that either mibefradil or NNC55-0396 may be more effective than verapamil in combination with CQ in treating the drug resistant parasites.

We also tested these combinations against a CQ resistant strain of Plasmodium chabaudi in vivo and showed that inclusion of a low dose of CQ (3mg/kg) with NNC55-0396 or mibefradil greatly increased the effect of these drugs, leading to ~15 and ~40 fold reduction in the treatment concentrations of NNC55-0396 and mibefradil, respectively (Fig. 3D-3F). Mibefradil is rapidly absorbed in vivo, reaching peak plasma concentration within 1-2 hours, and has an elimination half-life of 17-25 hours (39)(SOM), indicating that this drug may have a suitable pharmacokinetic profile for combination treatments. The results are consistent with the in vitro data from P. falciparum and suggest that the P. chabaudi homolog of pfcrt may be involved in CQ resistance in this rodent model, but requires further investigation.

Our analyses show that the majority of differential sensitivity of current P. falciparum populations to many compounds is linked to mutations in pfdhfr, pfmdr1, or pfcrt, suggesting that the number of parasite genes that can contribute to drug responses may be limited. Analyses of DCPs and genome-wide MS and SNP genotypes from the progeny of two genetic crosses and field isolates identified many compounds interacting with wild type or mutant pfcrt or pfmdr1, which were confirmed using genetically modified parasites and drug combinations. In addition to the 32 highly potent pan-active compounds, the compounds active against wild type or mutant forms of PfCRT or PfMDR1 offer novel strategies for antimalarial therapies, and the loci and genes associated with drug responses provide a genetic basis to better delineate the nature of drug resistance in malaria.

Supplementary Material


Suppl Table 1

Suppl Table 10

Suppl Table 11

Suppl Table 12

Suppl Table 13

Suppl Table 14

Suppl Table 15

Suppl Table 16

Suppl Table 2

Suppl Table 3

Suppl Table 4

Suppl Table 5

Suppl Table 6

Suppl Table 7

Suppl Table 8

Suppl Table 9


This work was supported by the Intramural Research Program of the Division of Intramural Research at the National Institute of Allergy and Infectious Diseases, the National Human Genome Research Institute, and the Director’s Challenge Award Program, all at the National Institutes of Health. Funding for the studies from D.A.F. were provided by R01 AI50234 and the Burroughs Wellcome Fund. We thank Paul Shinn and Danielle Van Leer for compound management, Jennifer Wichterman for assay assistance, Richard Eastman and Connor O’Brien for comments on the manuscript, and NIAID intramural editor Brenda Rae Marshall for assistance. The raw data from our chemical screening has been deposited at PubChem with accession number 504749 (

Because all authors except D.A.F. are government employees and this is a government work, the work is in the public domain in the United States. Notwithstanding any other agreements, the NIH reserves the right to provide the work to PubMedCentral for display and use by the public, and PubMedCentral may tag or modify the work consistent with its customary practices. Rights outside of the U.S. can be established subject to a government use license.


Author contributions: J.Y. performed drug assay qHTS, parasite culture, and data analysis; C.-c. C. carried out qHTS and data analysis; R.L.J. and R.H. performed data analysis and writing; S.P. performed in vivo tests of drug combinations; A.L. performed isobologram analyses; D.A.F. transfected parasites and assisted with writing; T.E.W. provided progeny and writing; C.P.A. and J.I. carried out project planning, support, and writing; and X.-z. S. provided project conception, data analysis, and writing.

Competing Interest: The authors declare that they do not have any competing financial interests.

Supporting Online Material

Materials and Methods

Figs S1 to S3

Table S1 to S16


1. White NJ. Science. 2008;320:330. [PubMed]
2. White NJ, Olliaro PL. Parasitol Today. 1996;12:399. [PubMed]
3. Noedl H, et al. N Engl J Med. 2008;359:2619. [PubMed]
4. Bonnet M, et al. Malar J. 2009;8:192. [PMC free article] [PubMed]
5. Dondorp AM, et al. N Engl J Med. 2009;361:455. [PMC free article] [PubMed]
6. Chong CR, et al. Nat Chem Biol. 2006;2:415. [PubMed]
7. Weisman JL, et al. Chem Biol Drug Des. 2006;67:409. [PMC free article] [PubMed]
8. Kato N, et al. Nat Chem Biol. 2008;4:347. [PubMed]
9. Yuan J, et al. Nat Chem Biol. 2009;5:765. [PMC free article] [PubMed]
10. Guiguemde WA, et al. Nature. 2010;465:311. [PMC free article] [PubMed]
11. Gamo FJ, et al. Nature. 2010;465:305. [PubMed]
12. Rottmann M, et al. Science. 2010;329:1175. [PMC free article] [PubMed]
13. Huang R, et al. Sci Transl Med. 2011;3:80ps16. [PMC free article] [PubMed]
14. Inglese J, et al. Proc Natl Acad Sci U S A. 2006;103:11473. [PubMed]
15. Basco LK, Le Bras J. Am J Trop Med Hyg. 1993;49:301. [PubMed]
16. Brockman A, et al. Trans R Soc Trop Med Hyg. 2000;94:537. [PubMed]
17. Mu J, et al. Nat Genet. 2010;42:268. [PMC free article] [PubMed]
18. Seed J, et al. Crit Rev Toxicol. 2005;35:664. [PubMed]
19. Huang R, et al. Chem Res Toxicol. 2008;21:659. [PMC free article] [PubMed]
20. Rathod PK, et al. Proc Natl Acad Sci U S A. 1997;94:9389. [PubMed]
21. Wootton JC, et al. Nature. 2002;418:320. [PubMed]
22. Randić M. J Chem Inf Comput Sci. 1997;37:672.
23. Fidock DA, et al. Mol Cell. 2000;6:861. [PMC free article] [PubMed]
24. Ferdig MT, et al. Mol Microbiol. 2004;52:985. [PubMed]
25. Sa JM, et al. Proc Natl Acad Sci U S A. 2009;106:18883. [PubMed]
26. Lell B, Kremsner PG. Antimicrob Agents Chemother. 2002;46:2315. [PMC free article] [PubMed]
27. Kazzim OJ, et al. Ann Trop Med Parasitol. 2006;100:579. [PubMed]
28. Gershon P, Howells RE. Trans R Soc Trop Med Hyg. 1986;80:753. [PubMed]
29. Mungthin M, et al. Am J Trop Med Hyg. 2010;83:1005. [PMC free article] [PubMed]
30. Foote SJ, et al. Nature. 1990;345:255. [PubMed]
31. Cowman AF, et al. J Cell Biol. 1991;113:1033. [PMC free article] [PubMed]
32. Kyle DE, et al. Am J Trop Med Hyg. 1993;48:126. [PubMed]
33. Sowunmi A. Ann Trop Med Parasitol. 2003;97:103. [PubMed]
34. Singh N, Puri SK. Acta Trop. 2000;77:185. [PubMed]
35. Burgess SJ, et al. J Med Chem. 2010;53:6477. [PMC free article] [PubMed]
36. Patel JJ, et al. Mol Microbiol. 2010;78:770. [PMC free article] [PubMed]
37. Wellems TE. Nat Med. 2004;10:1169. [PubMed]
38. Hayton K, Su XZ. Curr Genet. 2008;54:223. [PubMed]
39. Billups SJ, Carter BL. Ann Pharmacother. 1998;32:659. [PubMed]