Abstract
Helicobacter pylori remains a major cause of gastric diseases, while increasing antimicrobial resistance highlights the need for novel therapeutic targets. Here, we developed a comprehensive computational target-prioritization framework integrating Rosetta Stone domain-fusion analysis, subtractive genomics, metabolic pathway mapping, structural modeling, and molecular dynamics simulations to identify pathogen-specific drug targets in H. pylori J99. Twelve conserved domain families comprising 746 proteins were identified, and domain abundance showed a weak negative correlation with versatility (r = −0.327), suggesting functional specialization through domain expansion. Sequential filtering based on essentiality, nonhomology to the human proteome and gut microbiota, subcellular localization, and druggability prioritized five candidate proteins, with the ATP-binding protein AAD06727.1 emerging as the most promising target. AlphaFold2 modeling produced a high-confidence structure containing a conserved nucleotide-binding pocket. Molecular docking identified 5′-fluorosulfonylbenzoyl-5′-adenosine (FSBA) as the highest-affinity ligand (−7.3 kcal mol–1), outperforming ATP and verapamil. Triplicate 1 μs molecular dynamics simulations of apo and halo AAD06727.1 demonstrated stable rmsd, RMSF, radius of gyration, solvent-accessible surface area, hydrogen-bond occupancy, and principal component profiles, indicating preservation of structural integrity following ATP binding. Comparative simulations with the human ABCA12 nucleotide-binding domain showed conserved global architecture while revealing distinct ligand-recognition characteristics. Together, these findings identify AAD06727.1 as a structurally stable, pathogen-specific ATP-binding protein that represents a promising candidate for future structure-guided anti-H. pylori drug development.
1. Introduction
Helicobacter pylori J99 is a spiral-shaped, Gram-negative bacterium that colonizes the human gastric mucosa. As one of the most thoroughly characterized H. pylori strains, its complete genome sequence was first reported in 1999. Originally isolated from an American patient with a duodenal ulcer, the J99 strain has a relatively compact genome of approximately 1.64 Mb and a GC content of 39%. The J99 strain is predicted to encode 1491 protein-coding genes. , Notably, approximately 4% of the H. pylori J99 proteome comprises outer membrane proteins (OMPs), a proportion considered an evolutionary adaptation to the hostile, acidic environment of the human stomach.
Genomic comparison of strains J99 and 26695 reveals an average amino acid identity of 90% between orthologs, although significant divergence occurs in strain-specific genes and paralogous family expansions. Beyond its role in gastric pathogenesis, J99 serves as a valuable model for studying bacterial adaptation and host-immune crosstalk. Its high-quality genome has facilitated the development of advanced molecular tools for investigating this pathogen. The “Rosetta Stone” protein concept identifies functional associations in which two distinct polypeptides in one organism are fused into a single gene product in another. Such fusion events imply functional linkage, as physical tethering reduces dissociation entropy and increases local partner concentration. Our study targets putative proteins derived from open reading frames, including hypothetical in vivo sequences, as high-value candidates for drug development.
Domains represent the primary units of protein functional analysis, acting either independently or synergistically. “Domain mobility” facilitates recombination across distinct protein contexts, whereas “domain versatility” describes the recurrence of domains across diverse architectures, independent of fusion or truncation events. Elucidating these combinatorial interactions is essential for understanding protein evolution and identifying therapeutic targets.
In response to the global emergence of multidrug-resistant H. pylori, this study employed an integrated computational framework to identify and validate a novel pathogen-specific therapeutic target from the H. pylori J99 proteome. Rosetta Stone-based domain fusion analysis, functional reannotation, and subtractive proteomics were combined to prioritize essential, nonhomologous proteins with minimal potential for host and gut microbiota cross-reactivity. The high-priority candidate was subsequently subjected to structural modeling, binding-pocket prediction, molecular docking, and comparative molecular dynamics simulations to characterize its structural stability, ligand recognition, and conformational behavior. This integrated workflow provides a robust framework for structure-guided antimicrobial target discovery and prioritizes a promising ATP-binding protein for future therapeutic development against H. pylori.
2. Materials and Methods
2.1. Data Acquisition and Sequence Annotation
The complete protein data set for H. pylori J99 (Taxon ID: 85963) was retrieved from the NCBI Protein database on May 23rd, 2024, and aligned to the reference genome assembly GCF_000008785.1 (RefSeq). At the time of retrieval, the NCBI Protein database contained 3976 protein records associated with this taxon, including multiple entries originating from RefSeq, GenBank, UniProtKB, and other sequence repositories. Given the significant therapeutic potential of uncharacterized gene products, the analysis focused specifically on hypothetical proteins, and duplicate entries corresponding to the same genomic locus across source databases were identified and removed to obtain a nonredundant data set. To establish a functional baseline, these sequences were processed using EggNOG-mapper, a tool that facilitates functional and domain annotation by mapping sequences to precomputed orthologous groups within the EggNOG database, thereby enabling the precise identification of conserved Pfam domains and functional motifs.
Identifying evolutionary linkages is essential for characterizing unannotated proteins. To analyze proteomic complexity, we generated a domain abundance versus versatility plot. Abundance, N(d i ), measures the frequency of a given domain and was calculated as the number of nonredundant proteins in which the domain d i occurs
where A(p) denotes the set of Pfam domains assigned to protein p by EggNOG-mapper. Versatility, V(d i ), quantifies the combinatorial diversity of a domain, its tendency to occur alongside a variety of different partner domains rather than in a single fixed combination, and was calculated, adapting the concept described by Kumar and Kumar (2016), as the number of unique domain partners with which d i co-occurs across all proteins containing it
Domains with N(d i ) > 10 occurrences were retained for analysis 1 to ensure sufficient combinatorial sampling. While domain shuffling and genetic duplication are generally considered major drivers of protein evolution and would predict a positive correlation between abundance and versatility, this relationship does not hold universally, as abundance-driven functional specialization can instead suppress combinatorial diversity within a domain family. Domains exhibiting high versatility relative to their abundance were, therefore, prioritized for structural characterization and downstream screening.
The relationship between abundance and versatility was assessed using Pearson’s correlation coefficient (n = 12 domains meeting the occurrence threshold), with a two-tailed p-value and 95% confidence interval (Fisher z-transformation) reported.
2.2. Subtractive Proteomics-Based Target Identification and Prioritization
A total of 62 putative H. pylori proteins identified through the Rosetta Stone approach were subjected to systematic target prioritization using PBIT v3 (Pipeline Builder for Identification of Targets), an integrated subtractive proteomics platform that sequentially performs host nonhomology screening, antitarget analysis, gut microbiota filtering, essentiality and virulence assessment, broad-spectrum conservation analysis, host–pathogen interaction (HPI) analysis, and druggability prediction. The screening workflow implemented in PBIT v3 was used to identify pathogen-specific, essential, and therapeutically tractable targets. All analyses were performed using the integrated modules of PBIT v3, accessed on 19 December 2025. The platform incorporates multiple reference databases, including the Database of Essential Genes (DEG), Virulence Factor Database (VFDB), Database of Fungal Virulence Factors (DFVF), Host–Pathogen Interaction Database (HPIDB), Pathogen–Host Interactions Database (PHI-base), PHISTO, DrugBank, Therapeutic Target Database (TTD), the Human Microbiome Project (HMP)-derived gut microbiome data set, and an EBI-curated proteome data set comprising 181 pathogenic microorganisms, as described by Chakraborty et al. (2024). Because these resources are preintegrated within PBIT v3, their individual versions and release dates are not user-selectable and correspond to those incorporated during platform development. The human reference proteome used for host nonhomology screening was obtained independently from UniProt. The filtering criteria applied at each stage of the subtractive proteomics pipeline are described below.
2.2.1. Host Nonhomology and Antitarget Screening
To minimize potential host cross-reactivity, the 62 candidate proteins identified in the previous screening step were compared against the human proteome using BLASTp. Proteins exhibiting ≥50% sequence identity with statistically significant similarity (E-value ≤0.005) were excluded from further analysis. The 50% sequence identity threshold was selected to eliminate closely related human homologues while retaining pathogen proteins that may share conserved functional domains but lack high overall sequence similarity. Therefore, sequence identity was considered the primary exclusion criterion rather than statistical significance alone.
2.2.2. Microbiota Compatibility and Essentiality Analysis
To minimize disruption of beneficial microorganisms and reduce the likelihood of dysbiosis, the remaining proteins were screened against the human gut microbiome database integrated within the PBIT v3 pipeline, comprising 226 representative bacterial genomes curated from the Human Microbiome Project (HMP). Proteins exhibiting significant similarity (E-value ≤0.005 and sequence identity ≥50%) were excluded. The microbiota-compatible proteins were subsequently evaluated for essentiality and virulence using the DEG, the VFDB, and the DFVF. Following the criteria implemented in PBIT v3, BLASTp (Basic Local Alignment Search Tool for protein sequences) searches were performed using an E-value threshold of ≤0.0001 and a minimum alignment length of 1% to identify proteins associated with experimentally validated essential genes and virulence determinants. ,
2.2.3. Broad-Spectrum Potential and HPIs
To evaluate broad-spectrum therapeutic potential, the prioritized proteins were screened against the proteomes of 181 pathogenic species curated from the European Bioinformatics Institute (EBI) database using the broad-spectrum analysis module of PBIT v3. Homologous proteins were identified using BLAST-based sequence similarity with an E-value threshold of ≤0.0001 and a minimum alignment length of 1%, following the methodology derived from the targetTB pipeline and subsequent broad-spectrum target identification studies.
The conserved candidates were subsequently analyzed for their involvement in HPIs using HPIDB, PHI-base, and PHISTO. Proteins exhibiting significant similarity to experimentally characterized HPI factors (E-value ≤0.0001; alignment length ≥1%) were prioritized as potential modulators of HPIs.
2.2.4. Assessment of Druggability and Therapeutic Potential
To evaluate therapeutic tractability, the prioritized proteins were screened against established drug-target repositories using the druggability module of PBIT v3. BLASTp searches were performed against DrugBank, the Therapeutic Target Database (TTD), and ChEMBL using an E-value threshold of ≤0.0001 and a minimum alignment length of 1%. Proteins exhibiting homology to experimentally validated therapeutic targets were considered to possess favorable potential for drug repurposing and structure-based inhibitor design. In addition, the membrane topology of the shortlisted proteins was predicted using DeepTMHMM, a deep learning-based sequence analysis tool that classifies proteins as globular, transmembrane, or signal peptide-containing and predicts the number and positions of transmembrane helices directly from the amino acid sequence. This analysis was performed to determine the predicted subcellular topology of the candidate targets and to assess their suitability for subsequent structural modeling and molecular docking studies.
To further evaluate potential off-target effects on beneficial gastric microorganisms, the final prioritized target (AAD06727.1) was additionally compared against representative gastric commensal bacteria using BLASTp. Representative species included Streptococcus mitis, Streptococcus salivarius, Neisseria subflava, Actinomyces odontolytica, and Prevotella spp., which are commonly reported constituents of the healthy gastric microbiota. Sequence similarity was assessed using the same criteria applied during host nonhomology analysis (E-value ≤0.005; sequence identity ≥50%).
2.3. Structural Modeling, Quality Validation, and Active Site Prediction of AAD06727.1
Following subtractive proteomics-based target identification, the FASTA sequence of AAD06727.1 was submitted to AlphaFold2 for high-accuracy tertiary structure prediction. The top-ranked conformation was selected based on integrated ipTM + pTM (interface and predicted Template Modeling) scores, with structural reliability validated using pLDDT confidence categories (90–100: very high; 70–90: confident; 50–70: low; <50: very low). To optimize stereochemistry, the AlphaFold2 model was processed with GalaxyRefine, which employs iterative side-chain rebuilding and MD-based (molecular dynamics-based) relaxation to reach the lowest-energy state. The final refined model was selected based on the lowest total energy score, ensuring superior local stereochemistry and structural integrity as validated in CASP (Critical Assessment of protein Structure Prediction) benchmarks. , PrankWeb 2.5.1, implementing the P2Rank machine-learning algorithm, was used for template-free ligand-binding site prediction. P2Rank ranks predicted binding pockets using its internal prediction score derived from local geometric and physicochemical properties of protein surface points. For each predicted pocket, PrankWeb additionally reports descriptors including ligandability probability, pocket volume, solvent exposure, and residue composition, which were used to interpret the predicted binding site. The highest-ranked pocket was selected for subsequent structural and docking analyses.
2.4. Structural Comparison with Experimentally Determined Homologues
To independently evaluate the predicted tertiary structure, the AlphaFold2 model of AAD06727.1 was compared to experimentally characterized ABC transporter ATP-binding domains identified by BLASTp against the UniProtKB/Swiss-Prot database. The closest structural homologue corresponded to the nucleotide-binding domain 2 (NBD2) of the human ABC transporter ABCA12 (UniProt accession Q86UK0). The aligned NBD2 region was extracted and structurally superposed onto the AlphaFold model using the super algorithm implemented in PyMOL v2.5.5. Structural similarity was assessed using the root-mean-square deviation (rmsd) of the aligned atoms.
2.5. Structure-Based Ligand Recognition and Binding Site Characterization
2.5.1. Preparation of Receptor Structures
The refined AlphaFold2 model of AAD06727.1, described above, was used as the bacterial receptor in the molecular docking. The ligand-free (apo) conformation served as the initial receptor for the docking studies. To assess structural conservation with the closest human homologue, the experimentally determined nucleotide-binding domain 2 (NBD2) of human ABCA12 (UniProt accession Q86UK0; structural model 7lkp.1.A) was retrieved from the UniProt/Swiss-Model repository. The region corresponding to the aligned 228-residue nucleotide-binding domain was extracted and used as the human receptor for comparative docking analysis.
To obtain the holo conformation of AAD06727.1, the ATP-bound complex identified from molecular docking was subjected to a 100 ns molecular dynamics (MD) simulation as follows: the steps mentioned in Section . The equilibrated structure obtained at the end of the simulation was subsequently used as the holo-receptor for additional docking analyses.
2.5.2. Ligand Preparation
Three ligands were selected for molecular docking studies: adenosine triphosphate (ATP; PubChem CID: 5957), 5′-fluorosulfonylbenzoyl-5′-adenosine (FSBA; PubChem CID: 21121712), and verapamil (PubChem CID: 2520). The three-dimensional structures of all ligands were retrieved from the PubChem database, with ATP used as the native ligand, FSBA as a well-characterized ATP analogue, and verapamil included as a comparative ABC transporter substrate and modulator. ATP was selected to validate the predicted nucleotide-binding pocket, whereas FSBA and verapamil were employed to evaluate ligand recognition and binding specificity toward the ATP-binding site of AAD06727.1. The ligand structures were prepared using AutoDock Tools by adding polar hydrogen atoms, assigning Gasteiger partial charges, defining rotatable bonds, and converting the structures into PDBQT format.
2.5.3. Molecular Docking Workflow
Receptor structures were prepared by removing nonprotein molecules, adding polar hydrogen atoms, assigning Kollman charges, and converting the proteins into PDBQT format using AutoDock Tools. Molecular docking was performed using AutoDock Vina (version 1.1.2). The docking grid was centered on the highest-ranked PrankWeb-predicted nucleotide-binding pocket encompassing the conserved Walker A and Walker B motifs. Grid center coordinates, box dimensions, exhaustiveness, number of output poses, and energy range are provided in Table S1. For each docking experiment, multiple binding poses were generated, and the pose with the lowest predicted binding free energy (kcal mol–1) was selected for further analysis. Protein–ligand interactions were visualized and analyzed using Ligplot+ and PyMOL.
2.5.4. Comparative Docking Strategy
A comparative docking strategy was employed to investigate ligand recognition by bacterial and human ABC transporter ATP binding domains. Initially, ATP, FSBA, and verapamil were independently docked to the apo-conformation of AAD06727.1 to identify their binding modes and relative binding affinities within the predicted nucleotide-binding pocket.
To evaluate the structural specificity of ligand binding, the same three ligands were subsequently docked to the human ABCA12 NBD2 structure by using identical docking parameters. Comparison of the predicted binding affinities and interaction profiles between AAD06727.1 and human ABCA12 provided a structural basis for assessing the relative selectivity of the bacterial targets. The ATP-bound AAD06727.1 complex exhibiting the most favorable docking pose was then subjected to a 100 ns molecular dynamics simulation to generate an equilibrated holo-conformation. Following equilibration, FSBA and verapamil were redocked into the holo structure to investigate whether ATP-induced conformational changes influenced ligand binding within the nucleotide-binding pocket. ATP was not redocked into the holo receptor because the binding site was already occupied during the molecular dynamics simulation.
2.6. Comparative Molecular Dynamics Simulations and Trajectory Analysis
To investigate the conformational stability of AAD06727.1 and its ATP-bound complex, all-atom molecular dynamics (MD) simulations were performed using GROMACS 2023.2 with the CHARMM36m force field. Simulation systems were prepared using the CHARMM-GUI Solution Builder, which generated the system topology, solvated each protein in explicit TIP3P water molecules, added counterions to achieve electroneutrality, and produced GROMACS-compatible topology and parameter files. Energy minimization was performed by using the steepest descent algorithm for 5000 steps until convergence. The minimized systems were subsequently equilibrated under positional restraints in the NVT ensemble for 125 ps at a 1 fs integration time step. Temperature was maintained at 303.15 K using the velocity-rescaling (V-rescale) thermostat.
Production simulations were performed with a 2 fs integration time step under periodic boundary conditions. Long-range electrostatic interactions were treated using the particle mesh Ewald (PME) method with a real-space cutoff of 1.2 nm, while van der Waals interactions were calculated using the force-switch scheme with switching initiated at 1.0 nm and a cutoff of 1.2 nm. Temperature was controlled at 303.15 K using the V-rescale thermostat, and isotropic pressure was maintained at 1 bar by using the C-rescale barostat. Covalent bonds involving hydrogen atoms were constrained by using the LINCS algorithm.
To ensure the reproducibility of the observed conformational behavior, three independent 1000 ns production simulations were performed for both the apo AAD06727.1 and ATP-bound (holo) AAD06727.1 systems using identical simulation parameters but different initial velocity seeds. Prior to comparative ligand docking, the ATP-bound complex obtained from molecular docking was subjected to an initial 100 ns MD simulation to generate an equilibrated holo-conformation. The final frame of this simulation was subsequently used as the receptor for docking FSBA and verapamil, thereby accounting for the ATP-induced structural relaxation and receptor flexibility. The use of molecular dynamics-refined holo conformations for subsequent docking has been shown to provide a more realistic representation of ligand-binding environments than docking directly to the initial docked complex or an apo structure. For comparison with the closest human structural homologue, single 1000 ns MD simulations were performed for the apo nucleotide-binding domain 2 (NBD2) of human ABCA12 and the corresponding FSBA-bound ABCA12 complex using the same simulation protocol.
Trajectory analyses were performed using standard GROMACS utilities. Backbone structural stability was evaluated by calculating the root-mean-square deviation (rmsd) with gmx rms, residue flexibility by the root-mean-square fluctuation (RMSF) using gmx rmsf, structural compactness by the radius of gyration (R g) using gmx gyrate, solvent accessibility by the solvent-accessible surface area (SASA) using gmx sasa, and protein–ligand hydrogen-bond occupancy using gmx hbond. For the triplicate AAD06727.1 simulations, the resulting trajectories were processed using a custom Python script that aligned the data sets according to identical simulation time points (or residue numbers for RMSF) and calculated the arithmetic mean and standard deviation at each point across the three independent simulations. The resulting mean ± standard deviation profiles were used as representative trajectories for all comparative analyses, including comparisons with the corresponding single 1000 ns simulations of human ABCA12 NBD2.
2.6.1. Principal Component Analysis
To investigate the large-scale collective motions of AAD06727.1 during molecular dynamics simulations, principal component analysis (PCA) was performed using the covariance matrix of Cα atomic fluctuations after least-squares fitting to the initial structure. Projection of the trajectories onto the first two principal components (PC1 and PC2) was generated using the GROMACS PCA utilities. For apo and ATP-bound (holo) AAD06727.1, three independent 1000 ns simulations were analyzed separately. The resulting two-dimensional PCA projection coordinates (PC1 and PC2) from the three replicate simulations were subsequently processed using a custom Python script that aligned trajectories with identical frame numbers and calculated the arithmetic mean of the PC1 and PC2 coordinates at each corresponding frame. The averaged projection was visualized as a two-dimensional scatter plot using Matplotlib with a frame-dependent color gradient to illustrate the progression of conformational sampling during the simulations.
3. Results
3.1. Functional Landscape and Domain Architecture of the H. pylori Putative Proteome
From the 3976 retrieved protein records, 1265 hypothetical proteins were identified. These sequences were analyzed using EggNOG-mapper for domain identification and orthology assignment, resulting in 959 proteins with functional annotations, whereas 306 proteins lacked orthologous matches, suggesting lineage-specific or highly divergent evolutionary origins (Table S2).
Notably, 19 proteins were identified as members of the Helicobacter outer membrane protein (HP_OMP) family. These proteins are critical for bacterial adhesion to host gastric epithelia and serve as primary determinants of environmental niche adaptation and virulence. Beyond OMPs, the analysis identified 15 sequences harboring the helicase conserved C-terminal domain (Helicase_C), a hallmark of the DnaB superfamily required for ATP-dependent DNA unwinding during replication. Furthermore, 15 proteins were assigned to the response regulator receiver (Response_reg) domain. These βα5-fold structures are central components of two-component signal transduction systems, translating gastric environmental stimuli into transcriptional responses. Additionally, 14 proteins were classified under the GmrSD restriction endonuclease C-terminal domain (DUF1524), which employs an HNH endonuclease motif to cleave foreign DNA, thereby providing a defense against bacteriophage infection.
3.2. Metabolic Adaptability and Stress-Response Domains
The proteomic survey identified several domain families essential for redox homeostasis and nutrient acquisition. Thirteen proteins were found to contain the 4Fe–4S dicluster domain (Fer4_8), a key prosthetic group for electron transfer. Similarly, the GmrSD N-terminal domain (DUF262) was identified in 13 proteins, featuring a pseudo-β-barrel structure that facilitates site-specific binding and asymmetric dimerization for the recognition of modified DNA. Additional findings included 12 proteins containing the Bax1-I domain, 12 from the DEAD/DEAH box helicase family (including the cold-stress helicase RhpA), 12 proteins featuring Radical SAM domains, and 12 containing 2Fe–2S motifs typically regulated by the Ferric Uptake Regulator (Fur). Finally, 11 proteins possessed Sel1 repeats and an additional 11 contained ABC transporter domains. These transporters are essential for the energy-dependent uptake of nickel and dipeptides, which are requisite for urease activity and overall bacterial fitness (Table S3).
3.3. Quantitative Assessment of Domain Abundance and Versatility
To evaluate evolutionary constraints on domain architecture, we conducted a dual-metric assessment of domain abundance and versatility (Table ). Following the approach of Kumar and Kumar (2016), domains with N(d i ) > 10 occurrences were retained to quantify biological necessity (abundance) and combinatorial capacity (versatility). The scatter plot (Figure ) illustrates the relationship between domain abundance (x-axis) and versatility index (y-axis) for the identified protein domains. Figure highlights an inverse trend: highly abundant domains, such as HP_OMP, exhibit lower versatility, suggesting functional specialization. Conversely, domains such as Radical_SAM, Response_reg, and Fer4_8 exhibit relatively high versatility, despite comparatively lower abundance, indicating participation in more diverse protein architectures.
1. Shared Domains with Their Abundance-Versatility Scores.
| sl. no | domain | abundance | versatility |
|---|---|---|---|
| 1. | HP_OMP | 19 | 0 |
| 2. | Helicase_C | 15 | 1 |
| 3. | Response_reg | 15 | 4 |
| 4. | DUF1524 | 14 | 2 |
| 5. | Fer4_8 | 13 | 4 |
| 6. | DUF262 | 13 | 2 |
| 7. | Bax1-I | 12 | 0 |
| 8. | DEAD | 12 | 1 |
| 9. | Fer2_3 | 12 | 1 |
| 10. | Radical_SAM | 12 | 4 |
| 11. | Sel1 | 11 | 3 |
| 12. | ABC_tran | 11 | 3 |
1.
Relationship between domain abundance and versatility for shared Pfam domains identified across the nonredundant hypothetical protein data set of H. pylori J99. Abundance, N(d i ), denotes the number of nonredundant proteins in which a domain occurs; versatility, N(d i ), denotes the number of unique co-occurring domain partners. Only domains with N(d i ) > 10 are shown (n = 12). Each point represents one domain, labeled by name.
Pearson correlation analysis (n = 12 domains) was performed to statistically evaluate this relationship (Figure S1). Abundance and versatility were negatively correlated (r = −0.327, R 2 = 0.107, p = 0.299, 95% CI [−0.76, 0.30]). This association did not reach statistical significance at the present sample size; however, the direction of the relationship is consistent with a functional trade-off, in which highly frequent domains tend toward specialization at the expense of combinatorial diversity. Given the weak coefficient of determination (R 2 = 0.107), abundance alone accounts for only a small fraction of the variance in versatility, indicating that additional, unmeasured factors likely govern a domain’s combinatorial diversity. This trend warrants confirmation in a larger domain set.
The scatter plot (Figure ) further reveals a cluster of high-abundance, low-versatility domains (e.g., HP_OMP, Bax1-I, and DEAD), likely representing domains that evolved through extensive internal duplication of specialized motifs. Conversely, domains with moderate-to-high versatility (Response_reg, Fer4_8, Radical_SAM, Sel1, and ABC_tran) occur across a comparatively narrower abundance range, favoring diverse tethering over numerical expansion. Together, these patterns suggest that the structural complexity of the H. pylori J99 proteome is shaped by a combination of domain recurrence and combinatorial diversification, with abundance alone providing limited explanatory power over versatility.
3.4. Domain Fusion-Based Functional Linkage Prediction and Interaction Networks
To elucidate functional connectivity within the H. pylori proteome, the five most prevalent domains, characterized by high abundance and versatility, were further analyzed for domain–domain interactions (DDIs) (Table S4). This mapping revealed physical and functional interfaces between core modules and their respective partners, resulting in a complex, multilayered interaction network (Figure S2A). By correlating priority scores with the structural layout, a clear functional hierarchy emerges within the domain association network. High-degree hub complexes, including Radical_SAM, Response_reg, and Fer4_8, function as central “star” nodes coordinating with diverse partners such as SPASM, CheW, and LUD_dom. These hubs correspond to maximum priority scores (5.0) and exhibit the highest versatility indices (up to ∼0.33), underscoring their roles as indispensable mediators in biological signaling and metabolic pathways.
The integration of the ranked domains bar chart with the domain interaction network provides a multifaceted view of protein domain organization (Figure S2B). The persistent co-occurrence within this interaction network of reciprocal pairs, specifically DEAD/Helicase_C and DUF262/DUF1524, suggests that these constitute mandatory functional modules that likely operate as inseparable units. These pairs occupy the midrange of the priority scale (scores of 2.0 to 3.0), reflecting specialized rather than global functional roles. Conversely, isolates such as HP_OMP and Bax1-I represent domains that function primarily as autonomous structural elements. These isolates receive the lowest priority scores (1.0), reinforcing their status as standalone components with minimal connectivity within this biological context.
3.5. Multicriteria Prioritization and Functional Characterization of Therapeutic Targets
To identify high-priority therapeutic targets in H. pylori J99, we employed a multistage computational pipeline to isolate candidates with optimal safety and efficacy profiles (Figure S3A). An initial ensemble of 62 proteins, identified via Rosetta Stone-based domain fusion, underwent nonhomology screening against the Homo sapiens proteome. This step excluded two candidates to prevent host cross-reactivity and toxicity; the remaining 60 sequences showed no significant similarity to human proteins or established antitargets (Tables S5 and S6). To protect beneficial commensal species, these 60 proteins were screened against the human gut microbiota, yielding 22 candidates with negligible microbial similarity (Table S7).
Functional evaluation of the 22 candidates confirmed their essentiality, with 18 specifically associated with characterized virulence factors (Table S8). We further prioritized these targets by correlating essential gene homologues (DEG) with virulence factor homologues (VFDB). The resulting scatter plot (Figure S3B) identifies high-impact candidates, such as AAD06132.1, which exhibits high essentiality (∼410 homologues) and virulence (>40 homologues). This alignment suggests that central hubs in the domain association network are critical for both pathogen viability and pathogenicity. Broad-spectrum conservation analysis using the PBIT v3 reference database (comprising proteomes from 181 pathogenic organisms) identified 1056 homologous protein alignments for AAD06727.1 under the specified BLAST criteria, indicating extensive conservation across the reference data set (Table S9). These candidates, exceeding our 500-homologue threshold, represent core, indispensable biological components and constitute potential candidate targets for broad-spectrum therapeutic intervention (Figure S4A,B). It should be noted that this conservation reflects computational sequence-level similarity across the reference proteomes and does not, by itself, demonstrate an experimentally validated antimicrobial spectrum.
To prioritize factors at the host–pathogen interface, candidates were mapped against HPI data sets, confirming interaction-related homology for 18 proteins (Table S10). Druggability assessment via BLASTp against DrugBank, ChEMBL, and TTD identified five proteins with >30% sequence identity to known therapeutic targets (Table S11). These candidates possess well-characterized structural pockets or domains already targeted by small molecules. Based on druggability depth, AAD06132.1 emerged as the top-ranked target with nine unique database hits, followed by AAD06727.1 with eight hits (Figure S5). The convergence of network priority, conservation, and validated druggability identifies these proteins as promising candidate targets for broad-spectrum intervention. AAD06727.1 was selected as the primary lead on account of its 40% identity with the Candida elongation factor 3 (TEF3). Functional reannotation identified AAD06727.1 (gene locus jhp_1141; GenBank protein accession AAD06727.1; UniProt accession Q9ZK06) as encoded on the complementary strand of the H. pylori J99 chromosome (AE001439.1; genomic coordinates 1273981–1274667) and encoded a 228-residue ABC transporter ATP-binding protein. GO and InterPro analyses confirmed its role in ATP binding and hydrolysis, with the ABC_tran domain (PROSITE: PS50893) spanning 97.8% of the sequence (Val3–Glu225), indicating a highly specialized architecture.
To further justify target prioritization, the five shortlisted proteins were compared based on their predicted structural architecture and suitability for structure-based drug discovery (Table S12). DeepTMHMM classified AAD06727.1, AAD05869.1, AAD06868.1, and AAD07036.1 as soluble globular proteins lacking transmembrane helices, whereas AAD06132.1 was predicted to be an integral membrane ABC transporter containing six transmembrane helices. Although AAD06132.1 exhibited the highest druggability support based on homology to known therapeutic targets (nine database matches), its membrane-associated architecture presents additional challenges for structural characterization, molecular docking, and inhibitor optimization. In contrast, AAD06727.1 demonstrated comparable druggability support (eight database matches) while possessing a compact 228-residue soluble ABC ATP-binding domain that spans approximately 97.8% of the protein sequence and contains the conserved Walker A and Walker B nucleotide-binding motifs. These characteristics, together with its broad-spectrum conservation and high-confidence AlphaFold structure, make AAD06727.1 a particularly promising candidate for structure-based drug design and molecular dynamics simulations. In comparison, AAD05869.1 displayed limited druggability support (one therapeutic target homologue), whereas AAD06868.1 and AAD07036.1 were predicted to function as transcriptional regulators rather than catalytic ATPases, making them less suitable for the inhibitor development strategy pursued in this study.
Although AAD06727.1 produced statistically significant BLAST hits against several human ABC transporter ATPase domains, the highest sequence identity was approximately 30%, which is substantially below the predefined exclusion threshold of 50%. These matches reflect the evolutionary conservation of ATP-binding domains rather than close overall homology.
BLASTp analysis identified homologous ABC transporter ATP-binding proteins in representative gastric commensals, including S. mitis, S. salivarius, N. subflava, Actinomyces spp., and Prevotella spp. However, all identified homologues exhibited sequence identities below the predefined exclusion threshold of 50% (approximately 30–46%), and therefore no gastric commensal protein met this exclusion criterion.
BLASTp analysis against publicly available H. pylori protein sequences demonstrated that AAD06727.1 is highly conserved across multiple H. pylori strains, exhibiting 99–100% sequence identity over the full-length 228-amino-acid sequence. This high degree of conservation indicates that AAD06727.1 is not unique to strain J99 but represents a conserved ATP-binding protein across diverse H. pylori isolates, supporting its suitability as a broadly applicable therapeutic target.
3.6. de novo Structural Modeling and High-Fidelity Validation
In the absence of an experimentally determined PDB template, the tertiary structure of AAD06727.1 was predicted using AlphaFold2, yielding a high-confidence model with a pTM score of 0.92 (Figure S6). Residue-level confidence analysis further demonstrated high reliability of the predicted ATP-binding pocket. The conserved Walker A motif (residues 35–42) exhibited pLDDT values ranging from 92.25 to 97.57 (mean = 94.77), while the Walker B motif (residues 147–151) showed pLDDT values ranging from 89.56 to 95.11 (mean = 93.19), indicating that the catalytic nucleotide-binding residues were modeled with high confidence (Table S13). Subsequent refinement using GalaxyRefine optimized side-chain orientations and local stereochemistry. Model 1 was selected based on a GDT-TS of 0.989 and a MolProbity score of 1.537. PROCHECK validation confirmed exceptional stereochemical quality, with 97.1% of residues occupying the most favored Ramachandran regions and no residues in disallowed regions (Figure S7A).
Surface quality and environmental compatibility were assessed using ERRAT and Verify3D. The ERRAT quality factor reached 98.1% (Figure S7B), while Verify3D confirmed that 80.26% of residues maintained a 3D–1D compatibility score ≥0.1 (Figure S8A). Additionally, a ProSA-web Z-score of −7.54 (Figure S8B) confirmed the model’s structural validity within the typical range for native crystal structures of comparable size.
Following validation, PrankWeb 2.5.1, which implements the P2Rank machine learning algorithm for ligand-binding site prediction, was used to identify potential druggable interfaces. The highest-ranked ligand-binding pocket (Rank 1) yielded a P2Rank score of 7.37 with a ligandability probability of 0.426, indicating the most probable ligand-binding region within the protein. Although P2Rank does not define a universal numerical threshold for druggability, higher scores represent greater confidence in ligand-binding potential, and the top-ranked pocket is generally considered the most suitable candidate for structure-based drug design. This pocket comprised a conserved residue cluster consisting of Tyr12, Lys16, Gly40, Lys41, Thr42, and Glu151, overlapping with the conserved Walker A (residues 35–42) and Walker B (residues 147–151) ATP-binding motifs, thereby supporting its functional relevance as the nucleotide-binding site. Notably, the ATP-binding residues identified by PrankWeb overlap with the conserved Walker A (Gly40–Thr42) and Walker B (Glu151) motifs, all of which exhibited high AlphaFold confidence (pLDDT > 89.5). The functional relevance of these residues was corroborated by the subsequent MD simulations, which confirmed high positional stability and participation in critical intramolecular hydrogen-bonding networks.
To further assess the structural reliability of the predicted model, AAD06727.1 was compared with experimentally characterized ABC transporter ATP-binding domains available in the Protein Data Bank. BLASTp analysis against the UniProtKB/Swiss-Prot database identified the nucleotide-binding domain 2 (NBD2) of the human ABC transporter ABCA12 (UniProt accession Q86UK0) as the closest experimentally characterized structural homologue, exhibiting approximately 30% sequence identity across 94% query coverage. Structural superposition of the AlphaFold-predicted AAD06727.1 model with the corresponding ABCA12 NBD2 structure using the PyMOL super algorithm yielded an rmsd of 1.88 Å over 1116 aligned atoms, confirming conservation of the canonical ABC ATPase fold despite relatively low overall sequence identity (Figure S9). This comparison served solely as an independent structural validation of the predicted nucleotide-binding architecture and was not used for model construction or ligand placement. Accordingly, all molecular docking and molecular dynamics simulations were performed directly on the AlphaFold-predicted structure of AAD06727.1. The ATP-bound (holo) conformation used for subsequent docking studies was generated by docking ATP into the AlphaFold model followed by 100 ns molecular dynamics equilibration, after which ATP analogues were docked into the equilibrated holo structure.
3.7. Molecular Docking Identifies FSBA as the Preferred Ligand for the ATP-Binding Pocket of AAD06727.1
To evaluate ligand recognition by the predicted ATP-binding pocket of AAD06727.1, molecular docking was performed using ATP, the ATP analogue 5′-fluorosulfonylbenzoyl-5′-adenosine (FSBA), and verapamil against the apo conformation of AAD06727.1. Comparative docking was subsequently carried out against the nucleotide-binding domain 2 (NBD2) of the closest human structural homologue, ABCA12, to assess relative binding behavior. Among the three ligands, FSBA exhibited the most favorable binding affinity toward apo AAD06727.1, with a docking score of −7.3 kcal mol–1, followed by ATP (−6.7 kcal mol–1) and verapamil (−4.8 kcal mol–1). Docking against human ABCA12 NBD2 yielded binding affinities of −6.1 kcal mol–1 for ATP, −6.0 kcal mol–1 for FSBA, and −5.5 kcal mol–1 for verapamil. Compared with the human homologue, FSBA exhibited approximately 1.3 kcal mol–1 stronger predicted binding to AAD06727.1, whereas ATP showed a smaller difference of approximately 0.6 kcal mol–1. These results indicate that although the ATP-binding fold is structurally conserved, differences in the local binding environment favor recognition of FSBA by the bacterial ATP-binding protein.
To determine whether ATP-induced conformational changes influenced ligand recognition, FSBA and verapamil were subsequently docked into the ATP-equilibrated holo structure of AAD06727.1 generated following a 100 ns molecular dynamics simulation. In the holo conformation, the docking score of FSBA decreased from −7.3 to −6.1 kcal mol–1, whereas verapamil retained an identical binding affinity of −4.8 kcal mol–1 in both the apo and holo structures. The reduced affinity of FSBA following ATP binding suggests that occupancy of the nucleotide-binding pocket alters its accessibility and local conformational environment without eliminating ligand recognition. Overall, the comparative docking analyses identified FSBA as the highest-affinity ligand for the predicted ATP-binding pocket of AAD06727.1, while ATP displayed the expected native binding behavior and verapamil consistently exhibited weaker interactions. Accordingly, FSBA was selected for subsequent structural analyses, whereas the ATP-bound complex served as the equilibrated holo conformation for the molecular dynamics simulations.
3.7.1. Protein–Ligand Interaction Analysis
The predicted binding modes of ATP, FSBA, and verapamil within the nucleotide-binding pocket of apo AAD06727.1, holo AAD06727.1, and human ABCA12 NBD2 were further examined using LigPlot+. Representative complete two-dimensional interaction diagrams for all docked complexes are provided in Figures S10–S12.
In the apo AAD06727.1 structure, ATP formed 10 hydrogen bonds with residues lining the predicted nucleotide-binding pocket, while FSBA established 8 hydrogen bonds. Verapamil exhibited only 2 hydrogen bonds, consistent with its lower predicted binding affinity. Despite forming fewer hydrogen bonds than ATP, FSBA displayed the most favorable docking score (−7.3 kcal mol–1), suggesting that additional noncovalent interactions, including hydrophobic contacts and van der Waals interactions, contribute to its enhanced binding stability. For the human ABCA12 NBD2 structure, ATP, FSBA, and verapamil formed 6, 9, and 2 hydrogen bonds, respectively. Although FSBA established the highest number of hydrogen bonds with ABCA12, its predicted binding affinity (−6.0 kcal mol–1) remained lower than that observed for AAD06727.1 (−7.3 kcal mol–1), indicating differences in the overall interaction environment between the bacterial and human ATP-binding domains.
Following generation of the ATP-equilibrated holo conformation of AAD06727.1, FSBA and verapamil were redocked to examine ligand recognition in the ATP-bound state. FSBA formed 6 hydrogen bonds, whereas verapamil established only one hydrogen bond with the holo structure. Concurrently, the predicted binding affinity of FSBA decreased from −7.3 kcal mol–1 in the apo conformation to −6.1 kcal mol–1 in the holo conformation, suggesting that ATP-induced conformational changes partially altered the accessibility and interaction network of the nucleotide-binding pocket while still permitting ligand accommodation. Overall, the interaction analyses support the docking results by demonstrating that all three ligands occupy the predicted ATP-binding pocket identified by PrankWeb. Among the tested compounds, FSBA exhibited the most favorable binding affinity toward AAD06727.1 and maintained an extensive interaction network in both the apo and holo conformations, supporting its selection for subsequent structural analyses.
3.8. Comparative Molecular Dynamics Analysis of AAD06727.1 (Apo and Holo Triplicates) Versus Human ABCA12 NBD2 (Apo and FSBA-Bound)
To obtain an equilibrated ATP-bound (holo) receptor for subsequent docking studies, the ATP-AAD06727.1 complex generated by molecular docking was subjected to a 100 ns molecular dynamics simulation (Figure S13). The backbone rmsd increased during the initial phase of the simulation, reflecting structural relaxation of the docked complex, and subsequently reached a stable plateau with only minor fluctuations throughout the remainder of the trajectory. The complex exhibited an average backbone rmsd of 0.183 nm, with a maximum deviation of 0.270 nm, indicating that the ATP-bound structure remained conformationally stable during equilibration. ATP remained accommodated within the predicted nucleotide-binding pocket throughout the simulation, allowing the surrounding binding-site residues to adopt a relaxed holo conformation. The final frame of the equilibrated trajectory was therefore extracted and used as the receptor for comparative docking of FSBA and verapamil.
To directly compare conformational behavior across the bacterial target and its closest human structural homologue and to consolidate the triplicate apo and holo AAD06727.1 simulations with the human ABCA12 NBD2 simulations into a single interpretive framework, the rmsd, RMSF, R g, SASA, and intermolecular hydrogen-bond results originally presented as separate Sections , 3.8.2, and 3.8.3 are combined here. Four simulated systems are considered throughout: (1) apo AAD06727.1, triplicate (three independent 1000 ns replicas); (2) ATP-bound (holo) AAD06727.1, triplicate; (3) apo human ABCA12 NBD2 and its FSBA-bound complex, single 1000 ns trajectories each; and (4) the direct apo-AAD06727.1-versus-apo-ABCA12-NBD2 comparison, using the triplicate mean for AAD06727.1 against the single ABCA12 trajectory.
3.8.1. Global Structural Stability and Compactness
Across all four panels (Figure ), every system reached a stable rmsd plateau after initial equilibration and remained at that level for the full 1000 ns production run, with no trajectory showing progressive drift. Numerically, the apo AAD06727.1 triplicate averaged 0.184, 0.239, and 0.152 nm across replicas 1–3 (Panel A), while the holo triplicate averaged 0.168, 0.166, and 0.189 nm (Panel B); holo values are, if anything, marginally lower and tighter across replicas than apo, indicating that ATP binding does not destabilize the backbone and may modestly narrow interreplicate spread. The human ABCA12 NBD2 apo- and FSBA-bound systems averaged 0.164 and 0.182 nm, respectively (Panel C), each within the same general range as the bacterial protein. The most informative comparison is panel D: the triplicate-mean apo AAD06727.1 rmsd (0.191 nm) is substantially lower than the single-trajectory apo ABCA12 NBD2 rmsd (0.329 nm), indicating that, under these simulation conditions, the bacterial ATP-binding domain samples a comparatively more rigid backbone than its human structural counterpart, while both proteins nonetheless preserve their overall fold throughout the simulations.
2.
Comparative backbone rmsd across all simulated systems. (A) Apo AAD06727.1 triplicate; (B) holo AAD06727.1 triplicate; (C) apo vs FSBA-bound human ABCA12 NBD2; (D) apo AAD06727.1 (triplicate mean) vs apo ABCA12 NBD2.
All systems maintained narrow, stable R g distributions with no evidence of progressive expansion or collapse (Figure ). The apo AAD06727.1 triplicate averaged 1.852–1.864 nm across replicas (panel A), while the holo triplicate averaged a consistently lower 1.812–1.816 nm (panel B), a modest and reproducible compaction upon ATP binding that is directionally consistent across all three replicas. Human ABCA12 NBD2 showed closely matched apo- and FSBA-bound R g (1.839 and 1.847 nm; panel C), indicating negligible compaction or expansion upon FSBA binding in the human domain. In the direct cross-species comparison (Panel D), the triplicate-mean apo AAD06727.1 R g (1.860 nm) was modestly higher than the single-trajectory apo ABCA12 NBD2 R g (1.795 nm), indicating that the bacterial domain adopts a marginally less compact, though equally stable, global fold relative to its human counterpart under these simulation conditions.
3.
Comparative radius of gyration (R g) across all simulated systems. (A) Apo AAD06727.1 triplicate; (B) holo AAD06727.1 triplicate; (C) apo vs FSBA-bound ABCA12 NBD2; (D) apo AAD06727.1 (triplicate mean) vs apo ABCA12 NBD2.
3.8.2. Residue Dynamics and Hydrogen-Bond Interaction Profiles
All five panels (Figure ) show the same qualitative architecture: a rigid core (RMSF predominantly below 0.20 nm) punctuated by discrete, high-amplitude peaks confined to loop and terminal regions. Average RMSF values were 0.113, 0.138, and 0.086 nm for the apo AAD06727.1 replicas (panel A) and 0.093, 0.113, and 0.122 nm for the holo replicas (panel B), showing that ATP binding does not broadly rigidify or destabilize the protein relative to the apo state, consistent with the rmsd comparison above. Human ABCA12 NBD2 showed closely matched apo- and FSBA-bound average RMSF (0.118 and 0.123 nm, respectively; panel C), indicating that ligand binding has a similarly modest effect on local flexibility in the human domain. Comparing the triplicate-mean AAD06727.1 profile (Panel D) against the single-trajectory ABCA12 NBD2 profile (Panel E), both proteins confine their largest fluctuations to solvent-exposed loops and termini while maintaining a comparatively stable, low-RMSF nucleotide-binding core, a shared architectural signature consistent with conservation of the ATP-binding fold across the bacterial-to-human evolutionary distance, even though the human domain reaches a higher maximum RMSF (up to 0.572 nm) than the bacterial protein (up to 0.449 nm).
4.
Comparative per-residue RMSF across all simulated systems. (A) Apo AAD06727.1 triplicate; (B) holo AAD06727.1 triplicate; (C) apo vs FSBA-bound ABCA12 NBD2; (D) apo AAD06727.1 triplicate-mean; (E) apo ABCA12 NBD2, single trajectory.
The bacterial ATP-AAD06727.1 complex maintained a higher average hydrogen-bond count (5.72, 4.37, and 9.02 bonds across replicas 1–3; panel A) than the human FSBA-ABCA12 NBD2 complex (4.56 bonds; panel B) and also sampled a wider instantaneous range (0–15 vs 0–8 bonds) (Figure ). This is consistent with ATP, the domain’s native, evolutionarily coadapted substrate engaging a more extensive and more variable hydrogen-bonding network than FSBA, a covalent/affinity-labeling ATP-competitive probe binding the human domain. In both systems, hydrogen bonds were repeatedly formed and broken rather than persistently absent, indicating that the respective ligand remained associated with its nucleotide-binding pocket throughout the full 1000 ns trajectory in both the bacterial and human complexes, despite the quantitative difference in bond count and variability.
5.
Comparative intermolecular hydrogen-bond occupancy. (A) ATP-AAD06727.1 (triplicate; replicas 1–3); (B) FSBA-ABCA12 NBD2 (single trajectory).
3.8.3. Solvent Accessibility and Structural Integrity
The SASA trends mirror the R g comparison (Figure ). Apo AAD06727.1 averaged 119–125 nm2 across the three replicas (panel A), while the holo triplicate averaged a consistently lower 117–118 nm2 (panel B), reinforcing the conclusion from Section that ATP binding is accompanied by a small, reproducible reduction in both compactness and solvent exposure rather than any destabilization. Human ABCA12 NBD2 apo and FSBA-bound states showed near-identical average SASA (119.36 nm2 and 119.87 nm2; panel C; difference of only 0.51 nm2), indicating that FSBA binding has essentially no effect on solvent-exposed surface area in the human domain, a smaller ligand-induced change than observed for the bacterial ATP-AAD06727.1 complex. In the direct cross-species comparison (panel D), the triplicate-mean apo AAD06727.1 SASA (121.43 nm2) and the single-trajectory apo ABCA12 NBD2 SASA (122.33 nm2) were closely comparable, indicating that, despite their evolutionary divergence and the rmsd /R g differences noted above, the two ATP-binding domains present a similar degree of overall solvent exposure.
6.
Comparative SASA across all simulated systems. (A) Apo AAD06727.1 triplicate; (B) holo AAD06727.1 triplicate; (C) apo vs FSBA-bound ABCA12 NBD2; (D) apo AAD06727.1 (triplicate mean) vs apo ABCA12 NBD2.
Taken together, the five comparative analyses converge to a consistent picture. (i) Reproducibility: the triplicate AAD06727.1 simulations (apo and holo) show close interreplica agreement across all four global metrics, supporting the reliability of the AlphaFold2-derived starting model. (ii) Effect of ATP binding on AAD06727.1: ATP binding is associated with a small, directionally consistent reduction in R g and SASA (compaction) without an increase in rmsd or RMSF, i.e., the ligand-bound state is, if anything, marginally more compact and equally or more stable than the apo state; that is, the opposite of a destabilizing interaction. (iii) Effect of FSBA binding on human ABCA12 NBD2: essentially neutral across all four metrics, with only a minor rmsd increase and negligible R g and SASA changes, indicating that the human domain accommodates FSBA with little structural perturbation. (iv) Cross-species comparison: apo AAD06727.1 is more rigid (lower rmsd) than apo ABCA12 NBD2 but marginally less compact (higher R g), while both proteins share the same overall architecture of a stable nucleotide-binding core flanked by flexible loops/termini, and both retain their ligand within the binding pocket over the full simulation as shown by sustained (if quantitatively different) hydrogen bond networks. These integrated observations support the structural and functional plausibility of AAD06727.1 as an ATP-binding domain while also providing the first direct, side-by-side dynamical comparison against its proposed human structural counterpart, requested during peer review.
3.9. Essential Conformational Dynamics of Apo and ATP-Bound AAD06727.1
PCA was performed to characterize the dominant collective motions sampled by apo AAD06727.1 (Figure A) and ATP-bound (holo) AAD06727.1 (Figure B) during the 1000 ns molecular dynamics simulations. The PCA plots shown represent the average projection of three independent simulations generated by averaging the corresponding PC1 and PC2 coordinates at each simulation frame.
7.
PCA (PC1 vs PC2) of (A) apo AAD06727.1 and (B) holo AAD06727.1. The plots represent the averaged conformational space derived from three independent 1000 ns molecular dynamics simulations, illustrating the collective motions of the protein.
The apo AAD06727.1 exhibited a relatively broad conformational distribution within the PC1–PC2 space, indicating that the ligand-free protein sampled multiple conformational substates during the simulations (Figure A). Although the sampled conformations occupied a continuous region of the conformational landscape, no evidence of abrupt transitions or isolated conformational clusters was observed, suggesting stable exploration of the accessible structural space throughout the simulations. However, the holo AAD06727.1 displayed a comparatively more confined distribution within the principal component space (Figure B). The reduced spread of the conformational ensemble indicates that ATP binding restricted the large-scale collective motions of the nucleotide-binding domain, resulting in a more structurally constrained conformational landscape while maintaining dynamic flexibility necessary for nucleotide recognition.
The overall conformational sampling observed in both systems was continuous throughout the simulations, as illustrated by the gradual frame-dependent color transitions across the PCA plots. Together with the rmsd, RMSF, radius of gyration, and SASA analyses, these findings indicate that ATP binding stabilizes the global conformational dynamics of AAD06727.1 without inducing major structural rearrangements, thereby supporting the structural integrity of the predicted ATP-binding pocket during long time scale molecular dynamics simulations.
4. Discussion
The identification of 1265 putative proteins underscores the significant genetic diversity and evolutionary plasticity inherent in the H. pylori J99 genome. Such genomic variation frequently arises from recombination and duplication events driven by selective pressures from the host immune response and the acidic gastric environment. The functional classification of 959 sequences via EggNOG-mapper highlights the complexity of the H. pylori proteome, emphasizing that a substantial fraction of the genome is dedicated to specialized physiological niches.
We identified 19 protein sequences belonging to the HP_OMP domain family, a finding of particular significance, given that OMPs serve as the primary interface between H. pylori and the human host. These proteins mediate critical functions, including the secretion of virulence factors and the maintenance of membrane integrity under environmental stress. The presence of these conserved domains suggests that H. pylori employs a sophisticated repertoire of surface-exposed proteins to ensure long-term colonization and evade host-mediated clearance. These identified OMPs represent high-priority candidates for further investigation as potential vaccine targets and diagnostic markers.
Network analysis identified key hub complexes, including Radical_SAM, Fer4_8, and Response_reg, which function as high-degree star nodes coordinating with partners such as SPASM, CheW, and LUD_dom. These hubs correspond to maximum priority scores (5.0), reflecting their central roles in metabolic and signaling pathways. The radical SAM superfamily (GO:0003824, GO:0051536), interfacing with 16 distinct domains, likely coordinates radical-based transformations essential for bacterial persistence. Similarly, the 4Fe–4S dicluster domain (Fer4_8) supports DNA repair and oxidative stress resistance (GO:0010181) through interactions with four partners. The response regulator receiver domain (Response_reg; GO:0000160) demonstrated exceptional connectivity with 25 domains and one motif, functioning as a signaling integrator for gastric environmental cues. Structurally, the Sel1-like repeat domain (GO:0005576) acts as an extracellular mediator via its superhelical architecture, facilitating protein–protein scaffolding at the host–pathogen interface. The ABC transporter domain (ABC_tran) exhibited the most extensive network (38 domains), consistent with its indispensable role in ATPase-coupled transmembrane transport (GO:0042626) and nutrient acquisition. Collectively, these high-connectivity domains define the core functional architecture of the H. pylori proteome.
The prevalence of the Helicase_C domain (15 protein sequences) underscores the vital role of the DnaB helicase, which in H. pylori requires its C-terminal domain and specific sequence insertions for hexameric assembly and replication fork progression. The diversity of the Response_reg domain (15 sequences) highlights a sophisticated reliance on two-component systems for acid adaptation, exemplified by the essential regulator HP1043, which operates through a potentially phosphorylation-independent mechanism. , The presence of the DUF1524 domain (15 sequences) in GmrSD-like enzymes suggests a robust defense against bacteriophage invasion via the recognition and cleavage of modified DNA. The abundance of the iron–sulfur cluster domains (Fer4_8 and Fer2_3, with 13 and 12 sequences, respectively) and the radical SAM domain (12 sequences) reflects a heavy dependence on complex redox chemistry and electron transport, essential for biosynthesizing precursors such as menaquinone to withstand host-imposed oxidative stress. Additionally, the RhpA-associated DEAD domain indicates specialized RNA-processing mechanisms indispensable for colonization. Although the Bax1-I domain was identified in 12 sequences, its role is likely endogenous and regulatory, as H. pylori primarily modulates host apoptosis through external pathways rather than direct domain-mediated interactions.
The implementation of abundance and versatility metrics provides a sophisticated framework for understanding architectural complexity. High abundance typically points toward housekeeping domains or essential structural scaffolds, such as the α-helical bundles of membrane proteins. In contrast, the versatility score reflects architectural plasticity: highly versatile domains often function as nodes in signaling networks, enabling H. pylori to integrate environmental inputs into coordinated cellular responses. The observed negative relationship between domain abundance and versatility suggests a potential evolutionary trade-off between numerical expansion and combinatorial diversification in the H. pylori J99 proteome. Although the correlation was weak and did not reach statistical significance (R 2 = 0.107, p = 0.299), its direction is consistent with the hypothesis that highly abundant domains preferentially evolve through repeated duplication to reinforce specialized biological functions rather than participating in a broad range of multidomain architectures. Previous studies have shown that protein diversity is largely generated through domain duplication and recombination, with only a limited subset of domains exhibiting high combinatorial versatility across genomes. , Conversely, domains exhibiting greater versatility despite lower abundance may contribute disproportionately to functional plasticity by participating in diverse domain architectures, thereby expanding the regulatory and metabolic repertoire of the organism. Because abundance explained only a small proportion of the observed variation in versatility, additional evolutionary factors, including selective constraints, domain age, and protein interaction networks, are likely to contribute to the observed domain organization. Together, these findings suggest that the structural complexity of the H. pylori proteome is shaped by the combined effects of domain recurrence and combinatorial diversification rather than by domain abundance alone.
Although H. pylori colonizes the gastric mucosa, screening against representative human gut microbiota was performed as part of the subtractive proteomics workflow to minimize potential off-target effects on beneficial commensal microorganisms. We acknowledge that gastric microbiota-specific screening may provide a more niche-specific assessment and consider this to be a limitation of the present study. Because ABC transporter ATP-binding proteins are evolutionarily conserved, homologous proteins were detected in representative gastric commensals. However, the observed sequence identities remained below the predefined exclusion threshold, suggesting that no closely related homologues were identified according to the criteria adopted in this study. Nevertheless, sequence similarity alone cannot fully predict inhibitor selectivity, and future comparative structural and experimental studies will be required to evaluate potential off-target effects on the beneficial gastric microbiota.
The broad conservation of ABC transporter ATP-binding proteins across bacterial species represents both an opportunity and a challenge for antimicrobial drug discovery. While conservation may facilitate the development of broad-spectrum therapeutics, it may also increase the potential for unintended interactions with homologous proteins in the beneficial microbiota. In this study, comparative sequence analyses against representative gut and gastric commensal bacteria did not identify homologues exceeding the predefined exclusion threshold of 50% sequence identity. However, because sequence conservation alone does not fully determine inhibitor specificity, future studies should incorporate comparative structural analyses, binding site characterization, and experimental validation to evaluate target selectivity and potential microbiota-associated off-target effects.
Several established therapeutic targets have been extensively investigated in H. pylori, including urease, which enables survival under acidic gastric conditions, DNA gyrase involved in DNA replication, Mur enzymes required for peptidoglycan biosynthesis, and proteins associated with the Cag pathogenicity island that contribute to host colonization and virulence. These targets have formed the basis for current antibacterial strategies but are increasingly affected by the emergence of antibiotic resistance. In contrast, AAD06727.1 represents a previously uncharacterized ABC transporter ATP-binding protein that was identified through an integrative computational prioritization strategy rather than prior biological characterization. The protein fulfilled multiple desirable target-selection criteria, including predicted essentiality, association with virulence, broad-spectrum conservation, absence of significant homology to human proteins and gut microbiota, and the presence of a structurally conserved ATP-binding pocket. These complementary characteristics suggest that AAD06727.1 constitutes a promising candidate for further experimental validation, alongside established H. pylori drug targets.
The selection of AAD06727.1 as a primary target highlights the efficacy of combining domain-based prioritization with stringent homology filtering. The high degree of nonhomology between AAD06727.1 and the human proteome suggests high therapeutic specificity, while the absence of homology with the human gut microbiome reduces the risk of dysbiosis. The structural predominance of the ABC_tran domain in AAD06727.1 underscores its metabolic importance. Our DDI profiling revealed that this domain participates in 38 distinct interaction interfaces, identifying it as a critical functional hub. The 40% sequence identity to Candida TEF3 is noteworthy; while TEF3 is a fungal translation factor, the conserved ATP-binding motifs suggest a shared susceptibility to small-molecule inhibitors targeting the nucleotide-binding pocket. Given that AAD06727.1 lacks human orthologs, these binding sites represent an evolutionarily constrained vulnerability. Our MD trajectory confirms that the Gly40–Lys41–Thr42 motif (P-loop) maintains a rigid configuration, which is essential for the energetic coupling required for substrate translocation. ABC transporter ATP-binding proteins have been extensively investigated as antibacterial targets in numerous bacterial pathogens because they provide the energy required for substrate transport and are frequently essential for cellular physiology. Therefore, the novelty of the present study does not arise from the ABC transporter family itself, but from the identification and comprehensive characterization of AAD06727.1, a previously uncharacterized H. pylori J99 protein that was originally annotated as a hypothetical protein. Unlike established H. pylori therapeutic targets, including urease, DNA gyrase, Mur enzymes, and CagA-associated virulence pathways, AAD06727.1 was prioritized through an integrative subtractive proteomics and structural biology workflow. Its predicted ATP-binding domain, broad-spectrum conservation, high-confidence structural model, and favorable druggability collectively support its potential as a promising candidate for future inhibitor development.
The docking analyses consistently identified FSBA as the most favorable ligand for the predicted ATP-binding pocket of AAD06727.1, suggesting that the nucleotide-binding domain retains the structural characteristics required for the recognition of ATP analogues. FSBA was originally developed as a reactive ATP analogue that covalently modifies conserved residues within the nucleotide-binding domains of ATP-dependent proteins, including ABC transporters, and therefore serves as a valuable probe for investigating ATP-binding sites rather than as a conventional inhibitor. The stronger predicted affinity of FSBA toward AAD06727.1 compared to the human ABCA12 NBD2, despite the overall conservation of the ATP-binding fold, suggests that differences in the local binding environment may be exploited for the development of selective bacterial inhibitors.
Docking to the ATP-equilibrated holostructure further demonstrated that ATP occupancy reduced the predicted affinity of FSBA while preserving ligand recognition within the nucleotide-binding pocket. Such behavior is consistent with the ATP-dependent conformational transitions characteristic of ABC transporter nucleotide-binding domains, in which ATP binding remodels the local architecture of the binding pocket without disrupting its functional integrity. The LigPlot+ interaction analyses further support these observations by showing that FSBA maintained an extensive hydrogen-bonding and hydrophobic interaction network in both apo- and holo-conformations, indicating that the predicted binding site remains structurally accessible despite ATP-induced conformational rearrangements. Collectively, these findings support the structural validity of the predicted ATP-binding pocket and provide a rationale for considering ATP-competitive molecules as potential starting points for the structure-guided development of inhibitors targeting AAD06727.1.
ABC transporters play fundamental roles in bacterial physiology by coupling ATP hydrolysis to the active transport of a wide variety of substrates required for cellular homeostasis, nutrient acquisition, and environmental adaptation. In Gram-negative bacteria, ATP-binding proteins provide the energy necessary for substrate translocation across the membrane through interactions with their cognate transmembrane permease components. Consequently, disruption of the ATPase activity can impair transporter function, compromise nutrient uptake, and reduce bacterial fitness, particularly under nutrient-limited conditions commonly encountered during host colonization. , On the basis of its predicted annotation and conserved ATP-binding architecture, AAD06727.1 is proposed to function as the nucleotide-binding component of an ABC transporter in H. pylori.
Long time scale molecular dynamics simulations were performed to evaluate whether the predicted ATP-binding fold of AAD06727.1 remains structurally stable under physiological conditions and to determine the influence of ATP binding on its conformational dynamics. Independent triplicate simulations demonstrated highly reproducible structural behavior for both the apo- and ATP-bound conformations, with all trajectories rapidly reaching equilibrium and maintaining stable backbone deviations throughout the 1000 ns simulations. The reproducibility across independent replicas indicates that the observed conformational behavior represents an intrinsic property of the protein rather than stochastic variation arising from individual trajectories.
Backbone rmsd, RMSF, radius of gyration, and SASA analyses consistently showed that AAD06727.1 preserves its overall tertiary structure during simulations. The majority of structural fluctuations were confined to solvent-exposed loops and terminal regions, whereas residues constituting the predicted nucleotide-binding pocket remained comparatively rigid. Such localized flexibility, accompanied by preservation of a stable catalytic core, is a characteristic feature of nucleotide-binding domains of ABC transporters and is required to accommodate ATP binding while maintaining structural integrity necessary for catalytic function. ,
Comparison of the apo- and ATP-bound conformations demonstrated that ATP binding did not induce major structural rearrangements. Instead, the ATP-bound structure exhibited slightly lower rmsd, radius of gyration, and solvent-accessible surface area values than those of the apo protein, indicating a modest increase in structural compactness following nucleotide binding. Similar ligand-induced stabilization has been reported for ATP-binding cassette proteins, where ATP binding primarily reinforces local interactions within the nucleotide-binding domain, while larger conformational transitions occur only after communication with the transmembrane domains during the transport cycle. , The stable intermolecular hydrogen-bond network observed throughout the ATP simulations further supports persistent nucleotide recognition, despite continuous thermal fluctuations expected in explicit-solvent molecular dynamics simulations.
PCA provided additional evidence that ATP binding influences the global conformational landscape of AAD06727.1. Whereas the apo protein sampled a broader conformational space, the ATP-bound complex occupied a comparatively restricted region of the PC1–PC2 landscape, indicating reduced large-scale collective motions upon ligand binding. Restriction of conformational sampling following ligand association has frequently been observed during molecular dynamics simulations and reflects stabilization of energetically favorable conformational substates rather than loss of essential protein flexibility. ,
Comparison with the closest human structural homologue, ABCA12 NBD2, further demonstrated that both proteins maintain highly stable ATP-binding folds throughout the simulations despite their evolutionary divergence. Although the human nucleotide-binding domain displayed slightly greater backbone flexibility than AAD06727.1, both proteins exhibited comparable RMSF, R g, and SASA profiles, indicating the preservation of their overall structural architectures. The observed conservation of dynamic behavior is consistent with the highly conserved structural organization of ABC transporter nucleotide-binding domains across prokaryotic and eukaryotic systems despite substantial sequence divergence.
Collectively, these molecular dynamics simulations demonstrate that the predicted ATP-binding pocket of AAD06727.1 remains structurally stable in both ligand-free and ATP-bound conformations over microsecond time scales. Importantly, the simulations are not presented as evidence of therapeutic efficacy but rather as validation that the predicted nucleotide-binding site remains conformationally stable and structurally accessible for ligand recognition. When integrated with the subtractive proteomics pipeline, structural modeling, comparative docking, and target prioritization analyses, these findings provide a robust structural framework supporting AAD06727.1 as a promising candidate for future experimental characterization and structure-guided inhibitor development.
As illustrated in Figure , the proposed model summarizes the potential physiological consequences of AAD06727.1 activity. Under functional conditions, ATP hydrolysis by AAD06727.1 is expected to energize substrate transport, thereby supporting the intracellular nutrient availability and cellular homeostasis. Conversely, impairment of ATP binding or hydrolysis would be expected to reduce transporter activity, limiting substrate import, and ultimately affecting bacterial growth and survival. This schematic represents a mechanistic hypothesis derived from the computational findings of the present study and the established transport mechanism of bacterial ABC transporters rather than direct experimental evidence for the transport of any specific substrate by AAD06727.1.
8.
Functional impact of AAD06727.1 on bacterial nutrient acquisition and viability. Comparative model of the H. pylori ABC Transporter: impact of AAD06727.1’s functional integrity on nutrient acquisition and cellular viability.
4.1. Limitations of the Study
The present study is based on an integrated computational target prioritization framework. Gene essentiality was inferred through homology to experimentally validated entries in the DEG, an approach widely employed in subtractive proteomics but one that may not fully capture strain-specific essentiality in H. pylori J99. Furthermore, although molecular docking, comparative structural analyses, and triplicate molecular dynamics simulations provide strong computational evidence supporting ligand recognition, conformational stability, and structural suitability of the predicted ATP-binding pocket, these findings do not constitute experimental validation. In addition, molecular dynamics simulations were performed for the apo- and ATP-bound conformations of AAD06727.1, whereas simulations of inhibitor–bound complexes were not included and therefore await future investigation. Consequently, AAD06727.1 should be regarded as a computationally prioritized therapeutic target whose predicted essentiality and druggability require confirmation through experimental approaches, including targeted gene disruption or conditional knockdown, ligand binding studies, ATPase activity measurements, and functional inhibition assays before its therapeutic potential can be conclusively established.
5. Conclusion
This study presents an integrated computational framework for the discovery and structural characterization of novel therapeutic targets in H. pylori. Starting from Rosetta Stone-mediated identification of putative proteins, a comprehensive workflow incorporating functional reannotation, subtractive proteomics, structural modeling, comparative analyses, molecular docking, and long-time scale MD simulations was employed to prioritize AAD06727.1 as a promising ATP-binding protein with characteristics desirable for antimicrobial drug development. Comparative docking with ATP and selected ATP analogues, together with apo and ATP-bound molecular dynamics simulations, demonstrated that the predicted nucleotide-binding pocket remained structurally stable and capable of ligand recognition across different conformational states. Structural comparison with the human ABCA12 nucleotide-binding domain further supported conservation of the ATP-binding fold while revealing sufficient sequence divergence that may facilitate the rational design of selective inhibitors targeting the bacterial protein while minimizing potential off-target interactions with the human homologue. Collectively, these findings identify AAD06727.1 as a computationally prioritized target that warrants further experimental investigation. Future studies should focus on biochemical characterization of its ATPase activity, validation of ligand binding through biophysical assays, functional assessment by genetic approaches, and high-resolution structural determination to support structure-guided inhibitor development. More broadly, the computational strategy presented here provides a reproducible framework for discovering and prioritizing antimicrobial targets in H. pylori and other bacterial pathogens, thereby facilitating the development of new therapeutic strategies to address the growing challenge of antibiotic resistance.
Supplementary Material
Acknowledgments
The authors acknowledge Nitte (Deemed to be University) for providing the necessary research facilities. The authors also thank Dr. Karthik Raman, Department of Data Science and AI, Wadhwani School of Data Science and AI, IIT Madras, Chennai, for valuable technical input and critical review of the manuscript.
The data supporting the findings of this study are available within the article and its Supporting Information files. Additional data are available from the corresponding author upon reasonable request.
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.6c05087.
Domain architecture and interaction analyses, statistical correlation analysis, subtractive proteomics workflow, broad-spectrum conservation, druggability assessment, AlphaFold2 structural prediction and validation, structural comparison with human ABCA12 NBD2, LigPlot+ interaction diagrams for all docked complexes, ATP-bound holo equilibration, AutoDock Vina docking parameters, domain annotation data sets, subtractive proteomics screening results, comparative target prioritization, and residue-wise AlphaFold2 pLDDT confidence scores for AAD06727.1 (PDF)
M.K.: Conceptualization, methodology, formal analysis, investigation, and writingoriginal draft. P.G.: Conceptualization, methodology, formal analysis, investigation, writingoriginal draft, writingreview and editing, and supervision. H.S.S.K.: Interpretation of results, writingreview and editing, and supervision. K.K.B.: Writingreview and editing.V.A.S.: Writingreview and editing. All authors have read and approved the final manuscript.
This work was supported by the Anusandhan National Research Foundation (ANRF), India, under the TARE Scheme (ref. No. TAR/2022/000231).
Not applicable, as this study does not involve human participants, animals, or clinical data.
The authors declare no competing financial interest.
References
- Alm R. A., Bina J., Andrews B. M.. et al. Comparative genomics of Helicobacter pylori: analysis of the outer membrane protein families. Infect. Immun. 2000;68(7):4155–4168. doi: 10.1128/IAI.68.7.4155-4168.2000. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bauwens E., Joosten M., Taganna J., Rossi M., Debraekeleer A., Tay A., Peters F., Backert S., Fox J., Ducatelle R.. et al. In silico proteomic and phylogenetic analysis of the outer membrane protein repertoire of gastric Helicobacter species. Sci. Rep. 2018;8(1):15453. doi: 10.1038/s41598-018-32476-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sugiyama N., Miyake S., Lin M. H.. et al. Comparative proteomics of Helicobacter pylori strains reveals geographical features rather than genomic variations. Genes Cells. 2019;24(2):139–150. doi: 10.1111/gtc.12662. [DOI] [PubMed] [Google Scholar]
- Veitia R. A.. Rosetta Stone proteins: “chance and necessity”? Genome Biol. 2002;3(2):INTERACTIONS1001. doi: 10.1186/gb-2002-3-2-interactions1001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gollapalli P., Rudrappa S., Kumar V., Santosh Kumar H. S.. Domain architecture based methods for comparative functional genomics toward therapeutic drug target discovery. J. Mol. Evol. 2023;91(5):598–615. doi: 10.1007/s00239-023-10129-w. [DOI] [PubMed] [Google Scholar]
- Cantalapiedra C. P., Hernández-Plaza A., Letunic I., Bork P., Huerta-Cepas J.. eggNOG-mapper v2: Functional annotation, orthology assignments, and domain prediction at the metagenomic scale. Mol. Biol. Evol. 2021;38(12):5825–5829. doi: 10.1093/molbev/msab293. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kumar H. S., Kumar V.. A comparison of Rosetta Stones in adapter protein families. Bioinformation. 2016;12(5):285–292. doi: 10.6026/97320630012285. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chakraborty S., Askari M., Barai R. S., Idicula-Thomas S.. PBITV3: A robust and comprehensive tool for screening pathogenic proteomes for drug targets and prioritizing vaccine candidates. Protein Sci. 2024;33(2):e4892. doi: 10.1002/pro.4892. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chakkyarath V., Shanmugam A., Natarajan J.. Prioritization of potential drug targets and antigenic vaccine candidates against Klebsiella aerogenes using the computational subtractive proteome-driven approach. J. Proteins Proteom. 2021;12(3):201–211. doi: 10.1007/s42485-021-00068-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Narthanareeswaran B., Hemavathy N., Ranganathan S.. et al. Integrative genomics and structural bioinformatics uncovers AMR-associated drug targets and pqsH inhibitors in multidrug-resistant Pseudomonas aeruginosa JJPA01. Mol. Divers. 2026;30:4171. doi: 10.1007/s11030-025-11365-6. [DOI] [PubMed] [Google Scholar]
- Jones J., Murphy C. P., Sleator R. D., Culligan E. P.. An exploratory in silico analysis of bacteriocin gene clusters in the urobiome. Microbiome Res. Rep. 2024;3(2):24. doi: 10.20517/mrr.2023.78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Shakibania T., Arabfard M., Najafi A.. A predictive approach for host-pathogen interactions using deep learning and protein sequences. VirusDisease. 2024;35(3):434–445. doi: 10.1007/s13337-024-00882-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Southan C., Sitzmann M., Muresan S.. Comparing the chemical structure and protein content of ChEMBL, DrugBank, Human Metabolome Database and the Therapeutic Target Database. Mol. Inform. 2013;32(11–12):881–897. doi: 10.1002/minf.201300103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Hallgren J., Tsirigos K. D., Pedersen M. D., Almagro Armenteros J. J., Marcatili P., Nielsen H., Winther O.. DeepTMHMM predicts alpha and beta transmembrane proteins using deep neural networks. bioRxiv. 2022:2022.04.08.487609. doi: 10.1101/2022.04.08.487609. [DOI] [Google Scholar]
- Jumper J., Evans R., Pritzel A.. et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583–589. doi: 10.1038/s41586-021-03819-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Heo L., Park H., Seok C.. GalaxyRefine: Protein structure refinement driven by side-chain repacking. Nucleic Acids Res. 2013;41(W1):W384–W388. doi: 10.1093/nar/gkt458. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Anderson R. J., Weng Z., Campbell R. K., Jiang X.. Main-chain conformational tendencies of amino acids. Proteins. 2005;60(4):679–689. doi: 10.1002/prot.20530. [DOI] [PubMed] [Google Scholar]
- Polák L., Škoda P., Riedlová K., Krivák R., Novotný M., Hoksza D.. PrankWeb 4: a modular web server for protein-ligand binding site prediction and downstream analysis. Nucleic Acids Res. 2025;53(W1):W466–W471. doi: 10.1093/nar/gkaf421. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Rigsby R. E., Parker A. B.. Using the PyMOL application to reinforce visual understanding of protein structure. Biochem. Mol. Biol. Educ. 2016;44(5):433–437. doi: 10.1002/bmb.20966. [DOI] [PubMed] [Google Scholar]
- Ohnuma S., Chufan E., Nandigama K., Jenkins L. M., Durell S. R., Appella E., Sauna Z. E., Ambudkar S. V.. Inhibition of multidrug resistance-linked P-glycoprotein (ABCB1) function by 5′-fluorosulfonylbenzoyl 5′-adenosine: evidence for an ATP analogue that interacts with both drug-substrate-and nucleotide-binding sites. Biochemistry. 2011;50(18):3724–3735. doi: 10.1021/bi200073f. [DOI] [PMC free article] [PubMed] [Google Scholar]
- El-Hachem N., Haibe-Kains B., Khalil A., Kobeissy F. H., Nemer G.. AutoDock and AutoDockTools for Protein-Ligand Docking: Beta-Site Amyloid Precursor Protein Cleaving Enzyme 1(BACE1) as a Case Study. Methods Mol. Biol. 2017;1598:391–403. doi: 10.1007/978-1-4939-6952-4_20. [DOI] [PubMed] [Google Scholar]
- Trott O., Olson A. J.. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J. Comput. Chem. 2010;31(2):455–461. doi: 10.1002/jcc.21334. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Laskowski R. A., Swindells M. B.. LigPlot+: multiple ligand-protein interaction diagrams for drug discovery. J. Chem. Inf. Model. 2011;51(10):2778–2786. doi: 10.1021/ci200227u. [DOI] [PubMed] [Google Scholar]
- Abraham M. J., Murtola T., Schulz R., Páll S., Smith J. C., Hess B., Lindahl E.. GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX. 2015;1:19–25. doi: 10.1016/j.softx.2015.06.001. [DOI] [Google Scholar]
- Qi Y., Cheng X., Han W., Jo S., Schulten K., Im W.. CHARMM-GUI PACE CG Builder for solution, micelle, and bilayer coarse-grained simulations. J. Chem. Inf. Model. 2014;54(3):1003–1009. doi: 10.1021/ci500007n. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Seeliger D., de Groot B. L.. Conformational transitions upon ligand binding: Holo structure prediction from apo conformations. PLoS Comput. Biol. 2010;6(1):e1000634. doi: 10.1371/journal.pcbi.1000634. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Nitharwal R. G., Paul S., Dar A.. et al. The domain structure of Helicobacter pylori DnaB helicase: the N-terminal domain can be dispensable for helicase activity whereas the extreme C-terminal region is essential for its function. Nucleic Acids Res. 2007;35(9):2861–2874. doi: 10.1093/nar/gkm167. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Machnicka M. A., Kaminska K. H., Dunin-Horkawicz S., Bujnicki J. M.. Phylogenomics and sequence-structure-function relationships in the GmrSD family of Type IV restriction enzymes. BMC Bioinf. 2015;16:336. doi: 10.1186/s12859-015-0773-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Benoit S. L., Holland A. A., Johnson M. K., Maier R. J.. Iron-sulfur protein maturation in Helicobacter pylori: identifying a Nfu-type cluster carrier protein and its iron-sulfur protein targets. Mol. Microbiol. 2018;108(4):379–396. doi: 10.1111/mmi.13942. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Tomb J. F., White O., Kerlavage A. R.. et al. The complete genome sequence of the gastric pathogen Helicobacter pylori. Nature. 1997;388(6642):539–547. doi: 10.1038/41483. [DOI] [PubMed] [Google Scholar]
- Bourret R. B.. Receiver domain structure and function in response regulator proteins. Curr. Opin. Microbiol. 2010;13(2):142–149. doi: 10.1016/j.mib.2010.01.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Schär J., Sickmann A., Beier D.. Phosphorylation-independent activity of atypical response regulators of Helicobacter pylori. J. Bacteriol. 2005;187(9):3100–3109. doi: 10.1128/JB.187.9.3100-3109.2005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lee B., Lee D.. Protein comparison at the domain architecture level. BMC Bioinf. 2009;10(S15):S5. doi: 10.1186/1471-2105-10-s15-s5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Forslund K., Henricson A., Hollich V., Sonnhammer E. L.. Domain tree-based analysis of protein architecture evolution. Mol. Biol. Evol. 2008;25(2):254–264. doi: 10.1093/molbev/msm254. [DOI] [PubMed] [Google Scholar]
- Walker J. E., Saraste M., Runswick M. J., Gay N. J.. Distantly related sequences in the alpha- and beta-subunits of ATP synthase, myosin, kinases and other ATP-requiring enzymes and a common nucleotide binding fold. EMBO J. 1982;1(8):945–951. doi: 10.1002/j.1460-2075.1982.tb01276.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Apic G., Gough J., Teichmann S. A.. Domain combinations in archaeal, eubacterial and eukaryotic proteomes. J. Mol. Biol. 2001;310(2):311–325. doi: 10.1006/jmbi.2001.4776. [DOI] [PubMed] [Google Scholar]
- Vogel C., Bashton M., Kerrison N. D., Chothia C., Teichmann S. A.. Structure, function and evolution of multidomain proteins. Curr. Opin. Struct. Biol. 2004;14(2):208–216. doi: 10.1016/j.sbi.2004.03.011. [DOI] [PubMed] [Google Scholar]
- Pace C. N., Fu H., Lee Fryar K.. et al. Contribution of hydrogen bonds to protein stability. Protein Sci. 2014;23(5):652–661. doi: 10.1002/pro.2449. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Smith P. C., Karpowich N., Millen L., Moody J. E., Rosen J., Thomas P. J., Hunt J. F.. ATP binding to the motor domain from an ABC transporter drives formation of a nucleotide sandwich dimer. Mol. Cell. 2002;10(1):139–149. doi: 10.1016/S1097-2765(02)00576-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Oldham M. L., Chen J.. Snapshots of the maltose transporter during ATP hydrolysis. Proc. Natl. Acad. Sci. U.S.A. 2011;108(37):15152–15156. doi: 10.1073/pnas.1108858108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Davidson A. L., Dassa E., Orelle C., Chen J.. Structure, function, and evolution of bacterial ATP-binding cassette systems. Microbiol. Mol. Biol. Rev. 2008;72(2):317–364. doi: 10.1128/MMBR.00031-07. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Rees D. C., Johnson E., Lewinson O.. ABC transporters: the power to change. Nat. Rev. Mol. Cell Biol. 2009;10(3):218–227. doi: 10.1038/nrm2646. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Locher K. P.. Mechanistic diversity in ATP-binding cassette (ABC) transporters. Nat. Struct. Mol. Biol. 2016;23(6):487–493. doi: 10.1038/nsmb.3216. [DOI] [PubMed] [Google Scholar]
- Amadei A., Linssen A. B., Berendsen H. J.. Essential dynamics of proteins. Proteins. 1993;17(4):412–425. doi: 10.1002/prot.340170408. [DOI] [PubMed] [Google Scholar]
- David C. C., Jacobs D. J.. Principal component analysis: a method for determining the essential dynamics of proteins. Methods Mol. Biol. 2014;1084:193–226. doi: 10.1007/978-1-62703-658-0_11. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data supporting the findings of this study are available within the article and its Supporting Information files. Additional data are available from the corresponding author upon reasonable request.