Selective Bias Virtual Screening for Discovery of Promising Antimalarial Candidates targeting Plasmodium N-Myristoyltransferase

Malaria remains a significant public health challenge, with Plasmodium vivax being the species responsible for the most prevalent form of the disease. Given the limited therapeutic options available, the search for new antimalarials against P. vivax is urgent. This study aims to identify new inhibitors for P. vivax N-myristoyltransferase (PvNMT), an essential drug target against malaria. Through a validated virtual screening campaign, we prioritized 23 candidates for further testing. In the yeast NMT system, seven compounds exhibit a potential inhibitor phenotype. In vitro antimalarial phenotypic assays confirmed the activity of four candidates while demonstrating an absence of cytotoxicity. Enzymatic assays reveal LabMol-394 as the most promising inhibitor, displaying selectivity against the parasite and a strong correlation within the yeast system. Furthermore, molecular dynamics simulations shed some light into its binding mode. This study constitutes a substantial contribution to the exploration of a selective quinoline scaffold and provides valuable insights into the development of new antimalarial candidates.


Introduction
Malaria is a disease caused by protozoa of the Plasmodium genus, speci cally the species P. falciparum and P. vivax 1 that cause the highest rates of mortality and morbidity, respectively.In 2022, global estimates surpassed 200 million cases, resulting in over 600 thousand deaths 2 .Vivax malaria, the most geographically widespread form of the disease, is predominantly found in tropical and subtropical regions.Cases caused by P. vivax account for approximately 72% of all malaria episodes occurring outside the African continent 3 .Notably, around 73% of reported cases in the Americas are concentrated in Venezuela, Brazil and Colombia 2,4 .P. vivax is biologically distinct from P. falciparum primarily due to the formation of hypnozoites, a latent stage that persists in the liver, posing a challenge for treatment due to the relapse episodes [5][6][7][8] .Additionally, P. vivax is characterized by its preference for reticulocyte infection 9,10 and the early production of sexual stages (gametocytes) observed in peripheral blood 11,12 .An additional concern is related to the growing resistance of P. vivax to chloroquine [13][14][15] .Despite being a frontline treatment for vivax malaria since 1946, current evidence shows the circulation of resistant strains in endemic areas across various continents, including the Americas 16 .Evaluating drug resistance in P. vivax is particularly challenging due to the di culties in in vitro culturing of the parasite.Distinguishing between treatment resistance and late relapses identi ed in in vivo assessments adds complexity to the analysis 17 .Given these challenges, there is a pressing need to explore new therapeutic options for P. vivax.
The Nmyristoyltransferase protein (NMT) plays a crucial role in catalyzing the transfer of the fatty acid myristate (C14) from the myristoyl-CoA molecule to the N-terminal glycine residue of numerous proteins 18 .In humans, this protein exhibits two isoforms, attracting attention in cancer research as important targets for cell survival 19 .In contrast, in various organisms, such as Candida albicans 20 , Trypanossoma spp. 21,22, Leishmania donovani 23,24 , Cryptosporidium parvum 25 , and Plasmodium spp. 26- 28, which express a single isoform of the protein, NMT stands out as a promising drug target.This has prompted extensive research to discover new therapeutic agents for diseases associated with these organisms 29 .Studies revealed that many Plasmodium proteins need this post-translational modi cation during various life stages of the parasite 28 , which is important for drug development of inhibitors.While inhibitors for Plasmodium NMT are currently described, achieving parasite selectivity remains a challenge.Efforts are underway to enhance the selectivity of these inhibitors for more effective targeting of the parasite.Structural studies revealed that the Plasmodium NMT enzyme is capable of forming an alternate conformation in the peptide binding pocket which preferentially binds some inhibitors compared to the human enzymes thus providing a rational hypothesis for how to achieve selectivity 30 .Structural studies of these selective hits support the identi cation of a distinct conformation of the Plasmodium enzyme which binds compounds preferentially compared to the human NMTs 25,31 .In this study (Fig. 1), we employed structural analysis and virtual screening of millions of commercially available compounds using validated, robust, and predictive computational models as lters.This process enabled us to select 23 candidates for subsequent in vitro evaluation.Our approach involved experimental testing within a developed NMT yeast system, followed by P. falciparum in vitro phenotypic and cytotoxicity assays.Subsequently, the selected candidates underwent evaluation against NMT enzyme.To deepen our understanding, molecular dynamics simulations were conducted to elucidate the binding mode of the most promising candidates.Through this synergistic integration of computational models and experimental evaluation, we successfully identi ed LabMol-394 as a promising compound targeting the translational process of malaria parasites.

Results
In silico studies Five high-a nity representative binding sites identi ed through structural analysis.
Utilizing data from PDB (https://www.rcsb.org/),we collected 59 structures of NMT from H. sapiens (isoforms 1 and 2) and P. vivax.To conduct structural analysis using the Bio3D R package, we ltered the crystals based on the presence of a ligand, absence of mutations, and high resolution (≤ 2 Å), resulting in a total of 30 structures (10 for H. sapiens and 20 for P. vivax).This comprehensive analysis involved aligning these structures followed by principal component analysis (PCA) to categorize similar conformations into clusters.Notably, our ndings revealed distinct binding modes between Plasmodium compared to H. sapiens (Fig. 2a).Building upon this insight, we focused on Plasmodium structures, identifying ve clusters showcasing variations in conformation and ligand binding (Fig. 2b).Subsequently, we selected the most representative structures based on inhibition (Ki or IC 50 ) from PDB IDs: 2YND, 2YNE, 4CAF, 4UFX and 6MB1.These structures served as the foundation for developing and validating both shape-based models and docking protocols.

Validation of PvNMT Shape-based models
The selected representative structures were employed to construct and validate shape-based models, as depicted in Fig. 3. Notably, the application of RefTverskyCombo score yielded area under the curve (AUC) values ranging from 0.69 to 0.86 across all models.Particularly, the model generated using query PDB ID 2YNE demonstrated the highest performance, achieving an AUC of 0.86, along with an Enrichment Factor (EF) of 7.76 and BEDROC (Boltzmann-Enhanced Discrimination of ROC) score of 0.75 at the top 10%.These ndings, summarized in Table 1, provide compelling evidence for the e cacy of the bestperforming shape-based model, which exhibits satisfactory metrics and hence serves as a reliable lter in the virtual screening process.

Virtual Screening prioritizes 23 potential hits
With the models duly validated, the virtual screening was conducted applying the following lters (Fig. 4): 1. Utilization of the best molecular shape-based model for PvNMT, where the top 10% (131,982 compounds) of the list progressed to lter 2 for subsequent docking and selective rescoring between P. vivax and human isoforms.2. Examination of the top 10% of the list based on the MMGBSA-ΔG score (≤ -70 kcal/mol).
Compounds demonstrating a difference of ≤ -50 kcal/mol between HsNMT isoforms were singled out, resulting in the selection of 290 compounds.3. Application of P. falciparum 3D7 (mandatory) and W2 classi catory QSAR models to predict the activity of compounds selected in the previous step (80 compounds), given the importance of the target in asexual blood stages.4. Implementation of cluster analysis to identify representative chemotypes.Additionally, analysis of MMGBSA poses and evaluation of predicted ADMET characteristics were conducted.This comprehensive approach led to the nal selection of 23 compounds (see Supplementary le, Table 1).

Experimental evaluation NMT yeast-based system for identi cation of potential inhibitors
We used a modi ed yeast strains system expressing the NMT targets of both H. sapiens and P. vivax that allows a fast and cost-effective identi cation of potential selective inhibitors and cytotoxic compounds.
Cells were cultivated and subjected to tests at 100 µM in triplicate for each strain (yPvNMT, yHsNMT1 and ScNMT).Optical density (OD) was measured over a 72-hour period.In this assay, compound inhibition is determined by the reduction in growth rate compared to the control (180 µL of cells and 20 µL of YPD medium) in the presence of compound.As a result, seven compounds (Fig. 5) presented an inhibition phenotype, with some demonstrating a degree of selectivity against the strain expressing P. vivax NMT strain, as evidenced by a reduction in growth ratio (yield) compared to the human strain.Notably, LabMol-394 exhibited a vivax phenotype, indicating potential selectivity towards the parasite isoform.

Four compounds exhibited antimalarial activity
To evaluate the antimalarial activity of the compounds, we conducted tests against chloroquine-sensitive Pf3D7 and drug-resistant PfDd2 and PfSB1A6 strains, using a concentration of 5 µM, as detailed in Table 2. Initially, we determined the percentage of parasite growth inhibition and subsequently generated dose-response (EC 50 ) curves.Four compounds, LabMol-392, LabMol-393, LabMol-394, and LabMol-395, exhibited inhibition exceeding 80% against 3D7 strain parasites (see Supplementary le, Fig. S3 and S4), with EC 50 values ranging from 0.098 to 1.10 µM.Notably, all compounds tested exhibited low cytotoxicity, as evidenced by the selective index (the ratio of selectivity to the parasite over mammalian cells) higher than 10 (except for LabMol 393), aligning with guidelines for malaria drug development 32 .Furthermore, hemolysis tests were conducted at a maximum concentration of 20 µM indicated no hemolytic activity for any of the compounds tested (Supplementary le, Fig. S5).
The most promising compound identi ed in the phenotypic screening is LabMol 395, demonstrating potent antimalarial activity below 0.17 µM against all three P. falciparum strains tested herein.Moreover, it exhibits remarkable selectivity for the parasite over mammalian cells (SI > 10), indicating signi cant potential for advancement in pre-clinical studies.

NMT enzymatic assays revealed potential inhibitors
We conducted a two-point NMT inhibition assay of the top 23 potential inhibitors.Activity percentages were determined at 2 µM and 20 µM to identify a potential activity of compounds.Unfortunately, IC 50 calculations were not feasible due to compound solubility limitations.Nevertheless, the two-point enzymatic screen identi ed four promising compounds -LabMol-391, LabMol-392, LabMol-393 and LabMol-394 -which exhibited a minimum of 52% decrease in enzyme activity at a concentration of 20 µM against PvNMT (See Fig. 6).Regrettably, dose-response curves for the identi ed hits were hindered by compound precipitation at concentrations exceeding 40 µM.Subsequently, all compounds showing decreased activity were also tested against the human enzyme (HsNMT) in dose response studies ranging from 40 to 0.160 µM.Importantly, none of the compounds demonstrated inhibition of the human orthologue (HsNMT > 40 µM, see Supplementary le, gures S6 and S7).These ndings highlight four hit compounds that selectively inhibit PvNMT over HsNMT, thereby validating the e cacy of our computational selectivity approach.

Molecular dynamics simulations to analyze the binding mode of LabMol-394
Molecular dynamics simulations of 300 nanoseconds were conducted to evaluate the stability and binding dynamics of LabMol-394 with PvNMT.During this simulation, several signi cant interactions were observed.Although LabMol-394 (Fig. 7) does not directly engage with a critical residue, Leu410, it consistently maintains π-stacking interactions with residues Tyr211 and Phe105, occurring approximately 51% and 63% of the time, respectively.Intriguingly, a persistent hydrogen bond with Ser387 and the hydroxyl group was observed for approximately 34% of the simulation duration, while a π-cation interaction with Tyr211 and nitrogen of the pyridine ring was maintained for approximately 51% of the time.

Discussion
The process of discovering compounds with biological activity is a signi cant challenge, demanding the application of robust and predictive methodologies to ensure reliable outcomes.Bioinformatics and cheminformatics analyses serve as foundational tools in the search for novel antimalarial agents.Their effectiveness is evident through validation and research efforts led by esteemed organizations such as MalDA, MMV, the Bill and Melinda Gates Foundation, and the pharmaceutical industry 33 .
A crucial aspect of antimalarial drug discovery involves the identi cation of novel targets essential to the parasite's life cycle and conserved among various Plasmodium species.This strategy aims to encompass species previously overlooked as harmful to humans 34 .Consequently, the integration of computational strategies plays a crucial role in the discovery process, enabling the exploration of innovative mechanisms of action for potential new antimalarials.
The NMT enzyme was validated as a target for antimalarial drugs in 2014 30 with many inhibitors initially repurposed from other organisms 21,[35][36][37] and through High Throughput Screening (HTS) campaigns 38,39 .The latest NMT inhibitors, highlighted by Rodríguez-Hernández and colleagues, particularly compounds 12b and 30a from the series of hybrids of DDD85646 30 and IMP-1002 40 PvNMT inhibitors studies, demonstrated signi cant potency in vitro against PvNMT, with IC 50 of 0.0368 µM and 0.089 µM, respectively.However, their e cacy against hypnozoites and schizonts was observed to be in the low micromolar range 31 .
In our study, we initially conducted an analysis of structural conformations within a collection of crystal structures containing the new PvNMT inhibitor.This analysis allowed us to discern the various binding modes employed by the inhibitors.In parallel, a gathered set of compounds from the literature that had been tested against PvNMT, rigorously validating our shape-based and docking computational models, in accordance with established good practices 41,42 .Subsequently, a virtual screening campaign was conducted.Our approach involved utilizing a predicted ΔG difference between Plasmodium and human proteins as a lter to reevaluate the docking scores.This strategy enabled us to prioritize compounds that showed potential selectivity for the parasite.As a result, we identi ed and prioritized 23 compounds with promising characteristics for further investigation.
During the experimental validation phase, seven compounds exhibited a phenotypic effect against the modi ed yeast NMT strains.Next, the compounds were assessed for their e cacy against P. falciparum strains, considering its signi cant identity (> 80%) of NMT between the species 43 .Notably, in P. falciparum strains assays, four of these compounds demonstrated signi cant activity against three parasite strains, all at a single concentration of 5 µM, without any indications of cross-resistance and no signs of cytotoxicity (< 50 µM) or hemolysis, which are indicative of a favorable chemical safety pro le.It's important to consider that the observed micromolar activity against the parasites may be in uenced by the compounds' permeability to erythrocytic cells.Moreover, various e ux pumps, such as mdr1 and mrp1 in Plasmodium, known for their role in resistance mechanisms [44][45][46] , could affect the activity of these compounds.In yeast cells, the primary e ux pump for drugs, the gene PDR5, was deleted in the strains we used 47 .This deletion leads to an increase in the concentration of compounds inside the cells, potentially explaining the divergent results observed for some compounds.One standout compound from these assays, LabMol-395, displayed an antimalarial e cacy below 0.17 µM against P. falciparum strains examined.However, this particular compound did not exhibit a yeast phenotype against NMT, indicating that its antimalarial activity likely involves a different target.
On enzymatic NMT assay, LabMol-394 exhibited modest activity, resulting in a 52% decrease in the activity of PvNMT at a concentration of 20 µM, displaying selectivity towards vivax over the human isoform.Molecular dynamics simulations revealed that LabMol-394 demonstrated interactions comparable to PvNMT inhibitors, including those used in the shape-based modeling.However, it lacked interaction with Ser319, previously suggested as a crucial residue in the proposed mechanism of NMT inhibition 30 .Throughout our simulations, LabMol-394 did not engage with a crucial residue, Leu410.However, it sustained signi cant π-stacking interactions with Tyr211 and Phe105 residues, while also forming a hydrogen bond with Ser387.Although Ser387 is not currently associated with inhibition in the literature, we hypothesize it as a potential mechanism.
Chemically, LabMol-394 features a quinoline scaffold extensively documented for its diverse biological activities 48 .This scaffold is particularly renowned for its well-established antimalarial effects, primarily attributed to its inhibition of β-hematin and the formation of an irreversible complex with the heme group 49 .These actions disrupt the development of the parasite in both liver and red blood cells.Notably, recent reports unveiled a novel quinoline inhibitor targeting the translation elongation factor 2 (PfEF2), demonstrating multistage antimalarial properties and currently progressing through clinical stages [50][51][52][53] .This discovery underscores the signi cance of heterocycles as crucial sources of chemical activities and introduces new mechanisms of inhibition aimed at pivotal targets in the Plasmodium life cycle.
The prioritized candidates in this study were experimentally validated and the results demonstrated a correlation between antimalarial activity and the absence of cytotoxicity.LabMol-395 emerged as promising candidate for target identi cation, due to its favorable antimalarial pro le.Furthermore, enzymatic assays revealed LabMol-394 as the most promising PvNMT inhibitor, exhibiting selectivity against the parasite.Remarkably, there was a good correlation between the results obtained from the yeast system and enzymatic assays in NMT.This study makes a signi cant contribution to the exploration of the quinazoline scaffold, showcasing activity against Plasmodium vivax NMT and offering prospects for further optimization and development.

Structural analysis
For structural analysis, we utilized the Bio3D R package 54,55 , focusing on crystals with assigned ligand, no mutations and showed high resolution (≤ 2 Å).This analysis comprised the alignment of these structures and then principal component analysis (PCA) of the conformations was performed.

Data collection and preparation
To distinguish active compounds from inactive ones, we constructed and validated shape-based models.
Initially, compounds tested against the enzymes of both P. falciparum and P. vivax were collected from the literature 24,35,[56][57][58][59] .Subsequently, employing the protocol outlined by Fourches and colleagues 60,61 , we processed an initial set of 152 compounds.Brie y, the hydrogens were explicitly included while excluding counter ions, inorganic salts, polymers, mixtures, and organometallic compounds.We set a threshold for inhibitor activity considering compounds with Ki or IC 50 values ≤ 1 µM as active and those with values ≥ 1 µM, as inactives.Additionally, speci c chemotypes, such as aromatic and nitro groups, were standardized and duplicates were analyzed following the criteria: (i) entries with identical reported outcomes were consolidated, keeping only one, while the redundant entry was eliminated, and (ii) if duplicates exhibited inconsistent biological activity, both entries were removed from the dataset.Subsequently, the nal dataset includes 137 unique compounds, comprising 67 inhibitors and 70 noninhibitors.To augment the chemical diversity of the dataset, we generated 36 decoys for each inhibitor.
The dataset, including actives, inactives and decoys, underwent appropriated hydrogen protonation state on neutral pH (7.4) using the Open Babel program 62 .Subsequently, for each molecule, 200 conformers were generated using the OMEGA 63 program, and AM1BCC charges 64,65 were estimated using the QUACPAC software 66 .Validated shape-based models were then built using the ligand conformations extracted from the Protein Data Bank (PDB) and employed as queries at the ROCS v. 3.4.2.1 program 67 .

Shape-based model development and validation
To assess the predictive performance of the shape-based models, the following metrics were analyzed: Receiver Operating Characteristic (ROC) curve, which provides a graphical representation of the true positive rate (sensitivity) against the false positive rate (1-speci city); Area Under the ROC Curve (AUC), which quanti es the probability of an inhibitor being ranked higher than an inactive compound when compared to a random selection by the query 41,68 ; Boltzmann-Enhanced Discrimination of ROC (BEDROC), which uses an exponential decay function to assign weights to actives that ranked higher in the list 69 ; Enrichment Factor (EF), that evaluates the fraction of actives found at the Top n% of the ranking compared to a random selection 70,71 .The statistical metrics were calculated using the following equations: Selective bias docking and post-processing with MMGBSA The protein structures of Plasmodium (PDB ID: 2YNE) 30 and H. sapiens (PDB ID: 5O6J) 72 were processed in the Protein Preparation Wizard module 73 within Maestro (Schrödinger 2021-4, http://www.schrodinger.com).This protocol includes addition of hydrogen atoms, calculation of ionization states at physiological pH (7.4 ± 0.5) using the Epik program 74,75 , and removal of structural waters with a distance greater than 3 Å from the side chains.Subsequently, protein hydrogen bond assignments and protonation states were re ned using PROPKA, followed by restrained minimization using the OPLS-2005 force eld 76,77 .Ligand preparation was carried out in the LigPrep module in Maestro (Schrödinger 2021-4, http://www.schrodinger.com),with ionization states set at physiological pH (7.4 ± 0.5) using the Epik program.Conformers were generated, retaining the stereoisomerism, and the geometry was minimized using the OPLS-2005 force eld.Three-dimensional coordinates of the grid were constructed using the Receptor Grid Generation (Schrödinger 2021-4, http://www.schrodinger.com)module at the region of the peptide binding site reported in the literature for P. vivax 30 (grid coordinates x = 24,29, y = 43,20, z = 64,66) and H. sapiens 72 (grid coordinates x = -18,31, y = 4,55, z = -19,80), have a size of 10 Å. Docking validation was performed with the Glide program 78 (Schrödinger 2021-4, http://www.schrodinger.com)using the PvNMT dataset prepared and protein structure of PvNMT.Standard precision (SP) was utilized, employing a score function that uses an exhaustive sampling search, recommended for virtual screenings campaigns 79 .Statistical metrics including AUC, EF and BEDROC were computed using an in-house work ow in KNIME 80 to assess the robustness of the docking program.The MMGBSA rescoring method was performed in the Prime module version 3.0 (Schrödinger 2021-4, http://www.schrodinger.com).This module computes the energy of the ligand-protein complex in solvent from the initial docking poses.Rescoring was applied during protocol validation and during virtual screening for the Top 10% of the ranked list.The system was composed of implicit solvent of the VSGB2 model 81 , parametrized on OPLS-2005 force eld.The hierarchical sampling method was employed, with active site residues kept rigid in the rst cycle and site conformations explored in the second cycle until nding the lowest ΔG value expressed in Kcal/mol 82,83 .

Virtual Screening
Following the validation of shape-based models and docking, augmented by MMGBSA rescoring, the virtual screenings were carried out using the Core library and Express-Pick collection stock libraries from the ChemBridge database ( https://www.chembridge.com/screening_libraries/)and Life Chemicals Antimalarial Screening Libraries (https://lifechemicals.com/screening-libraries/targeted-and-focusedscreening-libraries/antimalarial-screening-libraries).These combined libraries encompass over 1.5 million compounds in total.These data were compiled and processed following the mentioned protocol for ligand preparation.The rst lter was the best shape-based model for PvNMT, where the Top 10% of the list went to the selective docking and rescoring lter between Plasmodium and human NMT.The Top 10% analyzed with the MMGBSA-ΔG threshold ≤ -70 Kcal/mol for PvNMT and presents a difference of ≤ -50 Kcal/mol against HsNMT.The compounds selected in the previous step were predicted by the 3D7 and W2 malaria machine learning models(described in 84 ), considering that the target is essential during the asexual blood stage.For the nal selection process, cluster analysis was performed using DataWarrior software 85 to identify representative chemotypes and avoid redundancy by excluding very similar compounds.Additionally, a medicinal chemistry visual inspection of the MMGBSA pose was performed, and predictions of ADMET (Absorption, Availability, Metabolism, Excretion and Toxicology) were made using the SWISS-ADME server (http://www.swissadme.ch/) 86.Finally, the prioritized candidates (See Supplementary Data le 1) were purchased and validated through in vitro assays using yeast-modi ed strains expressing P. vivax and human NMT, as well as P. falciparum 3D7, Dd2 and SB1A6 strains, cytotoxicity and enzymatic assays to comprehensively evaluate the compounds' e cacy and safety pro les.

Molecular Dynamics simulations
The simulations were performed in the Desmond program version 6.9 87 on the Maestro platform (Schrödinger 2021-4, http://www.schrodinger.com).The system for the 300 nanoseconds simulation was built in the System Builder module and used the explicit solvent (H2O) model of the TIP3P type 88,89 , the periodic boundary conditions of the orthorhombic type with water buffer at 10 Å around the protein.Ion's sodium (Na + ) and chlorine (Cl − ) were added to neutralize the system, whose nal concentration was NaCl at 150 mM.The system was minimized in a cubic box using the OPLS4 force eld 90 and the isothermalisobaric NPT ensemble was used, in which the number of particles, pressure and temperature in the system were constant.To adjust the temperature, the Nose-Hoover thermostat 91 at 300 K, and the Martyna-Tobias-Klein barostato 92 at 1.01325 pressure was used.Data processing of Root Mean Square Deviations (RMSD), Root Mean Square Fluctuation (RMSF), and protein-ligand interactions were obtained in the Simulation Interactions Diagram module.
Trajectory cluster analysis was performed using the Trajectory clustering module on Maestro (Schrödinger 2021-4, http://www.schrodinger.com)to obtain the most representative conformations of the lowest energy of the most populated cluster.

Experimental
Construction of plasmids and yeast -modi ed strains By codon usage optimization for expression in Saccharomyces cerevisiae, the coding sequences of NMT from H. sapiens and P. vivax (here referred to as HsNMT and PvNMT, respectively) were synthesized, ampli ed by PCR, and subsequently, cloned into the Bam HI-Pst I sites of the yeast expression vector pCM188-URA3 (EUROSCARF), resulting in the generation of the following constructs: pCM188-URA3-HsNMT and pCM188-URA3-PvNMT.For more details on these constructs, please review the publication from Bilsland, E. and colleagues 93 .The DNA constructs, veri ed by sequencing, were used in this work for the yeast-based functional complementation system.S. cerevisiae strains and plasmids used are listed in Table 3. Phenotypic analysis using Bioscreen YPD media (20% glucose,2% peptone, and 1% yeast extract) was used for routine culturing of strains.
Frozen stocks on exponential growth phase were transferred to Falcon tubes with YPD media.Growth was maintained for an hour in a rotary shaker at 30°C at 200 rpm.Before the start of experiment, the medium was removed, and the pellet resuspended in new media to a nal OD 600 of 1.0.Subsequently the cultures were diluted to an OD 600 of 0.1 into 100-well honeycomb plates (Labsystems Oy) pre-aliquoted with the appropriate media and tests.Strains were cultivated for 3 days with the low shaking setting at 30°C with 20 minutes measurement intervals using a wide band lter (450 to 600 nm) using Bioscreen C (Labsystems Oy).At the end of the experiment, the data were treated and normalized in the PRECOG program 94 .Then an R script was used using a growth curve library, to compute the cell yield (K) values, which was the parameter that best ts the data 95 .The data was subsequently exported to the GraphPad Prism 9 program, and the growth of tests and control growth were analyzed by the non-linear logistic regression model.

Phenotypic antimalarial assays
The antimalarial activity assay was assessed on P. falciparum 3D7 (chloroquine-sensitive), Dd2 (chloroquine-resistant) and SB1 (atovaquone-resistant) strains 14 .The parasites were cultivated in RPMI 1640 medium (SIGMA-ALDRICH), supplemented with hypoxanthine 0.005%, glucose 0.2%, sodium bicarbonate 0.2%, O + red blood cells (RBCs) and A + human plasma 10%.The cultures were incubated at 37°C in a low oxygen environment (3% O2, 5% CO2, and 92% N2) as described by Trager et al. 96 Drug inhibition assays were performed as previously described 97 .Brie y, synchronizations with 5% D-Sorbitol solution were performed at 48-hour intervals before experiments to allow incubations with > 90% of the parasites in the ring stage.Assays were performed in a 96-well plate, with 0.5% parasitemia and 2% hematocrit, in the presence of 5 µM of compounds or the drug vehicle (DMSO), as a control.Artesunate was used as an antimalarial standard.After 72 hours of incubation, parasitemia was assessed by uorometry using SybrGreen uorescent dye 98 .The plates were read in a CLARIOStar plate reader (BGMtech) by uorescence at 490 nm excitation and 540 nm emission wavelengths.The growth inhibition values were expressed as percentages relative to the drug-free control, and EC 50 values were calculated by plotting Log dosing vs growth inhibition (expressed as percentages relative to the drug-free control) using GraphPad Prism 8.The experiments were carried out in three independent assays.

Hemolysis assay
The hemolysis assay was carried out according to Wang et al. (2010) 99 with modi cations.Suspensions of erythrocytes (2% hematocrit) were incubated with the compounds at 20 µM or the drug vehicle (DMSO), as a control, at 37°C, 5% CO 2 , for 4 hours.The reaction mixtures were centrifuged at 1000g for 5 minutes, and the absorbance of the supernatants was measured at 540 nm using a Biotek Synergy-HT spectrophotometer.The hemolytic rate was calculated in relation to the hemolysis of erythrocytes in 1% of Triton X100, which was taken as 100%, using GraphPad Prism 8.The experiments were carried out in triplicate, relative to two independent assays.

Cytotoxicity assays
The cytotoxicity was evaluated using the MTT (3-[4,5-dimethyl-thiazol-2-yl]-2,5-diphenyltetrazolium chloride) reduction assay for quanti cation of cellular reductase enzymes activity as an indirect measurement of cell viability using human hepatocarcinoma cell line (HepG-2) and broblast-like cell lines derived from monkey kidney tissue (COS-7 cells 100 ).Brie y, the cells were cultured in Dulbecco's Modi ed Eagle Medium supplemented with 10% heat-inactivated fetal bovine serum and 1% antibioticantimycotic solution (10,000 UI of penicillin, 10 mg of streptomycin in 0.9% sodium chloride; Sigma Chemical Co., Saint Louis, USA) at 37°C and 5% CO 2 .The assays were conducted in a 96-well plate at a density of 10 4 cells per well.The cells were then incubated with a serial dilution of the drugs, starting at 100 µM, along with non-treated controls, for 72 hours.After the drug treatment, MTT was added to the wells, and absorbance readings were obtained using a CLARIOStar plate reader (BGMtech) at a wavelength of 570 nm (OD570).Cellular viability was expressed as a percentage relative to the vehicletreated control.The CC 50 was calculated by plotting a log dose vs. viability curve using GraphPad Prism 8.The experiments were performed in two independent assays.

Enzymatic
To measure the activity of the puri ed PvNMT, an activity assay was adapted from Goncalves et al, 2012 101 .A fresh stock of 4x assay buffer was made consisting of 9.2 mM potassium phosphate (KH 2 PO 4 ), 69.7 mM sodium phosphate (Na 2 HPO 4 ), 2 mM EDTA, and 0.04% TritonX-100 at pH 7.0.Fresh working stock solutions were made by adding DMSO for a nal concentration of 1% or 5% and a concentration of 1x assay buffer.Another working stock solution was made by adding DMSO for a nal concentration of 10%, without assay buffer.The test compound solutions were made from the 20 mM stock.The test compounds were diluted in 10% DMSO to make a 10mM stock.From the 10 mM stock, the test compounds were further diluted into 20 µM and 2 µM stocks in 10% DMSO.A 10 µM stock of a control inhibitor, known as Control 02(DDD85646), was made, and then further diluted in 10% DMSO to make a 2 µM solution.10 µL of the test compounds were plated in triplicates in a black, closed bottom 96-well plate (Greiner Bio-One).For the negative and positive control, 10 µL of 10% DMSO was plated in quadruplicates.10 µL of the control inhibitor, Control 02, was plated in quadruplicates.The PvNMT protein was diluted in 1% DMSO containing 1x assay buffer for a nal concentration of 10 µM.The PvNMT protein was diluted further in 1% DMSO, and 50 µL was added to each well in the 96-well plate for a nal PvNMT concentration of 25 nM.For the negative control, 50 µL of 1% DMSO containing 1x assay buffer was added to the well instead of the PvNMT protein.The reaction was initiated by adding 50 µL of reaction substrate containing 10 µM PfARF (Gly-Leu-Tyr-Val-Ser-Arg-Leu-Phe-Asn-Arg-Leu-Phe-Gln-Lys-Lys-NH2 purchased from Innopep in San Diego, California), 10 µM Myr-CoA (myristoyl coenzyme A purchased from purchased from Med Chem 101 LLC in Plymouth Meeting, Pennsylvania), and 8 µM CPM (7-Diethylamino-3-(4'-Maleimidylphenyl)-4-Methylcoumarin purchased from Thermo Scienti c Life Technologies in Grand Island, New York) to each well.Fluorescent readings were immediately taken of the plate using Spectra M2 plate reader (Molecular Devices) with excitation at 385 nm and emission at 485 nm.Reading continuously were taken in one-minute intervals for 45 minutes.Background uorescence and noise were determined by replacing each constituent of the reaction individually with 1% DMSO containing 1x assay buffer, and values were deducted from experimental samples.The percent inhibition of each test compound was calculated using Prism (GraphPad Software, Inc).

Declarations Tables
Table 2 is available in the Supplementary Files section.

Figures
Figure 1 Overview Table2.docx of the study's rationale for the development of new inhibitors targeting PvNMT.The process encompasses initial data acquisition from the Protein Data Bank (PDB) and literature-derived inhibitors, followed by structural analysis, shape-based and docking model development, and validation.Subsequently, a virtual screening campaign is conducted to prioritize compounds for in vitro validation, utilizing NMT yeast-based assays, malaria phenotypical and cytotoxicity assays, and NMT enzymatic assays.The most promising candidates undergo molecular dynamics simulations to analyse their binding modes.

Figure 2 Principal
Figure 2

Figure 5 Growth
Figure 5

Table 1
Validation of shape-based models employing different queries and RefTverskyCombo score.

Table 3
Plasmids and strains used in this study.