Designing dual inhibitors against potential drug targets of Plasmodium falciparum -M17 Leucyl Aminopeptidase and Plasmepsins

Abstract Malaria is one of the major diseases of concern worldwide, especially in the African regions. According to a recent WHO report, 95% of deaths that occur due to malaria are in the African regions. Resistance to present antimalarial drugs is increasing rapidly and becoming a problem of concern. M17 Leucyl Aminopeptidase (PfM17LAP) and vacuolar Plasmepsins (PfPM) are two important enzymes involved in the haemoglobin degradation pathway of Plasmodium falciparum. PfM17LAP regulates the release of amino acids and PfPM mediates the conversion of haemoglobin proteins to oligopeptides. These enzymes thus play an essential role in the survival of malaria parasites inside the human body. In the present study, we used in-silico molecular docking, simulation and Molecular Mechanics Poisson-Boltzmann Surface Area (MMPBSA) studies to find potential dual inhibitors of PfPM and PfM17LAP using the ChEMBL antimalarial library. Absorption, distribution, metabolism, excretion and toxicity (ADMET) profiling of the top ten ranked molecules was done using the BIOVIA Discovery Studio. The present investigation revealed that the compound CHEMBL426945 is stable in the binding site of both PfPM and PfM17LAP. In this study, we have reported novel dual-inhibitors that may act better than the present antimalarial drugs. Communicated by Ramaswamy H. Sarma


Introduction
Malaria is caused by Plasmodium spp. and transmitted through the bite of the female Anopheles mosquito. Plasmodium falciparum is one of the most lethal apicomplexan parasites among the Plasmodium species. According to WHO statistics, in 2019, globally, 229 million cases of malaria and 409,000 deaths were reported. In 2020, 241 million malaria cases and 627,000 deaths were reported worldwide (World Health Organization, 2020). Deaths due to malaria are more in developing and under-developed countries due to poor sanitation facilities. Hence, African regions share 95% of the malaria cases and 96% of the overall deaths caused by malaria.
To treat malaria infection, Artemisinin is being used either alone or in combination with other drugs, which is known as Artemisinin combination therapy (ACT). ACT has many advantages over using Artemisinin alone; for example, it helps to overcome the resistance against other drugs, has high efficacy, fast action and less likelihood of resistance development (Siddiqui et al., 2021). In October 2021, WHO approved the use of the malaria vaccine RTS, S/AS01, for children in nations with a high prevalence of P. falciparum (WHO, 2021).
ACT therapies have consistently shown better results both in terms of efficacy and their control over malaria. But in recent years, resistance against ACT has also been reported from some areas of Western Cambodia (Leang et al., 2015) and failure of ACT has been increasing gradually (Mok et al., 2011). Artemisinin resistance is also emerging and spreading independently in mainland Southeast Asia, Pursat, the Western border of Thailand, Southern Myanmar and Vietnam (Kyaw et al., 2013;White et al., 2014). A single nucleotide polymorphism in the Kelch13 gene of the Plasmodium parasite has been reported to cause resistance related to Artemisinin (Noreen et al., 2021).
Many natural molecules are used to cure plenty of diseases, including malaria. Epoxyazadiradione, natural chemical molecules extracted from neem seed oil are known to possess activity against Plasmodium falciparum (Thillainayagam et al., 2019). In vitro and in silico studies have shown that pyrazoline and pyrazole derivatives can also act as effective antimalarial agents (Thillainayagam et al., 2020). Fluoromethyl ketones and Vinyl sulfones are reported as inhibitors of the Plasmodium falciparum haemoglobin pathway (Belete, 2020).
Plasmodium spp. utilise different proteases in different stages of their life cycle. Aminopeptidase enzymes remove N-terminal amino acids from short peptides with high specificity. Two aminopeptidases, namely P. falciparum M1 Alanyl Aminopeptidases (PfM1AAP) and P. falciparum M17 Leucyl Aminopeptidases (PfM17LAP) have an important role in the growth and development of the parasite inside red blood cells by regulating the release of amino acids that are required for the growth of the parasite (Ruggeri et al., 2015). Both enzymes are present in the cytosol of parasites and are responsible for the survival of parasites. They also provide nutrients for the growth and development of parasites. Gene targeting and in-vivo experimental results have shown that inactivation of PfM17LAP causes inhibition of growth and, subsequently, death of parasites (McGowan et al., 2009;Stack et al., 2007). Expression data have shown that PfM17LAP expresses in all stages of Plasmodium life cycle, namely intra-erythrocytic, merozoite, sporozoite, early ring, early schizont, late schizont, late trophozoite, oocyst and ring stages (Lasonder et al., 2016;Otto et al., 2010;Zangh� ı et al., 2018). The essentiality of PfM17LAP, together with its ubiquitous expression in all stages of the life cycle strongly indicates that PfM17LAP can be a potential drug target against malaria.
In the present project, we used in-silico tools to identify novel inhibitors of PfM17LAP which may be used to stop the growth of malarial parasites and may prove effective even in cases of ACT resistance. PfM17LAP, a protein consisting of 528 amino acids, has an active site composed of two functional parts. The first part acts as a regulatory site and contains Mg 2þ , which can be readily substituted with other divalent metal ions, for example, Zn 2þ , Co 2þ and Mn þ2 . The second part serves as a catalytic site and contains Zn 2þ ions (Maric et al., 2009;McGowan et al., 2010). Metal replacement studies have found that PfM17LAP retains catalytic activity in the absence of metal ions at the regulatory site, but the removal of metal ions from the catalytic site leads to irreversible inactivity (McGowan et al., 2010).
For this work, we used the X-ray crystal structure of PfM17LAP (PDB ID -3KR4). We found Zn 2þ at the catalytic site, which was coordinated by Lys374, Asp379, Asp399 and Glu461. Besides the metal ions, two inhibitors of PfM17LAP, namely Bestatin (BES) and Compound4 (Co4; phosphinic dipeptide analog) were also present in the 3KR4. BES is a natural inhibitor of leucine aminopeptidases isolated from the culture filtrates of the Streptomyces olivoreticuli (Wilkes & Prescott, 1985). It has antibiotic and antimalarial activities as well. In an earlier study, the inhibitory constant (Ki) of BES and Co4 for PfM17LAP were found to be 25 nM and 13 nM, respectively (McGowan et al., 2010).
Another vital enzyme of the haemoglobin degradation pathway of Plasmodium is Plasmepsins (PfPMs) (Nasamu et al., 2020). Overall, haemoglobin degradation is a threestep process. In the first step, PfPMs degrade haemoglobin into oligopeptides that, in turn, break down into dipeptides which finally break down into amino acids using aminopeptidases ( Figure 1). Plasmodium uses PfPMs for intra-erythrocytic haemoglobin degradation of the human host, which is its nutrient source (Kumar & Ghosh, 2007). There are 10 different types of PfPMs in Plasmodium spp. which are expressed at different stages of its life cycle. Through the amino acid sequence analysis, 10 different types of PfPM were found. PfPM I-IV are active in the digestive vacuole and have 50-70% sequence identity among each other (Nasamu et al., 2020). They play important roles in processes like degradation of protein, maturation of secretory proteins, invasion and egress (Liu et al., 2005). PfPM-V is active in effector export and shares 19-23% sequence identity with other PfPMs. PfPM VI-VIII are active during the transmission stage and they share 31-36% sequence identity among each other. PfPM IX-X are involved in egress and invasion and share 37% sequence identity between each other (Nasamu et al., 2020). PfPM-III is also known as Histidine Aspartic Protease (HAP), due to the presence of Histidine in its active site. Previous studies have shown that removal of any one of the PfPM I-IV doesn't affect the parasite (Nasamu et al., 2020). However, the removal of more than one PfPMs among the four has an inhibitory effect on the growth of parasites.
It has been shown that removal of any one of the PfPM I-IV doesn't affect the parasite (Nasamu et al., 2020). However, the removal of more than one PfPMs among the four has an inhibitory effect on the growth of parasites. A high degree of similarity among PfPM I-IV indicates that a single inhibitor can be used to block the function of all four proteins. Hence, considering the importance of PfPM I-IV in parasitic activity, we used them as a potential drug target along with PfM17LAP and screened the potential inhibitor using molecular docking techniques. Multiple sequence alignment of PfPMs has revealed that amino acid residues TYR17, VAL105, THR108, LEU191, LEU242, GLN275 and THR298 are integral to their functioning. These amino acid residues are conserved across malarial strains but not in humans (Freire, 2002;Valiente et al., 2008).
We used the ChEMBL antimalarial library for screening, which contains bioactivity information for every molecule. After screening, we selected the top ten molecules based on their docking scores for post-docking analysis and molecular dynamic (MD) simulations to assess their utility as an inhibitor using in-silico techniques.

Protein preparation
The crystal structure (2Å structural resolution) of the PfM17LAP protein was obtained from the Protein Data Bank (PDB ID -3KR4). We defined the active site residues in the structure of PfM17LAP based on the binding of co-crystallized ligand BES. Structural summaries from the PDBSum database revealed that residues ASP379, LYS386, ASP399, ASP459, GLU461 and GLY489 of PfM17LAP form hydrogen bonds with BES.
Retrieved protein structures were prepared and the lowest energy conformation of protein was obtained using the protein preparation wizard of BIOVIA Discovery Studio. Energy minimization in the loop regions was done using the CHARMm forcefield, incorporated in the BIOVIA Discovery studio.

Chemical library preparation
For the purpose of screening, the ChEMBL antimalarial database was used. ChEMBL is a manually curated collection of bioactive molecules that are probable drug candidates with drug-like properties (Gaulton et al., 2017). The information about each ChEMBL drug candidate was obtained from different assays reported in the literature. For molecular docking, the compounds were prepared using the ligand preparation modules of the BIOVIA Discovery Studio, and a maximum of 10 tautomers were generated for each compound. During library preparation, isomers and 3D coordinates were generated for each compound after fixing bad valencies. A total of 281,896 compounds were generated after library preparation.

Molecular docking studies
The compounds generated via library preparation were taken for molecular docking studies. The FAST ligand conformation method was used to generate ligand conformations. A total of 255 lower energy conformations were generated for every compound. The LibDock module of BIOVIA Discovery Studio was used for molecular docking. It is a flexible docking method based on a high-throughput molecular docking algorithm developed by Diller and Merz (2001). LibDock uses polar and apolar hotspots to generate protein site features. The top 100 lower energy conformers of each compound were saved after the docking process.

Post-docking analysis
Ten compounds that obtained maximum LibDock scores were selected for post-docking analysis. Protein-ligand interactions were calculated and visualised using Ligplotþ (Laskowski & Swindells, 2011). Binding energies between the protein and the ligand were calculated using X-SCORE (Wang et al., 2002) (Equation (1)). X-SCORE is a consensus scoring function in which binding energy is calculated using three empirical scoring functions, HPScore (hydrophobic pair score), HMScore (hydrophobic match score) and HSScore (hydrophobic surface score) (Equation (1)). In X-SCORE, hydrogen bonding, van der Waals forces, hydrophobic interactions and deformation penalties are calculated between the protein and ligand (Obiol-Pardo & Rubio-Martinez, 2007).

Molecular dynamic simulations
MD simulations of protein and protein-ligand complexes were performed using GROMACS version 2020.4 (Abraham et al., 2015) using the Gromos53a5 forcefield (Oostenbrink et al., 2004). For MD simulation, the protein was placed in the centre of the cubic box with a distance of 10 Å radius around the protein and solvated using the Extended Simple Point Charge (SPCE) water model. Ligand topologies were generated by the Automated force field Topology Builder (ATB) server (Malde et al., 2011). The Isothermal-isochoric and Isothermal-isobaric ensembles were run before the minimization step at 300K temperature and 1 bar pressure for 500 picoseconds (ps). About 50,000 minimization steps were done before the final production run. The Particle Mesh Ewald method (Darden et al., 1993) was used to determine the long-range electrostatic interactions and Lennard Jones potential along with Coulomb's charge was used for the short-range interactions, with the distance of 1.2 nm. A Berendsen thermostat was used for temperature coupling  (Bussi et al., 2007). The Linear Constraint Solver (LINCS) algorithm was employed to constrain the bonds with hydrogen atoms. After equilibration, each system was set up for a 100 ns long production run at 2 femtoseconds (fs) time step. Trajectory information was obtained by saving the atomic coordinates after every 10 ps. The atomic movement during MD simulation was analysed using the inbuilt GROMACS modules and visualised using the XMGRACE tool.

Binding free energy predictions
MMPBSA analysis was carried out to predict the binding free energies of the protein-ligand complexes. MMPBSA calculations were done with the help of g_mmpbsa for Gromacs (Kumari et al., 2014). The calculation was done by combining the ensembles obtained from the trajectories of protein-ligand simulated complexes. A total of 500 frames between 50 and 100 ns simulated trajectories with a time duration of 100 ps were used for the calculation of binding free energies. Calculation of binding energy was carried out in 3 steps, in the first step potential energy in vacuum was calculated, in the second step polar solvation energy was calculated and finally in the third step, non-polar solvation energy was calculated. Among the different models available for the calculation of non-polar solvation energies, we used the Solvent Accessible Surface Area (SASA) model. Polar  solvation energy was calculated implicitly using the Poisson-Boltzmann (PB) equation (Good, 2006). As our system was neutral, we used the linear PB equation to calculate polar solvation energy instead of the non-linear PB equation that can be used for highly charged systems. Finally, the binding energy contribution of each residue was calculated using g_mmpbsa that used the previously calculated energy terms, molecular mechanics potential energy (E MM ), polar and nonpolar both in bound and unbound forms and subsequently calculates contribution of residues in binding energy (Kumari et al., 2014).

ADME and toxicity predictions
Pharmacokinetic properties, namely absorption, distribution, metabolism and excretion (ADME) of potential drug molecules, are important properties to measure the movement of a drug in the body with respect to time. In the present study, ADME properties were predicted with the help of the ADME descriptor algorithm protocol in BIOVIA Discovery Studio. This algorithm can predict six properties, namely human intestinal absorption, aqueous solubility, Blood Brain Barrier penetration (BBB), plasma protein binding (PPB), binding to Cytochrome P450 2D6 (CYP2D6) and hepatotoxicity of the small molecules. Optimal levels of ADME properties were used according to Ponnan et al. (2013). It can also measure the drug-likeness properties of small chemical or biological molecules. An important parameter to assess during an animal trial is the toxicity of the drug. It is an estimate of the damage that may be caused by a potential drug molecule to the organ in humans or animals and poisonous reactions that might occur in the body after drug administration/intake. A good quality potential drug molecule should be less toxic. A potential drug molecule proceeds to clinical trials only after it shows less or no toxicity in animals.
In the present work, the toxicity of a drug is predicted using the toxicity descriptor algorithm TOPKAT (TOxicity Prediction by Komputer Assisted Technology) in the BIOVIA Discovery Studio.

Off-target binding
A potential drug molecule may show interactions with protein(s) that differ from the intended target. Unspecific off-target binding is the primary reason behind the failure of drugs during clinical trials. Therefore, assessment of the potential off-target binding capacity of lead compounds is a crucial step of the drug discovery process. Prediction of probable off-targets not only provides an opportunity to avoid the adverse effects along with amplification of specificity by modifying molecules, but it is also helpful in identifying new targets of existing drugs.
In this work, we used the SwissTargetPrediction web server (Daina et al., 2019) to predict the non-specific binding of the top ten aminopeptidase inhibitors identified in our study. SwissTargetPrediction has a library of bioactive molecules whose information of targets and interactions is fetched from the ChEMBL database. For any given bioactive molecule, it predicts probable targets by checking the similarity between query and database molecules. It examines the similarity between target molecules and other molecules which can act as probable off-targets through a similarity index called Tanimoto index. Lead molecules with Tanimoto index more than the threshold value (0.65) may be probable off-target binders (Daina et al., 2019). We also checked the off-target binding of lead molecules based on their binding site similarity using PatchSearch (Rey et al., 2019). PatchSearch prediction is based on local non-sequential search of similar regions, called patches, on the surface of other proteins.

High throughput virtual screening of ChEMBL malarial library-Molecular docking of PfM17LAP
As P. falciparum is becoming more and more lethal with resistance spreading worldwide, there is an urgent need to find novel molecules with high potency. The selected target PfM17LAP is known to play an important role in the growth of the parasite by regulating the release of amino acids.
On the binding site of BES, we docked the compounds of the ChEMBL antimalarial library. We selected compounds having top ten LibDock scores. Since a high LibDock score indicates strong binding of ligand to the target, therefore we predict these compounds may act as potential inhibitors and can inhibit protein activity by binding on the active site. LibDock score of the reference ligand i.e. BES was 130.07 while the top scoring ligand from the ChEMBL antimalarial database was CHEMBL369831 with a LibDock score of 230.38. Analysis of top binding pose obtained by molecular docking showed that ligand CHEMBL369831 binds in the binding pocket along with the metal ions Zn þ2 and Mg þ2 (Figure 2). The top 4 LibDock scoring compounds, CHEMBL369831, CHEMBL426945, CHEMBL224244 and CHEMBL176888, had 9, 6, 9 and 5 hydrogen bonds; 51, 114, 47 and 54 non-bonded interactions with their corresponding binding protein, respectively (Table 2 and Figure 4). Post docking analysis of X-SCORE showed that CHEMBL426945 has the maximum binding energy of À 11.56 kcal/mol among the top 10 compounds, followed by CHEMBL176888 which has binding energy of À 11.05 kcal/mol. X-SCORE binding energies showed that the binding affinity of top 10 scored compounds was higher than the reference compound BES which has a binding affinity of À 8.75 kcal/mol.
It has been reported that compounds containing benzodioxole rings in their structure have antimicrobial and antiparasitic activity (Kamau et al., 2011;Shahavar Sulthana et al., 2015). We observed the presence of two benzodioxole rings in its structure CHEMBL369831. Our analysis also revealed that the O26 atom, part of the benzodioxole ring, formed two hydrogen bonds with nitrogen atoms of residues ALA387 and ALA388. The second benzodioxole ring made hydrophobic interactions with GLY390 and SER391, whereas the O45 atom formed hydrogen bonds with LYS374, THR486 and ASP399 and coordinated with Zn þ2 . The O41 atom of the compound CHEMBL176888 formed hydrogen bonds with LYS374, THR486 and coordination with Zn þ2 and Mg þ2 besides forming coordination with O35. Among the top four hits, compound CHEMBL426945 had a different binding mode (Figure 3) due to the presence of a polyene chain. We feel that the compound CHEMBL426945 may have antibacterial and antiparasitic activities because polyene analogues are known to possess these activities (Buzzini et al., 2008;Yamasaki et al., 2011).
Interestingly, all hydrogen bonds formed by the compounds CHEMBL369831, CHEMBL426945 and CHEMBL176888 were with oxygen atoms. The O1 atom of the reference compound BES forms hydrogen bonds with ASP399, THR486 and coordination with Zn þ2 . Our results showed the presence of a free terminal oxygen atom that makes more favourable interactions with PfM17LAP. Residues LYS374, ASP399, GLU461, ARG463, THR486, LEU487 and GLY489 were the main residues in the binding pocket. Residues LEU487, THR486, ASP399, LYS386 and LYS374 were the most important interacting residues, forming hydrogen bonds with top ten ranked compounds. GLU461, ASP459, ASP399 and ASP379 were the most important residues making coordination with metal ions. Nonactive site interacting amino acids were further analysed using MD simulations and MMPBSA analysis. Overall, docking results showed that top scored compounds from the ChEMBL library have more affinity than the reference ligand BES.

High throughput virtual screening of ChEMBL malarial library-Molecular docking of PfPMs I-IV
PfM17LAP is upregulated in many stages of the malaria parasite life cycle. Therefore, inhibiting its activity at one specific stage may not be sufficient to give potent antimalarial activity. Hence, we have also used PfPMs, which are also well known to play an important role in haemoglobin degradation, invasion and egress, as a potential drug target. Analysis of experimental data from the ChEMBL database showed that the top ten hits obtained after screening against PfM17LAP have already been reported to possess inhibitory activity against PfPM I and II. This indicates that the top hit compounds may act as probable dual inhibitors. To confirm this, we performed molecular docking analysis of the top ten hits against the four types of PfPMs found in the digestive vacuole. LibDock scores of the top ten hits against PfPM I-IV range from 162 to 213 and X-SCOREs range from À 9.31 to À 11.92 kcal/mol (Table 3a). The results (Table 4) suggest that the inhibitory activities (Ki) of top hits against PfPM-I and PfPM-II range from 0.4 to 6.0 nM and 1.1 to 15 nM, respectively. Therefore, these top hits may act as dual inhibitors for these enzymes. We also noted that despite the absence of any significant similarity between PfM17LAP and PfPM-I either at sequence or structure level, the top ten hits are still able to inhibit these two enzymes efficiently along with other PfPMs.
Molecular docking results have indicated that PfPM-I and CHEMBL369831 can form 4 hydrogen bonds within a distance range of 2.80-3.18 Å and 40 non-bonded interactions. CHEMBL426945 formed 2 hydrogen bonds within a distance range of 2.85-2.90 Å and 66 non-bonded interactions with PfPM-I (Table 3b and Figure 5). With PfPM-II, CHEMBL369831 and CHEMBL426945 each formed 3 hydrogen bonds, 70 and 69 non-bonded interactions, respectively (Table 3b). PfPM-III and CHEMBL369831 can form 2 hydrogen bonds within a distance range of 2.82-3.11 Å and 63 non-bonded interactions. CHEMBL426945 formed 2 hydrogen bonds within a distance range of 2.63-2.98 Å and 62 non-bonded interactions with PfPM-III. With PfPM-IV, CHEMBL369831 formed 2 hydrogen bonds within a range of 2.80-2.94 Å and 70 non-bonded interactions. CHEMBL426945 formed 6 hydrogen bonds within a range of 2.66-3.30 Å and 69 non-bonded interactions with   (Table 3b). Considering all these results, the obtained top ten hits can be considered as potential dual inhibitors.

ADMET and mutagenesis analysis of top ten scoring compounds
ADME analysis (Table 5) showed that all ten top scoring drug compounds were predicted to have poor absorption except the reference ligand BES. Solubility level was good for the top eight ligands and extremely good for the remaining two ligands namely, CHEMBL1086195 and CHEMBL5801112. BBB penetration value was unknown for the top scoring ligands including reference ligand BES. As per TOPKAT predictions, all top ten scoring ligands are non-toxic to humans except CHEMBL191130. Furthermore, all ligands were non-mutagens.

Off target binding results
Off-target predictions (Table 6) by SwissTarget Prediction showed that only two compounds (CHEMBL191130 and CHEMBL372618) of ten compounds have off-target binding prediction score 1. The remaining eight compounds have offtarget binding prediction scores of less than 0.32. This indicated only two compounds might show off-target binding with human proteins. Among ten compounds, for eight compounds the off-target protein was predicted as Cathepsin and for the remaining two compounds, Delta opioid receptor and Renin were predicted as a probable off-target binder protein.
The activity data against human cathepsin D along with inhibition activity values are shown in Table 4. Though the data shows that CHEMBL191130 and CHEMBL372618 compounds have off-targets in humans (cathepsin D), but their activity (Ki) range is very high as compared to PfPM-I and PfPM-II.

Molecular dynamic simulation analysis
The stability of protein-ligand complexes was determined through the MD simulations. For this, we selected the top 3 poses along with reference molecule BES based on X-SCORE. Therefore, comparative stability analysis was done for the apoprotein and four protein-ligand complex systems namely Protein-BES, Protein-CHEMBL369831, Protein-CHEMBL426945 and Protein-CHEMBL176888. A total of 100 ns MD simulations were performed on each system and stability was determined by calculating Root Mean Square Deviation (RMSD). The average RMSD for backbone atoms of native protein was in the range of 3.59 ± 0.74 Å. The average RMSD for backbone atoms of compounds BES, CHEMBL369831, CHEMBL426945 and CHEMBL176888 bound to PfM17LAP was found to be in the range of 3.96 ± 0.51 Å, 7.16 ± 0.84 Å, 4.01 ± 0.78 Å and 8.56 ± 1.61 Å, respectively ( Figure 6). The average radius of gyration for these compounds was 25.77 ± 0.16 Å, 25.75 ± 0.22 Å, 26.00 ± 0.19 Å, 26.35 ± 0.30 Å and 26.20 ± 0.21 Å, respectively ( Figure 8). Secondary structure analysis revealed that there are no major changes in the structure throughout 100ns for protein as well as protein-ligand complexes. Hydrogen bond analysis revealed that compound BES formed a greater number of hydrogen bonds than CHEMBL369831, CHEMBL426945 and CHEMBL176888 and the average number of hydrogen bonds formed were 3.57 ± 0.82, 2.13 ± 1.13, 1.59 ± 1.04 and 1.97 ± 1.34, respectively. PfM17LAP has four loops in its structure, loop 1: 247-271, loop 2: 374-392, loop 3: 496-499 and loop 4: 538-553. Root Mean Square Fluctuations (RMSF) analysis showed that the protein-CHEMBL426945 system has overall minimum fluctuations, specifically the active site residues corresponding to loop 2 were found to be more stable than other systems and apoprotein (Figure 7). Overall results Table 5. ADME, toxicity and mutagenesis prediction values for reference and top ten scoring ligands. Predictions are done with the help of the ADME predictor and TOPKAT from BIOVIA Discovery Studio. ADMET descriptors and their rules/keys have been mentioned below the Absorption: human intestinal absorption -0: good; 1: moderate; 2: poor; 3: very poor. Solubility: aqueous solubility-0: extremely low; 1: very low, but possible; 2: low; 3: good; 4: very good; 5: extremely good. BBB: blood brain barrier penetration -0: very high penetrant; 1: high; 2: medium; 3: low; 4: undefined. CYP-2D6: cytochrome p450-2d6-0: non inhibitor; 1: inhibitor. PPB: plasma protein binding -0: weak absorption; 1: high absorption. Hepatotoxicity: 0-nontoxic to humans; 1: toxic to humans. Mutagenicity: lesser negative values indicate no mutagenicity; higher negative values indicate more mutagenicity. ADMET scale adopted from (Ponnan et al., 2013).  depict that top binding poses showed greater stability in the active site. Further binding energy of simulated complexes were determined with the help of MMPBSA studies. Hydrogen bond analysis showed that an average of 4 hydrogen bonds were present in each protein-ligand complex. The protein in complex with CHEMBL369831 showed 4 consistent hydrogen bonds (Figure 8) throughout the simulation; the stability of these hydrogen bonds was further validated by binding energy studies.
The number of rotatable bonds in a molecule may cause disturbances in the active site interaction of protein with ligand. Among the five simulated systems (apoprotein, Protein-BES, Protein-CHEMBL369831, Protein-CHEMBL426945 and Protein-CHEMBL176888) Protein-CHEMBL426945 and Protein-BES showed stable hydrogen bond interactions throughout the 100ns simulation run, and their RMSD is lower than the apoprotein. The remaining systems are not maintaining stable hydrogen bond interactions. Due to the presence of more rotatable bonds there is a constant change in hydrogen bond interactions during simulations, which ultimately causes changes in interaction residues of the protein, this may cause more deviation in the backbone of RMSD of the protein. VMD movie analysis of Protein-CHEMBL426945 and Protein-BES systems revealed that compound CHEMBL426945 was stable throughout the simulation while the reference compound BES was moving away from the active site. The overall MD simulation analysis revealed that compound CHEMBL426945 is stable in the active site of protein throughout a 100 ns simulation, and this compound may be the potential dual inhibitor for PfM17LAP as well as plasmepsins.

Binding free energy calculations
A total of four systems were considered for the calculation of binding free energies. Calculation of the free energies of apoprotein, unbound ligand and bound protein-ligand complex was done separately. As discussed in the methods section, potential energy in vacuum, polar and non-polar solvation energies were calculated in three steps. The contribution of each amino acid residue in binding was calculated, which helps to find out important residues in ligand binding. Molecular docking and binding energy calculation results showed that compound CHEMBL426945 has the highest binding energy of À 199.592 ± 29.959 kJ/mol. Compounds CHEMBL369831 and CHEMBL176888 showed binding energies of À 185.690 ± 23.286 kJ/mol and À 140.455 ± 22.939 kJ/ mol, respectively (Table 7).
Reference compound BES showed positive binding energies with dielectric constant in the range of 2-8. To confirm the MMPBSA results, we performed the simulation of the BES complex twice (100 ns) followed by MMPBSA studies of the two top scored poses. We did not observe any changes in binding energies after repeated simulation studies. Earlier studies using MD simulations on PfM1AAP have shown  dissociation of BES from the active site of the enzyme (Yang et al., 2018). To the best of our knowledge, no such reports are available for PfM17LAP as of now.
From comparative analysis, we observed that compound CHEMBL426945 showed higher binding energies as compared to other top hits. The binding energies of residues range between 10 and À 20 kcal/mol (Figure 9). Along with defined binding residues, residues present in the defined binding sphere (LibDock) also contributed to binding. In case of compound CHEMBL369831, the binding energy contribution of residues LYS374, ALA388, LUE487 and GLY489, which are a part of its binding site is À 1.35 kcal/mol, À 4.18 kcal/ mol, À 4.19 kcal/mol and À 1.39 kcal/mol, respectively. While in the case of compound CHEMB426945 for the residues LYS374, LYS386, THR486 and LEU487, which are a part of its binding site it is À 1.80 kcal/mol, À 2.78 kcal/mol, À 0.95 kcal/ mol and À 6.87 kcal/mol, respectively, and in compound CHEMBL176888 for the residues LYS374, LEU487 and GLY489, which are part of its binding site it is À 0.83 kcal/mol, À 0.88 kcal/mol and À 2.56 kcal/mol, respectively.

Conclusion
As resistance against the first line antimalarial drugs is increasing rapidly, there is an urgent need to find novel antimalarial molecules with greater potency. The aim of this study was to find novel compounds that can simultaneously inhibit multiple proteins (PfM17LAP and PfPM I-IV) of haemoglobin degradation pathway of Plasmodium and may also prove to be promising inhibitors in combination with Artemisinin or individually in cases of ACT resistance. We selected the haemoglobin degradation pathway because it is essential for the survival of Plasmodium within the erythrocytes. In the present study, antimalarial data from ChEMBL was used for screening. The results showed that  CHEMBL369831 and CHEMBL176888 showed greater affinity towards both enzymes with the highest LibDock and X-SCOREs. Furthermore, the stability of PfM17LAP in complex with CHEMBL39831 was studied using MD simulations. The results of the MD simulation showed that the obtained top hits are stable in the active site. Bioactivity data from the ChEMBL database showed that molecule CHEMBL369831 has a minimum reported Ki of 0.5 nM. Further binding energy of the selected top hits was calculated using MMPBSA studies. Results showed that CHEMBL426945 has higher binding energy compared to the other top hits, including the reference compound BES. Overall docking results, post-docking analysis, MD simulation and binding energy analysis showed that CHEMBL426945 might be the potential inhibitor for PfM17LAP. Molecular docking and post docking analysis revealed that molecules CHEMBL369831 and CHEMBL426945 showed higher LibDock and X-SCORE against PfPM I-IV. The experimental data from the ChEMBL database and inferences drawn from our present investigation, together suggest that the top ten hits seem to act as dual inhibitors of PfM17LAP and PfPM I-IV. Off-target binding results showed that these compounds are selective towards targeted malarial proteins. Furthermore, clinical studies are required to confirm this molecule as a potential antimalarial drug.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
We

Authors' contributions
The work was designed and conceptualised by GRD, MK and NS. The experiments were carried out by GRD and AJ. Data analysis was done by GRD, AJ, MK and NS. Resources were provided by MK and NS. The manuscript was prepared and written by GRD and AJ. All authors reviewed and edited the manuscript.