Atomic level and structural understanding of natural ligands inhibiting Helicobacter pylori peptide deformylase through ligand and receptor based screening, SIFT, molecular dynamics and DFT – a structural computational approach

Abstract Helicobacter pylori is a Gram-negative microaerophilic gastric pathogen, responsible for the cause of peptic ulcer around half of the global population. Although several antibiotics and combination therapies have been employed for H. pylori-related gastric ulcer and cancer regiments, identifying potent inhibitors for specific targets of this bacterium will help assessing better treatment periodicity and methods to eradicate H. pylori. Herein, 1,000,000 natural compounds were virtually screened against Helicobacter pylori Peptide deformylase (HpPDF). Pharmacophore hypotheses were created using ligand and receptor-based pharmacophore modeling of GLIDE. Stringent HTVS and IFD docking protocol of GLIDE predicted leads with stable intermolecular bonds and scores. Molecular dynamics simulation of HpPDF was carried out for 100 ns using GROMACS. Hits ZINC00225109 and ZINC44896875 came up with a glide score of −9.967 kcal/mol and −12.114 kcal/mol whereas; reference compound actinonin produced a glide score of −9.730 kcal/mol. Binding energy values of these hits revealed the involvement of significant Van der Waals and Coulomb forces and the deduction of lipophilic forces that portray the deep hydrophobic residues in the S1pocket of H. pylori. The DFT analysis established the electron density-based features of the molecules and observed that the results correlate with intermolecular docking interactions. Analysis of the MD trajectories revealed the crucial residues involved in HpPDF – ligand binding and the conformational changes in the receptor. We have identified and deciphered the crucial features necessary for the potent ligand binding at catalytic site of HpPDF. The resulting ZINC natural compound hits from the study could be further employed for potent drug development. Communicated by Ramaswamy H. Sarma


Introduction
As a result of the current global antibiotic resistance crisis, some microorganisms, especially gram-negative bacteria, namely Pseudomonas aeruginosa, Klebsiella pneumoniae, Staphylococcus maltophila, and Acinetobacter baumannii cause major treatment challenges (Wenzel & Edmond, 2000). In that series Helicobacter pylori (H. pylori) is another Gramnegative microaerophilic organism that is considered major threat to human health in developing countries. Besides its etiological role in developing and causing peptic ulcer disease, dyspepsia, gastric cancer and lymphoma, the pathogen has also been associated in iron deficiency anaemia and idiopathic thrombocytopenic purpura (Siddique et al., 2018). Helicobacter pylori is one of the major human pathogens for which resistance-related attributes established by the pathogen cause treatment failures, diagnostic difficulties and uncertainty in clinical interpretation of therapeutic outcomes and thereby constitutes a serious threat to human health (Tshibangu-Kabamba & Yamaoka, 2021). Globally half of the population is infected by this group 1 carcinogen which necessitates an advantageous regimen for the efficient treatment (IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, 1994; IARC Working Group on the Evaluation of Carcinogenic Risks to Humans, 2012). Even though antibiotic therapies function as a rampart for eradicating this bacterial infection, the emergence of multidrug resistance has hampered this approach (Hu et al., 2016).
Peptide deformylase (PDF) is an attractive bacterial protein that could be considered an excellent target by structural biologist for antibacterial drug discovery PDF functions as a metallo protease, that co-translationally catalyzes the removal of a formyl group (carried by initiator methionine) from the N-termini region of polypeptides biosynthesized in prokaryotes. This deformylation is considered the first step during the N-terminal Methionine excision pathway (Boularot et al., 2004). This step is crucial in prokaryotes for bacterial protein synthesis as it is essential for bacterial proliferation. Even though PDF homologue has been identified from human, no effect has been discovered on cytoplasmic protein synthesis (Cai et al., 2006;Syed, 2016). Human PDF (HsPDF) has been suggested as an evolutionary remnant as it lacks any functional role. Also mutation in HsPDF prevent its activity similar to prokaryotic PDF (Hernick & Fierke, 2010). The native enzyme possesses metal in the state of iron. It has also been isolated with various cations in its active site, including cobalt, nickel, and zinc, due to sensitivity (Thorarensen et al., 2001). First 3D structure to be identified for PDF was of Escherichia coli (E. coli) (Groche et al., 1998). Since, several structures have been identified. Although overall differences have been described and identified from Gram-positive and Gram-negative bacteria, the active site region is to be confined and conserved Kreusch et al., 2003). PDFs are classified into three types based on homology and sequence analogue data. This Include Type I in Gram Negative, Type II in Gram positive and newly identified Type III in archaeal, kinetoplastids, leishmanial and trypanosomal homologs (Jaiprakash et al., 2015). The 'def' gene of H. pylori encodes its PDF. Helicobacter pylori PDF (HpPDF) contains three a-helices, seven b-strands, and four 3 10 helices in its overall structure shares its similarity with other PDFs like E. coli, Thermotoga maritime, P. aeruginosa etc (Cai et al., 2006;Lin et al., 2010). The insight into the interaction of amino acid side chains of the active site, metal-binding site and catalytic site with several inhibitors provided a unique platform for designing substrate analogue inhibitors or antibiotics targeting PDF. Three major motifs are found in bacterial PDFs. This includes GXGXAAXQ, HEXXH, and EGCLS where X can be of any hydrophobic atoms. The active site residues His138 and His142 of the HEXXH motif, Cys96 of EGCLS motif and water molecule coordinate tetrahedral by cobalt ion at the active site region in HpPDF. Residues like GLY95, ILE45, Leu97, Glu139, Val171, His138 domains are present in the crucial drug/substrate binding site S1'pocket . A distinctive variation in structure of HpPDF has been reported when comparing with PDFs of various other organisms. This include the structural difference in the CD loop of HpPDF and the confirmation of C-terminal helix that contain the Leu152 and Ser153 residues, which is different from Type I and Type II PDFs (Cai et al., 2006;Jaiprakash et al., 2015). Binding of small organic molecules in these sites can alter or perturb the functions of target proteins by inhibiting or activating their normal functions and so they have been widely used to understand the molecular mechanisms (Stockwell, 2004). Studies on HpPDF S1 pocket and inhibitors binding on the target are limited. We performed a computational structural study with molecular and atomic level analyses to design ligand and receptor-based hypotheses models to conduct High throughput Virtual screening (HTVS) of 1,000,000 of natural molecules and further DFT and Molecular dynamics simulation to investigate the responsible features of hits and thereby validate the binding of natural molecules on S1 pocket of HpPDF with respective features.

Protein preparation and receptor grid generation
The X-ray crystallographic structure of HpPDF with co-crystal ligand Actinonin (PDB ID: 4E9B in PDB format with 1.9 Å resolution was retrieved from the protein data bank. The selected complex was prepared using the protein preparation wizard workflow of GLIDE (Schr€ odinger-2017) (Sastry et al., 2013). Crystallized free water molecules beyond 5 Å distance and RNA were removed; hydrogen and missing side chains were added; loops and atoms were filled and assigned respectively. Hydrogen bonds assignment tool was employed to optimize the hydrogen bond. Impref module optimized the position of hydrogen bonds and all were kept in place . Further using heavy atom convergence of 0.3 Å the protein was subsequently minimized using OPLS 2005 force field (Vanajothi et al., 2020). Prepared protein was subjected for grid generation using the grid generation tool. Bound actinonin of PDB ID: 4E9B was selected to make the grid.

Ligand preparation
ZINC15, the free database provides data of over 120 million commercially available compounds that are categorised into different subsets where molecules are represented in ready to dock models (Irwin et al., 2012). For the present study ZINC natural product subset database was employed. In mol format structural coordinates of 1,000,000 natural compounds were retrieved. Further all the ligands were prepared using Ligprep module, Schr€ odinger 2017 (Mandal & Das, 2015). Ligands ionized at neutral pH 7 with ionizer and activated the desalt option. Ligands were further minimized using OPLS 2005 force field and generated one low energy conformation for each ligand . Ionization states for the compounds were generated at pH of 7.0 ± 2.0 with Epik and tautomeric conditions (Kikiowo et al., 2020;Shelley et al., 2008). Chiral centres were selected to retain the original states and avoid stereoisomer generation in-turn.
Energy-based pharmacophore modelling positive and negative hypothesis E-pharmacophore modelling was performed with crystal structure of HpPDF bound with an inhibitor. To perform the model generation, structures with PDB ID: 4E9B and PDB ID: 2EW7 were refined and selected as positive and negative complex respectively from Protein Data Bank (PDB). The structure of a protein in complex with a substrate or an inhibitor is a must to perform Energy-based pharmacophore modelling (E-pharmacophore) . To generate the structure-based pharmacophore model, the 3D structure of HpPDF was initially prepared using the protein preparation wizard of Schrodinger (Ma et al., 2019). Before this, ligands were extracted and prepared using Ligprep module of GLIDE. Minimised and conjugated gradient outputs of ligands were taken for Glide XP docking. XP descriptors write option and compute RMSD was enabled on this step. The output file was taken for E-pharmacophore modelling. The best pose was chosen manually from the pose viewer file that contained ligand posses and interactions. The selected pose was given as input to predict the pharmacophore. The Glide scoring function computes protein-ligand energetic terms and generate the pharmacophore. The Hbond acceptor (HBA), H-bond donor (HBD), features were used to create pharmacophore sites. Further, the complex was manually analysed to choose the best features, sites were ranked based on energy. Most favourable sites were selected based on this and using PHASE module, the hypothesis was generated.

Ligand-based pharmacophore
For the generation of pharmacophore model, a set of 21 inhibitors of prokaryotic Peptide deformylase especially for the SI pocket with pIC50 values ranging from 4 to 10 were selected from literatures and BRENDA database. The compounds selected for the study were minimised initially using Ligprep module. MMFFs force fields using an implicit GB/SA solvent model generated confirmations of the minimised ligand (a maximum of 1000) structures were given as input. Conformers were generated with confGen algorithm. The PHASE module of Schrodinger was exploited for the development of quantitative pharmacophore model. Thresholds set to 9.00 and 5.00 as training and test when applied yielded 16 actives and 5 inactive ligands. List of 21 compounds with their 1C50 value, PIC50 value and comment on activity are listed on (Table 1).

Hypothesis validation and enrichment
The Discriminative ability of the PHASE generated hypothesis to separate and define active compounds from inactive compounds was evaluated using hypothesis validation and enrichment calculation method. For this step, enrichment factor (EF 1% ), receiver operating characteristic (ROC), Boltzmann-enhanced discrimination of ROC (BEDROC) (BEDROC, a ¼ 20), robust initial enhancement (RIE) and enrichment factor efficiency (EFF) were considered to analyse the performance of the generated e-pharmacophore hypotheses for its discriminating ability (Nain et al., 2020;Oliveira et al., 2018). Decoys are molecules that possess physicochemical properties that are similar to as of drug molecules viz: molecular weight, ring count, hydrogen bond (H-bond) acceptor, H-bond donor and AlogP but they lack the druglike properties. For enrichment calculation method, the decoy library was downloaded from Schr€ odinger suite. This library of 1000 drug-like molecules was seeded with the 20 selected known inhibitors that possess different mode of activities. Further to calculate Good Hit (GH) scores, this combination of actives seeded with decoy molecule database was used to screen the pharmacophore models generated Selective screening using receptor-based pharmacophore model hypothesis ZINC compounds were subjected to the ligand and database screening tool of the PHASE module in GLIDE (Schr€ odinger 2017). ZINC compounds were imported as input ligands to be screened by the pharmacophore models generated. PHASE generated Hypothesis and ZINC ligands were added as input. As the first step of screening, a positive hypothesis was generated for ligand screening to filter the ligands that match with pharmacophore features. The output hit file was analysed and hits were imported as input to be screened by negative pharmacophore hypothesis. Hits (compounds matching with negative hypothesis) were excluded and Positive hits were selected based on fitness score.
Selective screening using ligand-based pharmacophore model hypothesis A Validated Hypothesis was taken for compounds to be screened from ZINC database comprising 1,000,000 natural compounds. The Screening was performed by using PHASE module-Ligand and Database Screening. Ligand-based 3 D pharmacophore screening was employed for HpPDF inhibitors using PHASE screening (Schr€ odinger 2017) (Geppert et al., 2010;Khan et al., 2016). In the current study pharmacophore model was used to identify the lead structure compound from ZINC natural compound database. The ligands were filtered based on the highest fitness score from the output file.

GLIDE HTVS and validation
Based on the efficiency of selected ligands to interact with the binding site residues, virtual screen aims for screening potential leads against a target. This script includes a strict selective filtration including rigid and flexible docking analysis for prediction of binding analysis and efficiency (Esther et al., 2017;Jang et al., 2018). Protocol for virtual screen comprises three steps: high-throughput virtual screening (HTVS), standard precision (SP), and extra precision (XP). For the present study HTVS and XP docking analysis were applied. The screened molecules were ranked in ascending order based on G-scores and Glide energy. Virtual screening algorithm recognises favourable hydrophobic, hydrogen-   2-(5-bromo-1-cyclopropyl-2,2-dioxido-1,4-dihydro-3H-2,1,3-benzothiadiazin-3-yl)-Nhydroxyacetamide 7 69 (Apfel et al., 2000) 10.
where vdW is for Vander walls energy, Coul denotes columbic energy, Lipo indicates lipophilic term, Hbond defines the hydrogen bonding term, Metal for metal-binding term, BuryP defines penalty for buried polar groups, RotB is penalty for freezing rotatable bonds and Site denotes active site polar interactions. For the present work screened natural compounds from phase screen was virtually screened through HTVS and XP docking respectively. Before docking protocols, docking was validated using redocking. The reference ligand Actinonin was extracted from the corresponding PDB: 4E9B crystal complex. The Ligprep tool was used to prepare the ligands initially. HpPDF was prepared using Protein preparation wizard. Receptor grids were generated with the corresponding ligand-binding site for the respective structure. Further using Glide ligand docking method, prepared ligand was redocked to the binding sites selected, in order to scrutinize the lowest the binding energy of respective native ligand on the S1 pocket of PDFs.

ADME property analysis
Drug likeliness of the hits was analysed using Qikprop module which in-turn aids in rational drug search (Kalirajan et al., 2019). Properties were examined and calculated using the ADME (Absorption, Distribution, Metabolism, Excretion filtration criteria. This algorithm calculates the pharmaceutical properties, including human oral absorption, Molecular weight, H-bond acceptors, H-bond donors, QPlogS (Aqueous solubility), QPlogPo/W (octanol/water partition coefficient) and QPlogBB (brain-blood partition coefficient), QPlogS (Aqueous solubility), central nervous system activity, HERG K þ channel blockage (logHERG), log Ps (Rate of brain penetration), apparent Caco-2 (QPP Caco), MDCK cell permeability (QppMDCk), Lipinski rule of five violations and Jorgensen rule of three for the input ligand structures Raza et al., 2017). Further, using the canvas module of Schr€ odinger 2017 the physiochemical properties of actinonin and ZINC hits were analysed. The hits from virtual screening were incorporated as input molecules for the drug likeliness and physiochemical property analysis.

Induced fit docking (IFD)
The prime tool was employed for IFD, as the protocol is developed to accurately predict the binding modes and associated structural changes of a ligand in receptor. To carry out IFD we selected hits from XP docking and ADMET analysis. The protocol begins by docking a ligand with receptor. Then the tool uses a reduced van der Waals radii and an increased Coulomb-vdW cut-off in the view of generating a diverse ensemble of ligand poses. During docking it further remove the highly flexible side chains temporarily. Thereby creates a closest conformer to the shape and binding mode of ligand molecules. The energy of the resulting complexes, is ranked based on the Glide Score.
Energy of the resulting complexes is ranked on the basis of Glide Score.

Molecular density functional
Molecular density Functional of ligands was performed using jaguar module of (Schr€ odinger-2017). The electronic molecular properties of hits from both ligand and structure based screening were calculated using Becke3 parameter and Lee-Yang-Parr correlation (B3LYP) functional method using 6-31 G(d,p) basic set level. Calculations were performed with the use of PBF solvation model and water was selected as solvent. For current study Frontier Molecular Orbitals (FMOS) viz: Highest Occupied Molecular Orbital (HOMO) and the Lowest Unoccupied Molecular Orbital (LUMO) energies, Molecular Electrostatic Potential Surfaces (MEPs), Local Reactivity Descriptors were analysed and determined out (Bendjeddou et al., 2016).

Molecular dynamics
Molecular dynamics simulations of all the complexes were done using GROMACS (Groningen Machine for Chemical Simulations) package (v. 2016.3) (Oostenbrink et al., 2004). GROMOS96 53a6 force field was applied and placed complexes in a box containing SPC216 water molecules. After setting Periodic boundary conditions Chloride ions were added to neutralize the net charge of the system to zero. Particle mesh Ewald (PME) algorithm was employed with a cut off of 1.0 nm for evaluating long range electrostatic interactions. Energy minimization interaction with varlet cut-off scheme was used. To minimize entire atoms, steepest descent integrator algorithm was employed to subject the whole system for 50,000 steps. Modified Berendsen thermostat and Parrinello-Rahman barostat were used for temperature and pressure coupling respectively. The system equilibrated with NVT ensemble, followed by 1 ns NPT ensemble equilibration at 1 bar pressure (Barge et al., 2021;Chinnasamy et al., 2020). For Apo HpPDF, co-crystal ligand (PDB ID: 4E9B) and ZINC bound hit complexes a simulation of 100 ns was performed with a time step of 2 fs at 300k on the equilibrated system. Prepared complexes were simulated to investigate the stability of the docked ligands. With respect to the simulation time root mean square deviation (RMSD) and root mean square fluctuations (RMSF) fluctuations, (Radius of Gyration) ROG plot and H-bond formation were noted and analyzed (Mandal & Das, 2015).

Result and discussion
Helicobacter pylori PDF (HpPDF) contains major three binding regions S1, S2, and S3 pockets, where S1 and S3 make the cavities and S2 links S3 as a saddle. Among the binding cavities, inhibitors binding to S1 cavities are of significant study interest (Cai et al., 2006). This is because S1 pocket has the major conserved hydrophobic pocket and the binding site for methionine side chain which in turn is required to remove the formyl group from the nascent N-formyl methionine peptide. However, an inhibitor binding to both the metal and active residues at this S1 pocket has been reported with good PDF inhibition among both Gram positive and Gram negative organisms (Smith et al., 2003). Whereas in S2 and S3 pockets variation are reserved with residue and also these regions are solvent exposed (Jaiprakash et al., 2015). S2 pocket sequence of gram negative bacteria is reported to found on the strand from 92 to 100 and S3 region found to be least conserved. For example Leu 125 in E. coli has been reported to be included in the S3 region (Smith et al., 2003). Very little data has been reported regarding these pocket regions in H. pylori.
Multiple sequence analysis of Peptide deformylase from five Gram-negative bacteria viz: E. coli, K. pneumonia, V. cholerae, H. influenza and H. pylori revealed that pocket regions are conserved. Whereas, CD loop region from 68 to 72 was found to be least conserved and highly variable in H. pylori (Supplementary File 1, Figure S1). The selection of HpPDF inhibitor for this study was based on two parameters (a) a natural origin, (b) Possessing the characteristics of a potent inhibitor and lacking the characteristics of the least active inhibitor.
The ligand actinonin is a naturally occurring antibiotic which act as potent and transition analogue inhibitor of all PDF (Chen et al., 2000). The simplicity of structure attracts the compound's usage in drug synthesis. Yet another feature of this compound is the presence of hydroxamate group, which make them very potent inhibitors of metallo enzymes. Several crystallographic studies of PDI have revealed that majority of PDIs are hydroxamic acid derivatives and these small molecules coordinate with the active-site metal atom (Hackbarth et al., 2002). The hydroxamate group of actinonin act as the chelating agent to bind the Fe 2þ ion of the Staphylococcus aureus PDF enzyme, which is inevitable for its activity (Singh et al., 2010). Ni 2þ , Zn 2þ , and Co 2þ can replace Fe 2þ metal ion with diverse catalytic activity without diminishing activity (Guilloteau et al., 2002;Ragusa et al., 1998). However Actinonin cannot be used as curative for number of reasons: (i) It has been reported with a moderate in vivo activity as it is pumped out by bacterial efflux pumps, (ii) Not only PDFs, it also targets several metalloenzymes (Antczak et al., 2007), (iii) It has also been reported to induce apoptosis (Gruji c & Renko, 2002;Xu et al., 1998). Nevertheless, this compound serves as reference molecule of novel antidotes of PDF. This is because of its high potency achieved through slow tight binding, a time dependent inhibitory mechanism (Fieulaine et al., 2016). For these reasons HpPDF bound with actinonin (PDB ID: 4E9B) was employed as reference complex to prepare pharmacophore hypothesis and for all the steps of this study.
Receptor-based pharmacophore would be more reliable for finding potential and specific inhibitors for metaloproteases since specificity is crucial for these proteins. Inhibitors specific for PDFs were taken from BRENDA-Database and literature survey (Eda & Jinka, 2019). Among the list of inhibitors identified so far Actinonin was reported with good inhibition of HpPDF with an MIC of 0.17 nm (Jaiprakash et al., 2015) and the above-mentioned characteristics of actinonin were taken to generate positive pharmacophore hypothesis. On the other hand another hypothesis was generated that had the least affinity and least binding with active site residues of HpPDF and was named negative hypothesis. For this, complex with PDB ID: 2EW6 was taken as reference since it was bound with N-Caffeoyltyramine. The structure and moieties of this compound have attributed for its least interaction. Ring confirmations have a null effect with HpPDF S1 binding pocket residues, as the region is deprived of aromatic amino acids including phenylalanine, tryptophan and Tyrosine around 5 Å of ligand.

Receptor-based E-pharmacophore
Positive and negative pharmacophore modelled complexes were generated using highly active and inactive ligand bound complexes respectively. Initial modelling generated an eight featured hypothesis for positive hypothesis (A1, A2, A3, A4, A5, D6, D7, D8, and D9) where, A denotes H-bond acceptor groups, D for H-bond donor groups and H for hydrophobic group. This hypothesis was further managed to include the functional groups involved in interactions. That included acceptors A3, A4 and A5 and Donors D6, D8, D9 (Figure 1a). Negative hypothesis was generated with four features: A1, A2, R11 and R12 (Figure not included).
Ligand-based pharmacophore modelling PHASE module generated seven different hypotheses with 4 to 6 features, including AAADHH, AAAHH, AADHH, AADDHH, AAAHHH, AAHHH, and ADDHH. Based on two statistical orders of highest survival parameters: (i) survival and (ii) survival inactive score, we characterised the robustness of the hypotheses ( Table 2). The survival ranged from 4.479 to 5.020 for the generated hypotheses. Hypothesis AADHH (Figure 1b) was considered the best hypothesis as it possessed a highest survival score of 5.020. The respective pharmacophore features were one H bond acceptor A, One H-bond Donor and two hydrophobic groups H.

Hypotheses validation
Ligand-based hypothesis AADHH recognized 16.7%, 18%, 23.7%, 35%, 55% and receptor-based hypothesis AAADDD recognised 35%, 71.4%, 71.4%, 76.2%, 81% for 1%, 2%, 5%, 10%, 20% actives respectively of total hits which represent the effective identification of the actives from the decoy ones. The enrichment screening for HpPDF was considered successful since the EF score was 48.62% for receptor-based hypothesis and 22.62% for ligand-based hypothesis portraying that the generated pharmacophore model is suitably enough for discriminating the known actives from the inactive molecules and also appropriate for retrieving active inhibitors. The curves potray the efficiency of the generated pharmacophores in identifying the active molecules. Further the Reciever Operating Curve (ROC) area model from study showed that the created model was good enough in discriminating the true and false positive ligands from binding, where it filtered about 80% of true positive ligands from the  active and inactive decoy list by AAADDD (Figure 2(a(ii))) and 70% of true positive by AADHH ( Figure 2(b(ii))). Generally, the plausible Score value for the ROC curve to be framed as efficient and suitable must be greater than or equal to 0.7. If the score seems closer to 1.0, it indicates a better hypothesis.
Receptor-based AAADDD hypothesis for the current study shows an ROC score of 0.84 and ligand-based AAADHH show an ROC 0.63 confirming both hypotheses as efficient and suitable in their own way. Further, the ROC plot of AAADDD and AADHH revealed an area under the accumulation of 0.87 and 0.63, respectively, pointing to a suitable ROC score to separate actives from inactive. The area under the accumulation curves (AUACs), ranks the probability of finding the actives before the relative rank. Area under the accumulation curves (AUACs) ranks the probability of finding the actives before the relative rank. Plausible value for AUAC between 0 and 1 is generally considered efficient. BEDROC is another Hypothesis ranking method in Enrichment analysis where the metric identifies the actives with the concept of early recognition. For this, the metric was set to specific virtual screening set up of a ¼ 20 (Marondedze et al., 2020). The BEDROC score for the hypothesis AAADDD and AADHH is found to be 0.93 and 0.50 respectively. The Robust initial enhancement (RIE) metric is then employed with a view of identifying the performance and ranking the actives. Higher the positive value of RIE better is the screening performance. The value of RIE for the hypothesis AAADDD and AADHH is 12.95 and 5.13 respectively. This higher positive value of RIE shows that the hypothesis can screen the compounds with better performance. EFF is the enrichment factor efficiency that distinguishes the active molecules from the decoy molecules. The value of Eff for the hypotheses AAADDD and AADHH is 0.972 and 0.914 respectively at 1% of enrichment factors with respect to N% sample size, which proves that both the hypotheses have the best capability to distinguish the actives from inactives. Plausible Eff value ranges between 0 to 1 where value towards 1 denotes efficient discrimination of actives and a value towards 0 indicates equal proportional rate of recovery of active and decoy (Nayak et al., 2019).
Favourable Enrichment values of the hypotheses AAADDD and AAADHH elucidate their strong predictive potential in identifying the suitable active ligands with high performance. The validated pharmacophore hypotheses were then used for screening against ZINC databases.
Screening of compounds using receptor-based and ligand-based pharmacophore models ZINC database comprising 1,000,000 natural compound subset was subjected to be screened by both hypotheses. The screening was done using Ligand and Database Screening tool of PHASE, where hypothesis and prepared ligand files were added as input. During receptor-based screening compounds screened by positive hypothesis was further screened using negative hypothesis. Negative screening yielded 54677 compounds matching to negative pharmacophore features. These negatively yielded when excluded produced a hit file.
Ligand-based screening yielded a hit file with 600,000 of compounds. Further from the output file, natural compounds were filtered based on the fitness score that is >1.8. We got a maximum fitness of 2.27 for receptor-based screening and 2.23 for ligand-based screening. Score value below 1.8 found to be poorly matching with features. We further superimposed the hypothesis AAADDD with hit molecule ZINC00225109, to verify the alignment of features and it was found that pharmacophore features A3, A4, A5 mapped with oxygen atoms at the hydroxamic tail regions and the cyclohexane ring, Hydrogen donors D6, D8, D9 mapped with hydrogen atoms of methyl group at the hydroxamic tail region (Figure 3(a)).
Whereas AADHH when aligned with ZINC44896875 found that hydrogen bond acceptors A2 and A5 mapped with the secondary nitrogen and carbonyl group of the ligand and hydrophobic groups mapped with the methyl groups and Donor D was mapped secondary carbon atom of the cyclohexane ring (Figure 3(b)).

Validation of GLIDE HTVS
Native ligand actinonin extracted and minimised from crystal structure when redocked on the S1 binding pocket of HpPDF reproduced almost similar binding pose in the active site of HpPDF. Redocking determines the lowest energy generated by Glide XP docking score function versus co-crystal ligands binding pose. The RMSD was calculated as 1.21 Å for both co-crystal (blue) and redocked ligand (red) by superimposing the two complex files. The RMSD value indicates that docking programme was successful in producing the native pose of ligand. Figure 4 shows the superimposed alignment of complex redocked in green and co crystal ligand in blue colour.
To perform the Glide docking validation, known compounds from the dataset created were docked in the active site of HpPDF S1 pocket using Glide HTVS program. Hits on top 10 scoring of screened compounds were analysed. Actinonin was screened among the top 10 scoring compounds, where N-Trans-Caffeoyltyramine was not screened among the top 10 indicating the discriminating ability of Glide among actives and inactives. Glide score of 20 known compounds taken were listed on (Supplementary File 1, Table S1). All the known inhibitors were found to bind within S1 pocket region of HpPDF ( Figure 5). A cartoon depiction of HpPDF is attached as Figure 5(a). Surface images of pockets are illustrated on Figure 5(b, c).
The selected compounds matching to the respective hypothesis with fitness score >1.8 from both receptor and ligand-based screening were subjected to molecular docking studies using HTVS and XP docking studies with vsw tool, as docking results were used as a filter to select the final compounds which interact with active site residues and thereby to predict the binding orientations of the respective hit compounds.
Based on the docking scores we obtained for both screening, compounds with glide score > À10.5 kcal/mol from receptor-based hypothesis screening and compounds with a glide score > À7 kcal/mol from ligand screening method were sorted for next induced fit step, as to filter out the leads.

Screening with glide score (induced fit docking)
The selected compounds matching to the hypothesis with high survival score from screening were subjected to molecular docking studies using Induced Fit Docking studies with vsw tool, since docking results were used as a postdocking filter to select the final compounds which interact with active site residues and further to predict the binding orientations of the respective hit compounds. Docking score details of hits are tabulated in Tables 3 and 4. Leads were subjected for ADME and the compounds were taken for IFD. ZINC00225109 produced a highest IFD score of À12.114 kcal/ mol and ZINC44896875 produced a highest glide score of À9.967 kcal/mol from receptor-based and ligand-based screening procedures respectively.
During rational drug designing, Structural interaction Fingerprint analysis (SIFT) is considered as an effectual molecular filter to pick molecules with advantageous binding  modes and interaction patterns with active residues of receptor protein at the binding site (Deng et al., 2004). Residual features such as backbone, side chain, hydrophobic, aromatic, acceptor, donor, polar, and charged groups in the neighbourhood of the ligands binding site was analysed using the SIFT analysis tool. We have observed and tabulated the aminoacids involved in respective interactions on (Supplementary File 1, Tables S2-S4). ZINC00225109 found to interact with most of the residues in the active site. It formed backbone interaction with Ile45, Gly46, Leu47, Lys93, Glu94, Gly95, Cys96, Leu97 and side chain interaction with Ile45, Gly51, Tyr92, Cys96, Leu131, Val134, His138, Glu139. Metal cobalt involved in side chain. Ile45, Tyr92, Cys96, Leu131, Val134 found to produce hydrophobic interaction. Whereas Gly51, His138, Glu139 found to be polar in nature. Glu139 found with charged interaction. Gly51, Leu97 as acceptor in nature and Tyr92 found with aromatic interaction with ZINC00225109. SIFT analysis of ZINC44896875 revealed that Ile45 was involved in Backbone, side chain and Hydrophobic interaction as that of ZINC00225109. His 138 and Glu139 found to be polar. Glu139 acted as charged residue, Tyr92 as aromatic and Gly95 as donor. It was observed that both hits ZINC00225109 and ZINC44896875 produced similar residual interaction as that of reference molecule Actinonin.
Binding mode analysis of ZINC leads from receptorbased screening with S1 loop of PDF The docking of all selected lead compounds showed similar common intermolecular interactions with the S1'loop of HpPDF. The amino acids involved in H-bond and Hydrophobic interaction and corresponding atoms and bond length (in Å) inside S1 pocket with respective leads with their glide score have been tabulated on (Table 3). Among the leads ZINC00225109 displayed a covalent bond with cobalt metal at the catalytic site within a distance of 2.08 Å. Previous studies of selective PDF inhibitors reported that inhibitors binding at these catalytic and S1 loop sites could be helpful in identifying and designing substrate analogue inhibitors (Cai et al., 2006). In our study, we observed that ZINC00225109 does interact with the cobalt metal ion of PDF enzyme and with S1 loop residues. ZINC00225109 displayed strong hydrophobic interactions with the S1 loop residues like Ile45, Leu47, Ala48, Leu97, Cys96, Ala135, Val134, Leu131, Tyr92. These charged residue accommodate the compound in proper position on the active site pocket (Rampogu et al., 2018). Further compound ZINC00225109 has hydrogen bond interaction with Gly95, Gly46, Ile45, Leu97, Gln51 through the amine group (H bond donor) and carboxyl moieties. The cyclohexane ring of ZINC00225109 is believed to be involved in kinking of alpha helix through strong hydrophobic interaction with S1 loop binding residues. We observed similar interaction on native ligand docking on HpPDF. From the result, interestingly both the Actinonin and ZINC00225109 exhibit same binding affinity with similar residues. Actinonin and ZINC00225109 produced a glide core value of À9.730 kcal/mol and À12.114 kcal/mol respectively which portrays ZINC molecule's dominancy over Actinonin in binding. Actinonin makes hydrogen bonds with Gly95, Gly46, Ile45, Leu97, and Gln51 (Figure 6(a)). The hydroxamic moiety of ZINC00225109 produces specific interaction towards active site residues. Structural comparison of inhibitors shows that hydroxamic tail moieties bind in the same location but possess slightly different orientations. Further H1 and O1 atom of ZINC00225109 has made hydrogen bond with oxygen at carboxyl and hydrogen at amino terminal of Gly 95, O2 of ligand has entered into hydrogen bond formation with amine group of Ile45. O3 of ligand entered in a hydrogen bond formation with hydrogen at amino terminal of leu97 and H11 and O3 made hydrogen bond with Gly 51 in the side chain of PDF (Figure 6 (b)).
Binding mode analysis of ZINC leads from ligand-based pharmacophore with S1 loop of HpPDF The presence of pyrazine and nitro methyl propene (metal binding region) of lead compounds assist strong hydrophobic interactions with Ile45, Leu47, Leu97, Ala48, Cys96, Leu131, Tyr103 and Val134 which favours them to bind deeply in the S1' pocket of the HpPDF. Hit ZINC44896875 possess a covalent bond with cobalt metal at a distance of 2.49 Å. ZINC 44896875 shows strong hydrophobic interactions with the S1 loop residues like Ile45, Leu47, Ala48, Leu97, Cys96, Ala135, Val134, Leu131, Tyr103. Further compound ZINC44896875 has hydrogen bond interaction with Ile45, Leu97 and Gln51 through the amine group (H bond donor) and carboxyl moieties. ZINC44896875 produces a glide score value of À9.967. The O1 atom of ZINC44896875 has made hydrogen bond with H atom at side chain Gly 46, O1 of ligand also entered in a hydrogen bond formation with amine group from main chain of leu97 of PDF and N2 of ZINC44896875 has made hydrogen bond formation with amine group at main chain of Ile45 (Figure 7(c)). These H-bond and hydrophobic interactions are important as they promote the selective inhibition of PDF enzyme. The interactions and binding orientations of the leads sorted out from ligand-based screening during IFD with the S1' loop residues of HpPDF are tabulated in Table 4. H-bond and hydrophic interaction of lead Zinc compounds during docking procedures with HpPDF revealed that they are strong enough to bind with the S1 binding region active residues. Schematic 2 dimensional (2D) interactions of all the three ligands with active site residues of HpPDF were also plotted with Ligplot software. Ligplot representation is attached as supporting information (Supplementary File 1, Figure S2)

MMGBSA
The binding free energies values and their energy components were predicted for the ZINC00225109, ZINC44896875 and crystal ligand Actinonin with HpPDF using Prime MM-GBSA method. Results are tabulated on Table 5. Highest binding free energy is crucial parameter for determining the ligand selectivity. Contribution of components like Van der Waals (vdW), lipophilic (Lipo), Coulomb, Covalent and solvation was compared for ZINC00225109, ZINC44896875 and Actinonin with HpPDF to reveal crucial factors that contribute to ligand selectivity among the components, Van der Waals (vdW), coulomb, Polar solvation and covalent interactions displayed favourable interactions, whereas Lipophilic displayed non favourable interactions for HpPDF. DGbind is a concoction of various energies which in turn is the final output obtained from the above process. Polar and non-polar energies contribute separately for the generation of DG bind (kcal/mol). We examined the DG bind for ligands ZINC00225109, ZINC44896875 and the reference ligand Actinonin with receptor HpPDF. When determining the selectivity of a ligand, DvdW (Van der Waals interaction) contribution is considered as the main component. DvdW interactions were, À36.809 kcal/mol for ZINC00225109, À41.298 kcal/mol for ZINC44896875kcal/mol and À43.343 kcal/mol for   reference ligand respectively When considered, the Coulomb interaction (Charge-Charge interactions) between two small ions is stronger than the gravitational one. Despite it, coloumb interactions are considered the strongest forces working inside molecules (Israelachvili, 2011). From the study, it was clear that except ZINC44896875, the other two experiences a greater coloumbic interaction than weak Vander walls interaction. When comparing the binding energy of ligand hits, they show similar coloumbic interaction as that of reference ligand actinonin. Van der Waals force is also found to contribute to the conformational entropy as similar to coloumbic forces and can be due to hydrophobic residues at the binding region (Gao et al., 2012). Further ZINC00225109, ZINC44896875 and Actinonin showed moderate lipophilic scores for HpPDF because it has larger and deep S1 binding pockets with major hydrophobic residues. DG Lipo (kcal/mol) for ZINC00225109, ZINC44896875 is À12.108 kcal/mol and À14.729 kcal/mol and for Actinonin is À12.837 kcal/mol. This indicates that lipophilic contribution of ZINC hits is not crucial for PDF selectivity. Polar solvation energy revealed ZINC00225109, ZINC44896875 and actinonin with entropy of 21.130 kcal/ mol and 37.590 kcal/mol and 37.910 kcal/mol respectively. Positive polar solvation energy is considered as a good interaction in turn a negative value indicates an exothermic reaction. We observed that ranking of predicted binding free energies of hits and crystals were favourable and in good agreement with ligand screening and docking analysis. The contribution of conformational entropy of ZINC00225109 and reference ligand Actinonin were similar encompassing the relative and similar flexibility of these ligands inside the binding cavity of PDF of H. pylori. The results of MMGBSA indicate that the docking programs are accurate enough to score the hits.

ADMET property analysis
The ADMET properties for the lead five zinc ligands each from structure and ligand-based screening against HpPDF was predicted in-silico by utilizing qikprop module of Schr€ odinger suite and the results are tabulated in (Tables 6  and 7). Molecular weights of compounds are in the range of 184.196 to 236.224 for receptor-based and 166.179-245.283 for ligand-based screening. Evaluated numbers of hydrogen bonds donors (HBD) by the solute to water atoms in the fluid arrangement of complex were in the range of 3 to 6 and 0 to 4 respectively. ZINC44896875 screened through ligand-based pharmacophore hypothesis had 0 HBD. Vice versa evaluated number of hydrogen bonds acceptors acknowledged by the solute from water particles in the fluid arrangement of the complex are in the range of 5.750-10 and 2-6 for leads from both screening methods. The compound ZINC00225109 has the most elevated Lipophilicity QplogPo/ w value of 0.447 and compound ZINC44896875 had 1.112. Lipophilicity is another crucial physicochemical property of a drug molecule which determines its pharmacology efficacy and solubility towards lipid, absorption and penetration of membranes, solubility and binding of plasma proteins. Hit compound Zinc002255109 and ZINC44896875 possess a possibility to cross the blood-brain barrier QP log BB for brain/ blood of À1.330 and À0.408 (recommended value range is À3.0/1.2), Cell permeability (QPPCaco) of 225.070 and 130.211 respectively. Further Human Oral Absorption of both hits was 3 which is a plausible range. Physiochemical properties of hits analysed through canvas module is attached as Supplementary File 2. The results of in silico ADMET screening of compounds Actinonin ZINC00225109 and ZINC44896875 possessed best pharmacokinetic properties to be considered drug molecules.

DFT analysis
The density-functional theory is a successful computational modelling method to determine and calculate the electronic structure of atoms, molecules and solids using the fundamental laws of quantum mechanics (Fiolhais et al., 2003;Segall et al., 2002). It allows to interpretation and prediction the complex system's behaviour at an atomic scale (Parr,  (Chinnasamy et al., 2020). Positive electrostatic potential regions are depicted by a deepest blue colour indicate an excess positive charge while negative potential regions depicted in deepest red colour indicate an excess negative charge (Chinnasamy et al., 2015).

Molecular orbitals (HOMO and LUMO)
How molecules interact with other species is determined by Fronteir molecular orbital. HOMO, the outermost orbit acts as electron donors and LUMO innermost unoccupied orbit with free space to grab up electrons acts as electron acceptors (Gece, 2008). Hence HOMO and LUMO energies are directly related to ionization potential and electron affinity respectively. The difference between these orbital energies is called energy gap, which determines structures' kinetic stability and chemical reactivity (Bendjeddou et al., 2016). Smaller the energy gap more polarised the molecule is and considered as soft molecule (Padmaja et al., 2009). These molecules are associated with high chemical reactivity and low kinetic stability. Higher the energy gap, harder the molecules are and thereby signifying a state of higher excitation energy. HOMO represents regions that are most likely to be of nucleophilic attack and LUMO indicates sites most likely to be of electrophilic attack (Hussan et al., 2020). In Actinonin Homo (p donor) is delocalised over hydroxamic tail moieties with carbonyl, amine and hydroxyl group while LUMO (p acceptor) is delocalised all over the pyrrolidine ring except the tail region. Their corresponding energy value is À0.16777 eV and 0.00610 eV respectively. The energy gap (DE) has been calculated as 0.17387 eV (Figure 7(a)). In ZINC00225109 the HOMO is delocalised over cyclohexane ring and LUMO localised over hydroxamic tail moiety ( Figure  7(b)) and in ZINC44896875 the HOMO is localised all over the pyrazole ring and nitro methyl propene region (Figure 7 (c)). The HOMO and LUMO energy for ZINC00225109 was À0.021528 eV and À0.075464 eV. HOMO, LUMO energy for ZINC44896875 was identified as À0.241758 eV and À0.066838 eV ( Table 8). The (DE) has been calculated as 0.13982 eV and 0.17492 eV for both hits respectively (Table  8). The higher the negative HOMO value the more electrons stable to nuclei and the more negative the LUMO value is, the stronger the affinity for electrons are. When comparing the DE of hits with reference ligand actinonin both hits were successful in producing a similar DE as that of Actinonin.

Global reactive descriptors
From the Frontier molecular orbital energies (HOMO and LUMO), the Global reactive descriptors have been calculated on the basis of Koopmans's theorem. The global descriptive parameters like hardness (g), softness (S), Electronegativity (v), Chemical Potential (lÞ and Electrophilicity index (xÞ helps to identify the reactive nature of ligands (Sangeetha et al., 2017).
ChemicalPotentialðlÞ ¼ Àv; ElectrophilicityIndexðxÞ ¼ l2 2g When comparing the descriptive parameters of screened ZINC ligands with reference molecule Actinonin, the hardness and softness of reference compound PDB ID: 4E9B (Actinonin) and ZINC44896875 were similar, however compound ZINC00225109 had a softness of 10.3102 eV. Hardness and softness are inversely proportional to the reactive nature of the compound according to the Maximum Hardness Principle (MHP). As per the principle, the reactive nature of a compound decreases with an increase in the hardness and the stability increases at this time. But an increase in softness makes the compound highly reactive and less stable (G azquez, 2006;Hussan et al., 2020). In this study ZINC00225109 found to be softened and reactive compared to the reference compound actinonin and ligand-based pharmacophore screened hit ZINC44896875.
Another fundamental descriptor on chemical reactivity of atom is Ionization energy. Small ionization energy indicates high reactivity of atoms and molecules and vice versa. ZINC44896875 was calculated with comparatively smaller Ionization energy (0.021528 eV), when compared with other two. If the electronic chemical potential is lower, then the species is considered as stable forming stable complex in the receptor. Chemical potential of all the compounds were found to be low, however ZINC00225109 possessed a very low chemical potential of À0.026968 eV than ZINC44896875 (À0.154298 eV). The electrophilicity index measures the tendency of the species to accept electrons. A lower electrophilicity index indicates species are of good reactive nucleophile and higher value indicates a reactive and good electrophile. In this study when comparing the descriptors it could be inferred that ZINC00225109 is found to be more reactive with a lowest electrophilicity index (00079 eV), highest softness, lowest hardness, negative chemical potential values (Table 9). All the descriptor values calculated for hit molecules correlate with molecular docking.

MEP surface analysis
The molecular electrostatic potential (MEP) map provides data regarding the sensitivity of molecular sites toward electrophilic (electron deficient) and nucleophilic (electron rich) attacks (Ganesan et al., 2018). It usually displays the charge distribution of the molecules and there by ligand receptor interactions. Visual method with colour code help to understand the polarity of molecule (Sureshkumar et al., 2020). The potential value ranges in increasing order from red < blue colour. Highly potential electrophilic (negative) in red colour and nucleophilic (positive) in blue (Ranjith et al., 2017). In actinonin highly negative region are capped in red colour and include atoms like O4, O13, O2 and O42 at hydroxamic tail and centre region of ligand which in turn were found to interact with highly hydrophobic residues Leu97, Gln51, Gly95, Ile45 and metal Cobalt (CO) of HpPDF. The visualization in Figure 8(a) clearly indicates the importance of  carbonyl group at these positions as electron acceptors as LUMO is practically capped over this region. The vicinity of hydrogen atoms of amine and methyl group could be sensitive towards nucleophilic attack and hence capped as highly positive potential regions. For ZINC00225109 MEP was visualised as negative potential localised around O2, O11 and O15.
In which O11 and O15 found to be highly potential and found to interact with major residues Leu97, Gly95 and also bound with CO. Highly potential positive region found on NH groups. H 24 of N9 and H27 and 28 of N16 found to be highly positive potential. This result coordinate with docking results as these Hydrogen atoms haven found to interact with Gly51, Gly46 and Gly95 respectively (Figure 8(b)). In the case of ZINC44896875 the negative potential region has been capped around O atom at the nitro methyl propene region and amine group at 7th and 16th position. Where atom O has found to interact with hydrophobic residue Leu97 moreover covalent bond was made with the Co metal of HpPDF. Further confirms the success of screening process in keeping Co groups that interact to important residue Leu97 at the active site. Positive potential (blue) region was localised over all the hydrogen molecules (Figure 8(c)).

Molecular dynamics simulation
Molecular dynamics (MD) simulation was performed for a period of 100 ns on GROMACS software, on protein-ligand complex for evaluating the stability and conformational changes of selected inhibitor in the active regions of receptor at different time scale intervals. Simulation of protein-ligand complex with Actinonin and ZINC hits (ZINC00225109, ZINC44896875) was performed to study the conformational changes of receptor during simulation period and further the trajectories were analysed to scrutinise the position of ligand in the binding pocket and explicit protein dynamics over the period of time. Apo and crystal ligand complex were employed as control for comparing the simulation of protein-novel zinc hit complexes throughout the evaluation study by molecular dynamics

Molecular dynamics simulation trajectory analysis
Backbone RMSD, RMSF, H-Bond and ROG were analysed for all the simulated complexes, with respect to the simulation time of 100 ns. MD simulation was performed in the presence (complex) and absence (apo) of inhibitors thereby to check the system stability. Variations in RMSD and RMSF of protein back bone atoms are charted to scrutinize the protein dynamics. RMSD analysis showed initial fluctuations similar for all the four complexes till 40 ns. We observed a wobbling protein structure up to 40 ns. The Apo protein had a sudden deviation from 0.3Åto 0.35 Å unlike other three bound complexes. Apo protein showed a maximum deviation up to 0.35 Å. A steady fluctuation was maintained by both crystal and ZINC00225109 complex. But ZINC44896875 showed a higher fluctuation of back bone atoms than crystal complex towards the end of simulation period. Actinonin bound complex had a maximum fluctuation of 0.3 Å and minimum of 0.2 Å. ZINC00225109 exhibited deviations up to 25 ns and further showed a stable deviation and an overall RMSD below 0.26 all through the simulation time. Whereas ZINC44896875, as time progressed exhibited a major fluctuation at 86 ns and its RMSD value reached 0.36 Å. At the same time ZINC00225109 was found to be successful in reproducing the stability all through the simulation like crystal complex (Figure 9(a)). Residues Glu139, Leu 97 was found to have major role of interactions at the binding pockets with both hits. The lack of fluctuations in the crystal and ZINC complexes, and presence of fluctuations at regions of Apo protein clearly indicates the tight binding of ligand in the S1 pocket and stability of complexes. On the RMSF plot, the peak indicates residual protein areas that undergo more fluctuation. Based on the RMSF values, the residues involved are distinguished and stated as stable or fluctuating. Fluctuating regions are observed as loop regions and includes mostly N-Terminal and C-Terminal regions in a protein secondary structure. As these terminal fluctuations are common, we looked for uncommon fluctuation and stability by the residues. The results of RMSF analysis indicates that in apo protein and all bound complexes (Figure 9(b)) significant conformations were centred on the residues 60-70 and also towards the C terminal residues. When comparing ZINC4486785 and ZINC00225109, the latter seems steady and exhibited maximum conformational deviation of 0.3 Å. The fluctuations were similar to Apo protein.
We observed that Residues Val22, Glu66, Asp67, Gly68, Val69, Gln70, Met91, Tyr92, Lys93, Gly95, Phe120 had high fluctuation than more than average in Apo protein. Compared to Apo protein these residues found to be stable in bound complexes. Residues Asn 65 -Cys75 forms the CD loop region in HpPDF that has been found with distinctive structural difference compared to other PDFs and has been reported to be conformationally flexible (Cai et al., 2006). To support the binding efficiency of ligand we observed that residue Met 91 which is flexible in Apo protein is at binding pocket region of HpPDF and found to be stable in ZINC00225109 complex in contrast to all other which signifies the stable binding of this ligand with residue. When compared to Apo protein less confirmations were found at ligand binding regions including Ile45, Gln51, Gly46, Gly96, Tyr103, Glu139 which were involved in forming hydrogen bond confirmation with ZINC ligands. The fluctuation of these residues in corresponding region at receptor is less than 0.5 Å. Regions with high confirmations can be consider and correspond to the loops in surface exposed region of receptor. ZINC002255109 and Actinonin found to be similar in stable/fluctuation all through the simulation period. The binding affinity of the protein is based on the steady formation of H bonds during the whole scale of production MD run trajectories. The number of hydrogen bonds formed between the residues of protein and the compounds over the simulation time is crucial, so that, it could reinforce the activity of the agonists. HpPDF residues at site were forming H bonds with bound ligands and we discerned it to be conserved and properly maintained over the course of the simulation time. ZINC00225109 complex found to form a maximum of 6 and an average of 3 hydrogen bonds all over the simulation time (Figure 9(c)). At the same time ZINC44986785 only had a maximum of 2 and an average of 1 h bond. Radius of gyration for all the four complexes have analysed to measure the distance between backbone atoms and their axis of rotation. On Scrutinizing the Rg plot, we have deciphered that the Rg of the Apo HpPDF protein was 1.56-1.52 whereas Crystal ligand and bound complexes had an average Rg of 1.55 nm. Similar to RMSD Apo and ZINC44896875 displayed fall and rise in Rg. Actinonin (Crystal ligand) and ZINC00225109 displayed comparable deviations to other two (Figure 9(d)).
Simulation studies decipher that receptor-based pharmacophore screened compound was comparatively better in exhibiting stability over the simulation time than ligandbased. ZINC00225109 was found to be stable through the dynamics procedure and successfully reproducing the reference molecule interactions and stability on dynamic studies.

Discussion
This study discussed here is a part of an extensive study we have carried out on H. pylori receptors and natural inhibitors, including structural characterization of other potential targets. The study aims to understand the underlying features responsible for and design a competitive inhibitor for H. pylori PDF enzyme by scrutinizing the structural features responsible for their action. We have performed a computational analysis using ligand and receptor-based pharmacophore modelling, high throughput virtual screening, DFT and molecular dynamics analysis. Initially, 1,000,000 of natural compounds from ZINC database were retrieved. A total of 21 PDF antagonists reported in literature and BRENDA database are selected to perform ligand and receptor-based pharmacophore screening to understand structural characteristics responsible for their biological activity. Two types of pharmacophore hypotheses were created using ligand and receptorbased methods to understand the enrichment of hypotheses in screening the hits. Both hypotheses were validated individually to score them in discriminating the training sets. HTVS of the natural database screened with the HpPDF is performed to rank molecules based on their binding energy. Through the docking procedures binding pocket was characterised. We have identified that residues Ile45, Gln51, Gly46, Gly96, Leu97, Tyr103, Glu139 are important in the S1 pocket region. ADMET and MMGBSA analysis were performed to identify and characterise the drug likeness and energy involvement of hits. We observed that both hits were successful in having pharmacophore kinetic properties to be considered as drug molecule, which in turn predicted the non-reactivity of hits in accordance with reference compound actinonin. Further we identified that ZINC00225109 and reference compound actinonin exerted a greater coloumbic interaction than Van der Waals force. However due to the presence of hydrophobic residues lipophilic scores of hits were moderate. The contributions f conformational entropy of both ZINC ligand hits and crystal ligand Actinonin is the same encompassing the relative and similar flexibility of these ligands inside the binding cavity of HpPDF which further validate further and confirms that the docking programs are accurate enough to score the training ligands. The DFT analysis established the electron density-based features of the molecules and also we observed that the results correlate with docking intermolecular interactions. Molecular dynamics simulation of HpPDF was carried out for 100 ns using GROMACS where three different protein-ligand systems and an apoprotein system were simulated. Analyses like change in RMSD, RMSF, Radius of gyration and Hydrogen bond were scrutinised. On analysing the MD trajectories we have observed that hit screened from receptor-based hypothesis is stable than ligand-based screened hit. This could be attributed to the selection of features responsible for biological activity through pharmacophore screening.
Stability and hydrogen bond formation of ZINC00225109 was found to be better when compared with ZINC4489685 and Actinonin. We observed that residues Met 91, Leu97, Glu139 are responsible for stability of receptor HpPDF. This inference supports the study by Jianhua Cai et al., on X-ray crystallographic characterisation and validation of HpPDF as a potential target to inhibit H. pylori (Cai et al., 2006). From the current study we recommend ZINC00225109 can be taken for further analysis to be considered and developed as a potent molecule to inhibit H. pylori.

Conclusion
From this study we observed and identified the importance of active site residues and metal binding site interaction of novel molecules for designing potent drug with stable binding. We conclude that ligands with hydroxamic tails contribute for better and stable binding in the S1 loop residues rather than with ring structures. Presence of carbonyl and secondary nitrogen groups attribute for strong interactions and also regions with metal interactions attribute for ligands stable binding. Moreover, we observed that the size of compounds also attribute for its binding at hydrophobic regions. Smaller the size better is the binding at S1 pocket. We could found that residues Gly 95, Gly51, Ile45, Glu139, Leu97 important in making interactions. Simulation and DFT analysis supports this features importance in producing favourable and stable interactions at the binding region. Therefore these features could be considered important while designing a drug molecule with potential binding at S1 pocket of H. pylori. The insights into the comparative intermolecular interaction between the ZINC-HpPDF complexes against Actinonin -HpPDF reference compound complex deciphered that ZINC hits have occupied the essential active residues of HpPDF by aid of intermolecular forces and hence, can be, used in the development of inhibitor for HpPDF. We conclude that ZINC hits exhibited considerable affinity and substantial interactions with active site residues in the respective complexes and can be employed in developing of HpPDF inhibitors as supported by HTVS, ADMET, MMGBSA, DFT and Molecular simulation analyses.