Structural and systems-level analysis of polyphenolic scaffold compatibility at immune checkpoint interfaces: a PD-L1 dimer–focused in silico study

Structural and systems-level analysis of polyphenolic scaffold compatibility at immune checkpoint interfaces: a PD-L1 dimer–focused in silico study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Structural and systems-level analysis of polyphenolic scaffold compatibility at immune checkpoint interfaces: a PD-L1 dimer–focused in silico study Muhammad Rafif Pratama Putra Shanizal This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8841071/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Context Protein–protein interaction interfaces of immune checkpoints present persistent challenges for molecular modeling due to shallow topology, limited pocket definition, and conformational flexibility. Although polyphenolic compounds have been widely explored in immune-related computational studies, their structural compatibility with immune checkpoint interface architectures remains poorly defined at the scaffold level. In this study, we investigated interface-focused structural compatibility of representative polyphenolic scaffolds at immune checkpoint proteins, with primary emphasis on the programmed death-ligand 1 (PD-L1) dimer interface and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) included as a structural comparator. The analysis reveals reproducible pose convergence and non-random residue footprint overlap at the PD-L1 dimer interface, supported by stable intrinsic interface architecture during apo-state molecular dynamics simulations. Network-based analysis further situates predicted ligand-associated targets within immune-related interaction neighborhoods, providing systems-level context consistent with a structurally permissive signaling environment. These findings characterize interface-level structural compatibility rather than functional immune checkpoint inhibition and generate testable hypotheses for subsequent experimental studies. Methods Interface-preserving molecular docking was performed using AutoDock Vina with multi-seed sampling to assess spatial compatibility, pose convergence, and residue-level footprint overlap at immune checkpoint interfaces. Docking validation included redocking benchmarks and decoy-based evaluation. Molecular dynamics simulations of the apo PD-L1 dimer were conducted using GROMACS with a classical all-atom force field to characterize intrinsic interface stability, residue flexibility, interfacial contacts, and hydrogen-bond persistence. Predicted molecular targets of representative polyphenols were identified using similarity-based target prediction tools and analyzed through protein–protein interaction network construction and topological analysis using Cytoscape. All computational workflows were executed using standard molecular modeling and network analysis software, with full methodological details provided in the main text and Online Resource 1. Biological sciences/Computational biology and bioinformatics Biological sciences/Drug discovery PD-L1 dimer Polyphenols Network pharmacology protein–protein interaction docking interface-focused docking scaffold compatibility Figures Figure 1 Figure 2 Figure 3 Figure 4 1. INTRODUCTION 1.1 Immune checkpoints in cancer immunotherapy Immune checkpoint pathways regulate the magnitude and duration of T-cell activation, maintaining immune homeostasis and preventing excessive immune responses that may lead to tissue damage or autoimmunity. Among the best-characterized immune checkpoints, cytotoxic T-lymphocyte–associated protein 4 (CTLA-4) and programmed death-ligand 1 (PD-L1) act as dominant regulators of antitumor immunity through distinct yet complementary mechanisms.[ 1 ] Despite their biological relevance, immune checkpoint protein–protein interaction interfaces pose persistent challenges for molecular modeling due to shallow topology, conformational flexibility, and limited canonical binding pockets. CTLA-4 primarily regulates early T-cell priming within secondary lymphoid organs by competing with CD28 for the costimulatory ligands CD80 and CD86, thereby attenuating initial activation signals. CTLA-4 was therefore included not as a primary target, but as a structurally distinct immune checkpoint interface to enable comparative assessment of scaffold compatibility across divergent protein–protein interaction architectures. In contrast, PD-L1 suppresses effector T-cell function in peripheral tissues through engagement with programmed death-1 (PD-1), contributing to immune tolerance within the tumor microenvironment.[ 1 ], [ 2 ], [ 3 ] Therapeutic blockade of CTLA-4 and the PD-1/PD-L1 axis using monoclonal antibodies has fundamentally altered cancer treatment, yielding durable clinical responses across multiple malignancies.[ 2 ], [ 4 ] Nevertheless, antibody-based immune checkpoint inhibitors are frequently associated with immune-related adverse events, high production costs, and limited tissue penetration. These challenges have motivated exploration of complementary strategies that expand immune checkpoint modulation beyond biologic therapeutics.[ 5 ], [ 6 ] 1.2 Small-molecule modulation of immune checkpoints Small molecules represent an attractive alternative or adjunct to antibody-based therapies due to their oral bioavailability, scalable synthesis, and tunable pharmacokinetic profiles. However, immune checkpoint proteins pose substantial challenges for small-molecule targeting because their functional regions are dominated by relatively flat and solvent-exposed protein–protein interaction interfaces rather than deep catalytic pockets.[ 7 ] Advances in protein–protein interaction drug discovery have demonstrated that such interfaces can be engaged by small molecules through distributed binding mechanisms involving aromatic stacking, hydrophobic surface complementarity, and electrostatic anchoring rather than classical lock-and-key interactions.[ 8 ], [ 9 ] Within this framework, polyphenolic compounds have emerged as structurally relevant chemical scaffolds, as their extended aromatic systems and multiple hydrogen-bonding functionalities are well suited for interaction with shallow protein surfaces rather than classical active-site pockets.[ 10 ], [ 11 ] In this context, polyphenols are considered as chemically diverse scaffolds for probing interface permissiveness rather than as candidate therapeutic inhibitors. Emerging structural and computational studies further indicate that PD-L1 can adopt oligomeric assemblies that expose ligand-accessible interface pockets not apparent in monomeric conformations. Small molecules binding at the PD-L1 dimer interface have been shown to stabilize alternative protein arrangements and indirectly interfere with PD-1 recognition through steric or conformational mechanisms.[ 12 ], [ 13 ], [ 14 ] From a molecular modeling perspective, the PD-L1 dimer interface remains comparatively underexplored relative to monomeric binding sites, particularly with respect to scaffold-level interface compatibility. These findings highlight receptor oligomeric state as a critical determinant of interface-level ligand accommodation and underscore the importance of preserving native quaternary structure in computational evaluation of immune checkpoint interactions. 1.3 Rationale and objective of the study Despite increasing interest in small-molecule engagement of immune checkpoint interfaces, many computational studies remain focused on single targets or monomeric protein structures, potentially overlooking interface-specific binding opportunities relevant to immune regulation. In particular, systematic evaluation of compounds capable of engaging both CTLA-4 and PD-L1 within a unified structural and systems-level framework remains limited. Unlike many prior studies that focus on monomeric PD-L1 structures or single-target affinity ranking, this work preserves the PD-L1 dimer architecture and evaluates interface-level compatibility across two immune checkpoints within a unified framework. The present study addresses this gap by applying an integrated in silico strategy to examine selected polyphenolic compounds as interface-compatible scaffolds in relation to two major immune checkpoints. Molecular docking was performed against CTLA-4 and the PD-L1 homodimer to assess ligand accommodation at structurally relevant surface cavities. Molecular dynamics–based structural characterization was incorporated to describe the intrinsic integrity of the PD-L1 dimer interface, providing dynamic context for interface-focused docking interpretation. In parallel, network-based analyses were conducted to examine convergence of predicted molecular targets on immune-related signaling pathways.[ 15 ], [ 16 ] Despite extensive computational exploration of immune checkpoint-related ligands, few studies have explicitly addressed interface-preserving modeling strategies that prioritize spatial compatibility, residue-level footprint convergence, and intrinsic interface stability over affinity-based ranking. As a result, scaffold-level permissiveness at immune checkpoint protein–protein interfaces remains poorly defined from a structural modeling perspective. Specifically, this study aims to characterize the docking behavior of selected polyphenols toward CTLA-4 and PD-L1, assess the influence of PD-L1 dimeric architecture on interface-level ligand accommodation, and integrate structure-based docking results with network analyses to explore pathway-level relevance of dual immune checkpoint engagement. Through this approach, the study provides a structural and systems-level rationale for dual immune checkpoint interface compatibility while clearly distinguishing computational insight from functional or therapeutic validation. 2. MATERIALS AND METHODS All computational analyses were conducted using a combination of Windows 11 and Ubuntu Linux environments. Molecular docking was performed using AutoDock Vina via PyRx, molecular dynamics simulations were carried out using GROMACS, and network analyses were performed using STRING and Cytoscape. Detailed software versions and computational environment specifications are provided in Online Resource 1. 2.1 Protein structure preparation Three-dimensional structures of immune checkpoint proteins were obtained from the Protein Data Bank (PDB), a curated repository of experimentally determined biomolecular structures.[ 17 ] The extracellular domain of cytotoxic T-lymphocyte–associated protein 4 (CTLA-4) was retrieved under PDB ID 1I85 at a resolution of 2.3 Å, while programmed death-ligand 1 (PD-L1) was obtained under PDB ID 5J89 at a resolution of 2.45 Å. Both proteins crystallize as homodimers. For CTLA-4, chain A was selected for docking analysis, as the reported ligand-accessible surface is contained within a single protomer. In contrast, PD-L1 chains A and B were retained to preserve the dimeric architecture, which exposes an inter-protomer interface pocket previously reported to accommodate small-molecule binding.[ 13 ], [ 14 ] CTLA-4 was included as an architectural comparator rather than as a primary inhibitory target. Docking to CTLA-4 was performed to assess whether polyphenolic scaffolds capable of accommodating the PD-L1 dimer interface also exhibit compatibility with a second immune checkpoint surface. The analyzed CTLA-4 site represents an interface-adjacent permissive surface, enabling cross-interface comparison without implying direct blockade of the B7-1 binding site.[ 3 ] Protein preparation was performed using AutoDockTools following standard protocols.[ 18 ] Detailed preparation steps are provided in Online Resource 1. 2.2 Ligand preparation Six polyphenolic compounds were selected based on structural diversity and reported immunomodulatory relevance: gnetin C, quercetin, resveratrol, kaempferol, epigallocatechin gallate (EGCG), and curcumin. Ligand structures were retrieved from the PubChem database in SDF format.[ 19 ] Ligand structures were prepared and energy-minimized using standard workflows. Full ligand preparation procedures are described in Online Resource 1. 2.3 Binding pocket identification and molecular docking Potential ligand-binding cavities were identified using CB-Dock2, an automated docking platform that detects protein cavities and estimates optimal grid centers based on structural topology.[ 20 ] For both CTLA-4 and PD-L1, the highest-ranked cavity was selected for subsequent docking analysis. Docking simulations were performed using AutoDock Vina implemented through PyRx.[ 21 ] Docking parameters were set as follows: exhaustiveness = 24, number of output binding modes = 9, and the AutoDock Vina v1.2.3 scoring function applied. All docking calculations were performed using a fixed random seed to ensure reproducibility across independent docking runs.[ 21 ], [ 22 ] Grid centers and dimensions for CTLA-4 and PD-L1 docking were defined based on CB-Dock2 cavity detection and are provided in Online Resource 1. For robustness assessment, each ligand was re-docked using five independent random seeds while keeping the interface-focused grid and all other docking parameters unchanged. For each ligand, mode-1, corresponding to the lowest predicted binding free energy, was selected for further analysis. 2.3.1 Ligand selection for comparative analysis Ligand prioritization was guided by docking-derived evidence rather than binding affinity alone. All six polyphenols were initially docked against CTLA-4 and PD-L1 using identical protocols. Gnetin C consistently exhibited the most favorable binding energies and reproducible top-ranked docking poses across both targets. Quercetin and resveratrol were retained for comparative analysis based on reproducible docking poses and consistent occupation of the primary binding pocket across independent docking runs. Importantly, these ligands occupied overlapping regions of the CTLA-4 and PD-L1 binding cavities, supporting comparable binding modes rather than nonspecific or peripheral interactions.[ 18 ], [ 22 ] For each ligand–target pair, multiple docking poses were generated in a single docking run, and all output poses were retained for comparative analysis. Binding energies reported in the main text correspond to the top-ranked pose (mode-1), while mean binding energies across all generated poses are provided in Online Resource 1 to reflect overall pose stability rather than single-score optimization. 2.4 Docking result validation Docking validation was performed using a combination of pose reproducibility analysis, benchmark redocking of a co-crystallized PD-L1 interface ligand, and decoy-based enrichment testing to assess scoring robustness. Pose reproducibility was evaluated across independent docking runs using RMSD-based comparison. Detailed procedures are described in Online Resource 1. Given the shallow and solvent-exposed nature of protein–protein interaction interfaces, RMSD values in the range of ~ 2–3 Å were considered acceptable.[ 23 ], [ 24 ] 2.4.1 Interpretation boundaries of docking validation All docking validation procedures were intended to support structural plausibility and reproducibility of docking placement rather than prediction of binding affinity, binding persistence, or functional inhibition.[ 7 ], [ 22 ] 2.5 Interaction analysis Ligand–protein interactions were analyzed using standard visualization tools. Detailed interaction classification criteria are provided in Online Resource 1. 2.6 Residue footprint and hotspot consensus analysis Residue–ligand contact sets were extracted from the top-ranked docking pose (mode-1) using a distance-based criterion consistent with protein–protein interface analyses (heavy-atom distance ≤ 4.0 Å). Binary residue footprints were constructed for each ligand–target pair based on the presence or absence of residue-level contacts. Pairwise Jaccard similarity indices were computed to quantify overlap between ligand interaction footprints. Consensus hotspot residues were defined as residues contacted by at least two ligands within the same target interface. 2.7 Structural mapping of CTLA-4 docking site relative to the B7-1 interface To assess the spatial relationship between the predicted CTLA-4 docking site and the physiological B7-1 binding interface, the CTLA-4/B7-1 complex structure was used as a reference.[ 25 ] Docking-interacting residues identified from CTLA-4 docking analysis were mapped onto the CTLA-4 structure and visually compared with the B7-1 interaction surface using molecular visualization tools. This analysis was performed to determine whether the docking site overlapped with, or was adjacent to, the B7-1 interface and to classify the site as an exploratory interface-compatible surface rather than a direct blockade site.[ 7 ], [ 24 ] 2.8 Molecular dynamics simulation of the PD-L1 dimer To assess the intrinsic structural integrity of the PD-L1 dimer interface, an apo molecular dynamics (MD) simulation was performed. The PD-L1 dimer structure (PDB ID: 5J89), retaining both protomers, was used as the initial model to preserve the native dimeric interface architecture. The simulation was designed to characterize inter-protomer behavior and interface stability rather than ligand dynamics or binding persistence. MD simulations were carried out using GROMACS, a widely used molecular simulation engine optimized for biomolecular systems.[ 26 ] The protein was parameterized using the CHARMM36m force field, which provides improved accuracy for folded and flexible protein regions.[ 27 ] An unrestrained production MD simulation was performed for 10 ns with a 2 fs integration time step. Long-range electrostatic interactions were treated using the particle mesh Ewald (PME) method,[ 28 ] and all covalent bonds involving hydrogen atoms were constrained using the LINCS algorithm. Interface-level analyses reported in this manuscript were computed from the available trajectory segment generated during post-processing. The 10-ns simulation length was selected to characterize short-timescale interface integrity and inter-protomer behavior rather than long-term conformational transitions or ligand residence dynamics. 2.9 Molecular dynamics trajectory analysis Trajectory analyses were performed using standard GROMACS analysis utilities. Structural stability of the PD-L1 dimer was evaluated by calculating the root-mean-square fluctuation (RMSF) of backbone atoms, computed separately for each protomer to assess potential asymmetry in flexibility. Hydrogen bond analysis was performed to determine the number and temporal persistence of inter-protomer hydrogen bonds during the simulation, providing further insight into interface stabilization. All reported inter-protomer interface metrics (contacts, minimum distance, and hydrogen bonds) were computed over the current analyzed segment. All molecular dynamics analyses were computed over the full 10-ns production trajectory. 2.10 Docking–MD structural integration To provide structural context for docking predictions, the top-ranked docking pose for PD-L1 was overlaid onto a representative MD snapshot of the PD-L1 dimer. Backbone alignment was performed to enable qualitative comparison between the docking-predicted ligand placement and the MD-derived dimer interface geometry. This integration step was intended for structural interpretation at the interface level and does not constitute validation of ligand binding stability, binding persistence, or inhibitory activity.[ 29 ] No ligand-bound MD simulations or binding free energy calculations were performed. The representative MD snapshot was selected from the equilibrated portion of the analyzed segment to support qualitative interface-level interpretation. 2.11 Target prediction and network construction Putative molecular targets of the selected polyphenolic compounds were predicted using SwissTargetPrediction,[ 30 ] with analyses restricted to Homo sapiens. For each ligand, the top-ranked predicted targets were retained and merged into a unified target set for subsequent network analysis. Protein–protein interaction (PPI) networks were constructed using STRING,[ 31 ] limited to Homo sapiens and first-shell interactors. A minimum confidence score of 0.4 was applied, incorporating experimentally validated interactions, curated database evidence, and co-expression data, while excluding text-mining–derived associations. CTLA-4 and PD-L1 were retained as seed proteins to anchor the network around immune checkpoint–related interactions. Putative molecular targets were predicted using SwissTargetPrediction, a ligand-based target inference platform. Predictions were restricted to Homo sapiens . Approximately 10–20 top-ranked targets per ligand were retained and combined as a union set for network analysis. 2.12 Network topology and functional enrichment analysis Network topology analysis was performed using STRING and Cytoscape,[ 32 ] with node degree used as the primary centrality metric to identify highly connected hub proteins. The ten nodes with the highest degree values were designated as hub nodes for downstream interpretation. Functional enrichment analysis was conducted using the STRING enrichment module, including Gene Ontology Biological Process, KEGG, and Reactome pathway analyses. Enrichment results were evaluated against the Homo sapiens genome background, with multiple testing correction applied using the Benjamini–Hochberg method and a false discovery rate threshold of < 0.05. Enrichment analyses were intended to provide pathway-level context for predicted target convergence rather than to infer direct biological modulation. 3. RESULTS Docking- and structure-derived observations reported in this section describe interface-level spatial compatibility and reproducibility within the applied computational framework and do not constitute evidence of immune checkpoint inhibition or functional immunomodulation. 3.1 Docking performance of polyphenols on CTLA-4 Mode-1 docking affinities for CTLA-4 spanned a narrow range (−8.0 to −6.1 kcal/mol), enabling consistent ranking across ligands (Table 1). Gnetin C ranked most favorably, followed by a middle tier including curcumin and epigallocatechin gallate, while resveratrol, quercetin, and kaempferol exhibited lower affinities. Top-ranked poses were reproducibly identified across repeated docking runs without rank inversion, indicating stable docking behavior for CTLA-4. The predicted CTLA-4 binding site was subsequently mapped against the physiological B7-1 interaction surface to contextualize the docking site within known structural constraints (see Section 3.2.1). Table 1. Mode-1 docking affinities of polyphenolic compounds toward CTLA-4. Ligand Binding affinity (kcal/mol) Gnetin C −8.0 EGCG −6.8 Curcumin −6.8 Resveratrol −6.3 Quercetin −6.1 Kaempferol −6.1 3.2 Interaction profiling of CTLA-4–ligand complexes All three representative ligands occupied the same CTLA-4 binding pocket with overlapping orientations. Interaction profiles converged on a shared set of interface-adjacent residues, including LYS56, LYS78, ASP34, and ASP79, indicating consistent pocket usage rather than alternative-site binding. Differences in interaction density among ligands reflected relative docking ranks but did not alter overall binding location (Figure 1). 3.2.1 Structural positioning of the CTLA-4 docking site relative to the B7-1 interface Mapping of docking-interacting residues onto the CTLA-4/B7-1 complex structure revealed that the predicted ligand-binding cavity is spatially adjacent to, but does not fully overlap with, the B7-1 binding interface. Key residues recurrently involved in docking interactions (LYS56, LYS78, ASP34, ASP79, TYR83) localized near the interface boundary rather than the central B7-1 contact region. This spatial relationship supports classification of the docking site as an interface-compatible exploratory surface rather than a direct blockade site. 3.3 Residue-level footprint overlap and hotspot consensus analysis Residue-level footprint analysis was performed to quantify ligand interaction convergence at the CTLA-4 and PD-L1 interfaces using docking-derived contact sets. At the PD-L1 dimer interface, quantifiable overlap was observed among ligand interaction profiles. Pairwise Jaccard similarity indices ranged from 0.43 to 0.80, indicating convergence toward a shared interaction region rather than diffuse surface association (Table 2). Table 2. Pairwise Jaccard similarity indices of ligand interaction footprints. Target Ligand pair Jaccard index PD-L1 dimer Gnetin C – Quercetin 0.57 PD-L1 dimer Gnetin C – Resveratrol 0.43 PD-L1 dimer Quercetin – Resveratrol 0.80 CTLA-4 Gnetin C – Quercetin 0.25 CTLA-4 Gnetin C – Resveratrol 0.50 CTLA-4 Quercetin – Resveratrol 0.33 Consensus hotspot analysis further identified a subset of recurrent interface residues (Table 3). These findings provide quantitative support for interface-level permissiveness and suggest dual-interface compatibility of polyphenolic scaffolds. The observed clustering highlights conserved interaction hotspots at the PD-L1 dimer interface and lower but non-random convergence at the CTLA-4 interface-adjacent surface (Figure 2). Table 3. Consensus hotspot residues identified across representative ligands. Target Residue Ligands involved Frequency PD-L1 dimer Tyr56 3/3 100% PD-L1 dimer Ala121 3/3 100% PD-L1 dimer Met115 3/3 100% PD-L1 dimer Asp122 2/3 67% PD-L1 dimer Tyr123 2/3 67% CTLA-4 Lys78 3/3 100% CTLA-4 Lys56 2/3 67% CTLA-4 Asp34 2/3 67% CTLA-4 Tyr83 2/3 67% CTLA-4 Asp79 2/3 67% 3.4 Docking performance of ligands on PD-L1 dimer All ligands converged to the same PD-L1 dimer interface pocket across independent docking runs, consistent with benchmark redocking and pose distribution analyses as shown in Online Resource 1. Gnetin C ranked most favorably within this interface, followed by resveratrol and quercetin, while the remaining polyphenols clustered at lower ranks (Table 4). Preservation of the dimeric architecture altered ligand accommodation relative to monomeric PD-L1, highlighting the structural relevance of the inter-protomer interface. Table 4 . Mode-1 binding affinities of all screened polyphenols against the PD-L1 dimer. Ligand Binding affinity (kcal/mol) Gnetin C −8.4 Kaempferol −7.9 Resveratrol −7.5 EGCG −7.4 Curcumin −7.4 Quercetin −7.2 3.5 Interaction profiling of PD-L1 dimer–ligand complexes Interaction profiling revealed that all representative ligands occupied a conserved hotspot at the PD-L1 dimer interface, engaging residues contributed by both protomers. Recurrent contacts involved Tyr56, Asp122, Ala121, and Met115, supporting a shared binding mechanism consistent with validated docking placement (Figure 3). 3.6 Molecular dynamics characterization of the PD-L1 dimer interface To characterize the intrinsic structural behavior of the PD-L1 dimer interface independent of ligand binding, an apo molecular dynamics simulation was performed over a 10-ns production trajectory. Structural stability was evaluated using backbone flexibility, inter-protomer contact persistence, minimum inter-protomer distance, and hydrogen bond analysis (Figure 4). Backbone root-mean-square fluctuation analysis revealed comparable flexibility profiles between protomers A and B, with low fluctuations observed across interface-forming residues relative to solvent-exposed regions. Inter-protomer contact analysis demonstrated sustained interface engagement throughout the simulation, with only minor temporal variation. Consistently, the minimum inter-protomer distance remained within a narrow range, indicating maintained proximity between the two chains. Hydrogen bond analysis further supported interface stability, with a small but persistent number of inter-protomer hydrogen bonds observed throughout the trajectory. No interface separation events or large-scale disruption were detected during the simulated timescale. Collectively, these observations indicate that the PD-L1 dimer interface remains structurally cohesive under apo conditions, providing a stable architectural context for interface-focused docking interpretation rather than evidence of ligand binding stability. 3.7 Structural integration of docking and molecular dynamics To provide structural context for docking predictions, the top-ranked docking pose of gnetin C was overlaid onto a representative snapshot extracted from the equilibrated molecular dynamics trajectory of the PD-L1 dimer. Backbone alignment was performed to enable qualitative comparison between docking-predicted ligand placement and the geometry of the MD-derived dimer interface. The overlay indicated that the docking-predicted ligand placement spatially overlapped with an interface region characterized by low backbone fluctuation and persistent inter-protomer contacts in the apo MD simulation. This observation suggests that docking convergence occurred within a pre-organized interface region rather than a highly flexible surface groove. This qualitative overlay is consistent with the benchmark redocking results, which independently confirmed convergence of docking poses to the same PD-L1 interface region (Online Resource 1). 3.8 Predicted molecular targets of representative polyphenols Predicted molecular targets for gnetin C, quercetin, and resveratrol were obtained using SwissTargetPrediction and used to support subsequent network-based analyses. For each ligand, top-ranked predicted targets were retained based on probability scores and merged for comparative evaluation. Across the three ligands, target prediction revealed partial overlap, with several targets shared between at least two compounds and a smaller subset common to all three. Despite differences in individual target composition, the predicted target sets exhibited comparable distributions across major protein classes, including enzymes, kinases, and nuclear receptors (Table 5). A subset of predicted targets was associated with immune-related signaling and regulatory pathways based on functional annotation. These shared targets provided a rationale for protein–protein interaction network construction to examine connectivity patterns in relation to immune checkpoint proteins, without implying direct biological modulation. Table 5. Predicted molecular target classes of gnetin C, quercetin, and resveratrol Ligand Dominant target classes Representative targets Gnetin C Nuclear receptors, kinases, enzymes ESR1, ESR2, SRC, MAPK1 Quercetin Enzymes, nuclear receptors, kinases PTGS2, ESR1, MAPK1 Resveratrol Nuclear receptors, enzymes, kinases ESR1, SIRT1, MAPK1 3.9 Protein–protein interaction network analysis Protein–protein interaction network analysis was performed to explore connectivity among predicted molecular targets in relation to immune checkpoint proteins. Separate networks were constructed for CTLA-4–centered and PD-L1–centered target sets using STRING, with CTLA-4 and CD274 (PD-L1) included as seed nodes. Both networks exhibited statistically significant enrichment of protein–protein interactions relative to random expectation, indicating non-random connectivity within the predicted target space. In each network, the immune checkpoint protein occupied the dominant hub position, with secondary connectivity observed among immune-related signaling components. The PD-L1–centered network displayed higher overall connectivity compared with the CTLA-4–centered network, consistent with broader interaction context associated with the PD-L1 dimer interface. These network characteristics provide systems-level context for the convergence of predicted targets around immune checkpoint–associated interaction landscapes and are not intended to infer functional immune modulation. 3.10 Functional enrichment results Functional enrichment analysis was conducted separately for CTLA-4–centered and PD-L1–centered protein sets to provide pathway-level context for predicted target connectivity. Enrichment significance was assessed using Benjamini–Hochberg correction with a false discovery rate threshold of < 0.05. For the CTLA-4–centered network, Gene Ontology Biological Process analysis identified enrichment of immune-related processes associated with T-cell signaling and costimulatory regulation. KEGG and Reactome pathway analyses further highlighted pathways involved in immune communication and lymphocyte-associated signaling. These enriched terms reflected functional coherence among CTLA-4 and its directly connected interactors rather than isolated pathway associations. Functional enrichment of the PD-L1–centered network revealed a partially overlapping but broader enrichment profile. Gene Ontology terms were enriched for immune signaling and receptor-mediated processes, while KEGG and Reactome analyses identified multiple signaling pathways consistent with the denser connectivity observed in the PD-L1–centered network. The higher number of significantly enriched pathways reflected the increased interaction density within this network rather than differential biological effect. Across both networks, enrichment patterns supported non-random functional clustering of predicted targets around immune checkpoint–associated signaling landscapes. These results provide systems-level context for the observed network structures and are not intended to imply direct modulation of immune pathways or therapeutic activity. 4. DISCUSSION Most computational studies of small-molecule immune checkpoint modulation remain single-target and monomer-centric, primarily focusing on affinity ranking rather than interface-level permissiveness. In contrast, the present study reframes small-molecule engagement as an interface-compatibility problem across immune checkpoints rather than a target-specific inhibition problem. By jointly evaluating CTLA-4 and the PD-L1 dimer interface within a unified structural and systems-level framework, this work emphasizes spatial convergence, interface reuse, and architectural permissiveness rather than optimization of docking scores. This dual-interface perspective does not aim to identify potent inhibitors but instead proposes a structural hypothesis in which certain chemical scaffolds are inherently compatible with multiple immune checkpoint interfaces. Such compatibility may represent a prerequisite for rational interface-targeted ligand optimization rather than a surrogate for biological activity. 4.1 Binding behavior of polyphenols on immune checkpoint proteins This study shows that selected polyphenolic compounds exhibit reproducible docking compatibility with two immune checkpoint proteins, CTLA-4 and PD-L1, within the applied computational framework. Across both targets, docking results consistently ranked gnetin C as the most favorable ligand, while quercetin and resveratrol displayed intermediate or architecture-dependent docking performance. Although CTLA-4 is classically discussed in the context of its interaction with B7 family ligands, the present analysis does not aim to infer direct functional blockade. Instead, CTLA-4 is considered here as a structurally distinct immune checkpoint surface used for cross-interface comparison. From a structural perspective, immune checkpoint proteins expose multiple solvent-accessible regions beyond their canonical ligand-binding interfaces, allowing small molecules to engage interface-adjacent or permissive surfaces without necessarily disrupting physiological protein–protein interactions. In this context, the convergence of similar polyphenolic scaffolds on interface-adjacent regions of both PD-L1 and CTLA-4 supports a model of scaffold-level interface compatibility rather than target-specific inhibition.[ 3 ], [ 12 ] Importantly, the reported docking affinities should not be interpreted as absolute predictors of biological activity. Rather, they reflect relative compatibility between ligand physicochemical features and receptor surface topology. Polyphenols are characterized by extended aromatic systems and multiple hydrogen bond donors and acceptors, properties that favor interactions with shallow, solvent-exposed protein surfaces rather than deep catalytic pockets. Immune checkpoint proteins, which function primarily through protein–protein interactions, present such interfaces as potential sites for small-molecule engagement. This binding paradigm aligns with growing evidence that modulation of protein–protein interfaces by small molecules is feasible when ligands exploit distributed hydrophobic contacts, aromatic stacking, and electrostatic anchoring instead of classical lock-and-key complementarity.[ 7 ], [ 9 ] Several recent studies have highlighted that docking against protein–protein interaction surfaces yields narrower affinity ranges and higher pose degeneracy compared with enzyme active sites, emphasizing the importance of reproducibility and spatial convergence rather than absolute docking scores.[ 8 ], [ 33 ], [ 34 ] 4.2 Structural determinants of PD-L1 dimer interface engagement A central finding of this work is that preservation of the PD-L1 dimer architecture reveals a conserved interface pocket that is preferentially occupied by all representative ligands in docking analyses. Docking consistently identified residues such as Tyr56, Asp122, Ala121, and Met115 as recurrent contributors to ligand accommodation at the dimer interface, consistent with the presence of a structurally relevant interaction hotspot rather than a docking artifact. Molecular dynamics analysis further indicated that this interface region constitutes a structurally maintained element of the PD-L1 dimer within the simulated timescale. Reduced backbone flexibility at the interface relative to terminal and solvent-exposed regions, together with persistent inter-protomer contacts and hydrogen bonds, suggests that the identified pocket is pre-organized and structurally maintained under the simulated conditions. The recurrence of identical interface residues across chemically distinct ligands further argues against stochastic surface docking and supports the presence of a structurally permissive interface hotspot. Previous structural and biophysical studies have demonstrated that PD-L1 dimerization can be stabilized by small molecules and represents a functionally distinct structural state compared with the monomer.[ 12 ], [ 35 ] Targeting this interface has been proposed as a potential indirect strategy to influence PD-1 recognition through steric or conformational effects, as suggested by prior structural studies. The improved docking rank of resveratrol in the dimeric PD-L1 context compared with monomer-based docking further underscores the sensitivity of ligand accommodation to receptor oligomeric state. Among the investigated ligands, gnetin C formed the most extensive interaction network spanning both protomers, consistent with its favorable docking rank within the applied protocol. Quercetin and resveratrol exhibited progressively simpler interaction profiles, suggesting that molecular size, aromatic density, and spatial reach influence the capacity of polyphenols to bridge the PD-L1 dimer interface effectively. 4.3 Integration of docking and molecular dynamics insights The molecular dynamics analysis was designed to provide descriptive characterization of PD-L1 dimer interface integrity over the simulated timescale rather than statistical estimation across independent replicas. Although only a single apo trajectory was analyzed, multiple independent interface metrics exhibited consistent behavior across consecutive simulation windows with available data. Within these constraints, the MD results support interface-level interpretation of docking predictions by defining the structural environment of the receptor rather than validating ligand binding stability or functional modulation.[ 36 ] While longer and replicated simulations would be required to capture rare conformational events, the present timescale is sufficient to assess baseline interface cohesion and structural maintenance. Importantly, quantitative footprint overlap and hotspot consensus analyses indicate that ligand accommodation across CTLA-4 and PD-L1 interfaces is driven by shared residue hotspots rather than stochastic docking placement. The higher degree of footprint convergence observed at the PD-L1 dimer interface relative to CTLA-4 further supports the role of receptor architecture in shaping interface permissiveness, consistent with the conceptual framing of scaffold-level compatibility rather than direct functional inhibition. 4.4 Network-level interpretation of dual immune checkpoint modulation Beyond direct protein–ligand interactions, network analysis revealed that predicted molecular targets of the investigated polyphenols converge on immune-related signaling nodes associated with T-cell regulation, costimulatory signaling, and immune checkpoint pathways. In both CTLA-4– and PD-L1–centered networks, the immune checkpoint proteins emerged as dominant hubs, reinforcing their relevance as focal points for multi-target modulation. The convergence of docking-supported targets with network-identified central nodes supports a systems-level interpretation in which predicted ligand targets map onto interconnected immune-related pathways rather than isolated proteins. These network features reflect topological properties derived from the selected interaction database and do not imply direct modulation of all connected nodes by the investigated compounds. Within this framework, dual consideration of CTLA-4– and PD-L1–associated networks provides a conceptual basis for exploring pathway-level relationships between early T-cell priming and peripheral immune regulation, offering a mechanistic rationale for the dual-target hypothesis explored in this study.[ 37 ], [ 38 ] Network-level analysis was applied as a contextualization layer to relate structurally compatible scaffolds to pathway-adjacent protein sets, rather than as evidence of direct multi-target modulation. Within this constraint, network convergence highlights biological neighborhoods potentially compatible with the identified scaffold architectures. 4.5 Implications for immunotherapy development Current immune checkpoint therapies rely predominantly on monoclonal antibodies, which offer high specificity but are associated with limitations such as immune-related adverse events, high production costs, and restricted tissue penetration. In particular, antibody-based checkpoint blockade primarily targets extracellular interactions, whereas small molecules may enable alternative modes of interface engagement or allosteric modulation. Small-molecule modulators represent a complementary strategy, potentially enabling oral administration, tunable pharmacokinetics, and combinatorial use with antibody-based therapies.[ 39 ], [ 40 ] Although the present study does not claim direct inhibitory or therapeutic activity, the structural evidence presented here supports the feasibility of small-molecule engagement of immune checkpoint interfaces. Polyphenolic scaffolds, in particular, may serve as starting points for rational optimization rather than final drug candidates. Their interaction patterns highlight interface-compatible structural motifs that could be refined to enhance interface compatibility and interaction selectivity. 4.6 Study limitations and future directions Several limitations should be acknowledged. First, while molecular dynamics analysis was performed to assess intrinsic dimer stability, ligand-bound simulations and binding free energy calculations were not conducted. Consequently, conclusions regarding ligand stability over time remain inferential rather than quantitative. The observed consistency across independent docking seeds further supports that the identified PD-L1 dimer interface interactions are not artifacts of stochastic sampling. Second, all findings are derived from computational analyses and do not account for cellular context, protein expression levels, or downstream signaling effects. Docking scores and interaction profiles alone cannot predict immunomodulatory efficacy or safety. Ligand-bound molecular dynamics simulations were not performed in this study by design. The primary objective was to assess interface permissiveness and structural accommodation rather than ligand residence time, binding persistence, or inhibitory behavior. By decoupling interface stability from ligand dynamics, this study focuses on defining whether the PD-L1 dimer interface constitutes a pre-organized structural feature capable of accommodating chemically diverse scaffolds. Ligand-bound simulations and free energy calculations are necessary to evaluate binding stability and functional relevance but fall outside the scope of the present interface-level analysis. Future studies should incorporate ligand-bound molecular dynamics simulations, free energy estimations, and experimental validation using biochemical and cellular assays. Such efforts will be essential to determine whether interface-binding polyphenols can modulate immune checkpoint signaling in biologically meaningful ways. Accordingly, the findings presented here should be interpreted as structural and network-level hypotheses that require experimental validation to establish biological relevance. 5. CONCLUSION This study presents an integrated in silico framework to examine polyphenolic scaffolds in relation to two immune checkpoint proteins, CTLA-4 and PD-L1, by combining molecular docking, molecular dynamics–based structural characterization, and network-level analysis. Docking analyses demonstrated reproducible and spatially consistent interface-level accommodation of gnetin C, quercetin, and resveratrol across both targets, with gnetin C consistently ranked most favorably within the applied protocol. Preservation of the PD-L1 dimer architecture revealed a conserved inter-protomer interface pocket shared across ligands, highlighting the importance of receptor oligomeric state in shaping small-molecule interface compatibility. Apo molecular dynamics analysis was used to characterize intrinsic PD-L1 dimer interface integrity rather than ligand behavior. Persistent inter-protomer contacts, maintained hydrogen bonding, and reduced backbone flexibility at the interface support its classification as a structurally maintained and pre-organized architectural feature, providing context for interface-focused docking interpretation without assessing ligand binding stability or functional modulation. At the systems level, network analysis indicated convergence of predicted molecular targets on immune-related signaling nodes, with CTLA-4 and PD-L1 emerging as central hubs. This observation supports a dual-target concept framed in terms of structural compatibility and pathway connectivity rather than direct inhibitory or immunomodulatory activity. Overall, this work defines interface-level structural compatibility of selected polyphenolic scaffolds with immune checkpoint proteins and provides a structure-informed, hypothesis-generating framework to guide future ligand optimization and experimental validation, without implying immune checkpoint inhibition or therapeutic efficacy. Declarations Competing Interests The authors have no relevant financial or non-financial interests to disclose. Funding The authors declare that no funds, grants, or other support were received during the preparation of this manuscript. Author Contribution MRPPS conceived the study, designed the computational workflow, performed data collection and analysis, and wrote the manuscript. Acknowledgments The authors thank colleagues for constructive discussions and technical assistance related to computational analysis. Data Availability The datasets generated during and/or analysed during the current study, including molecular docking outputs, molecular dynamics trajectories, and processed analysis files, are available from the corresponding author on reasonable request. Publicly available protein structures used in this study were obtained from the Protein Data Bank. References S. C. Wei, C. R. Duffy, and J. P. Allison, “Fundamental Mechanisms of Immune Checkpoint Blockade Therapy,” Cancer Discovery , vol. 8, no. 9, pp. 1069–1086, Sep. 2018, doi: 10.1158/2159-8290.CD-18-0367. A. Ribas and J. D. Wolchok, “Cancer immunotherapy using checkpoint blockade,” Science , vol. 359, no. 6382, pp. 1350–1355, Mar. 2018, doi: 10.1126/science.aar4060. D. M. Pardoll, “The blockade of immune checkpoints in cancer immunotherapy,” Nat Rev Cancer , vol. 12, no. 4, pp. 252–264, Apr. 2012, doi: 10.1038/nrc3239. P. Sharma, S. Hu-Lieskovan, J. A. Wargo, and A. Ribas, “Primary, Adaptive, and Acquired Resistance to Cancer Immunotherapy,” Cell , vol. 168, no. 4, pp. 707–723, Feb. 2017, doi: 10.1016/j.cell.2017.01.017. M. A. Postow, R. Sidlow, and M. D. Hellmann, “Immune-Related Adverse Events Associated with Immune Checkpoint Blockade,” N Engl J Med , vol. 378, no. 2, pp. 158–168, Jan. 2018, doi: 10.1056/NEJMra1703481. F. Martins et al. , “Adverse effects of immune-checkpoint inhibitors: epidemiology, management and surveillance,” Nat Rev Clin Oncol , vol. 16, no. 9, pp. 563–580, Sep. 2019, doi: 10.1038/s41571-019-0218-0. M. R. Arkin, Y. Tang, and J. A. Wells, “Small-Molecule Inhibitors of Protein-Protein Interactions: Progressing toward the Reality,” Chemistry & Biology , vol. 21, no. 9, pp. 1102–1114, Sep. 2014, doi: 10.1016/j.chembiol.2014.09.001. D. E. Scott, A. R. Bayly, C. Abell, and J. Skidmore, “Small molecules, big targets: drug discovery faces the protein–protein interaction challenge,” Nat Rev Drug Discov , vol. 15, no. 8, pp. 533–550, Aug. 2016, doi: 10.1038/nrd.2016.29. A. Mullard, “Protein–protein interaction inhibitors get into the groove,” Nat Rev Drug Discov , vol. 11, no. 3, pp. 173–175, Mar. 2012, doi: 10.1038/nrd3680. H. Cory, S. Passarelli, J. Szeto, M. Tamez, and J. Mattei, “The Role of Polyphenols in Human Health and Food Systems: A Mini-Review,” Front. Nutr. , vol. 5, p. 87, Sep. 2018, doi: 10.3389/fnut.2018.00087. C. G. Fraga, K. D. Croft, D. O. Kennedy, and F. A. Tomás-Barberán, “The effects of polyphenols and other bioactives on human health,” Food Funct. , vol. 10, no. 2, pp. 514–528, 2019, doi: 10.1039/C8FO01997E. K. M. Zak et al. , “Structure of the Complex of Human Programmed Death 1, PD-1, and Its Ligand PD-L1,” Structure , vol. 23, no. 12, pp. 2341–2348, Dec. 2015, doi: 10.1016/j.str.2015.09.010. K. Guzik et al. , “Development of the Inhibitors That Target the PD-1/PD-L1 Interaction—A Brief Look at Progress on Small Molecules, Peptides and Macrocycles,” Molecules , vol. 24, no. 11, p. 2071, May 2019, doi: 10.3390/molecules24112071. L. Skalniak et al. , “Small-molecule inhibitors of PD-1/PD-L1 immune checkpoint alleviate the PD-L1-induced exhaustion of T-cells,” Oncotarget , vol. 8, no. 42, pp. 72167–72181, Sep. 2017, doi: 10.18632/oncotarget.20050. A. L. Hopkins, “Network pharmacology: the next paradigm in drug discovery,” Nat Chem Biol , vol. 4, no. 11, pp. 682–690, Nov. 2008, doi: 10.1038/nchembio.118. U. Chandran, N. Mehendale, S. Patil, R. Chaguturu, and B. Patwardhan, “Network Pharmacology,” in Innovative Approaches in Drug Discovery , Elsevier, 2017, pp. 127–164. doi: 10.1016/B978-0-12-801814-9.00005-2. H. M. Berman et al. , “The Protein Data Bank,” Nucleic Acids Research , vol. 28, no. 1, pp. 235–242, Jan. 2000, doi: 10.1093/nar/28.1.235. G. M. Morris et al. , “AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility,” J Comput Chem , vol. 30, no. 16, pp. 2785–2791, Dec. 2009, doi: 10.1002/jcc.21256. S. Kim et al. , “PubChem 2023 update,” Nucleic Acids Research , vol. 51, no. D1, pp. D1373–D1380, Jan. 2023, doi: 10.1093/nar/gkac956. Y. Liu, X. Yang, J. Gan, S. Chen, Z.-X. Xiao, and Y. Cao, “CB-Dock2: improved protein–ligand blind docking by integrating cavity detection, docking and homologous template fitting,” Nucleic Acids Research , vol. 50, no. W1, pp. W159–W164, Jul. 2022, doi: 10.1093/nar/gkac394. O. Trott and A. J. Olson, “AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading,” J Comput Chem , vol. 31, no. 2, pp. 455–461, Jan. 2010, doi: 10.1002/jcc.21334. N. S. Pagadala, K. Syed, and J. Tuszynski, “Software for molecular docking: a review,” Biophys Rev , vol. 9, no. 2, pp. 91–102, Apr. 2017, doi: 10.1007/s12551-016-0247-1. D. Kozakov et al. , “The FTMap family of web servers for determining and characterizing ligand-binding hot spots of proteins,” Nat Protoc , vol. 10, no. 5, pp. 733–755, May 2015, doi: 10.1038/nprot.2015.043. J. A. Wells and C. L. McClendon, “Reaching for high-hanging fruit in drug discovery at protein–protein interfaces,” Nature , vol. 450, no. 7172, pp. 1001–1009, Dec. 2007, doi: 10.1038/nature06526. C. C. Stamper et al. , “Crystal structure of the B7-1/CTLA-4 complex that inhibits human immune responses,” Nature , vol. 410, no. 6828, pp. 608–611, Mar. 2001, doi: 10.1038/35069118. M. J. Abraham et al. , “GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers,” SoftwareX , vol. 1–2, pp. 19–25, Sep. 2015, doi: 10.1016/j.softx.2015.06.001. J. Huang et al. , “CHARMM36m: an improved force field for folded and intrinsically disordered proteins,” Nat Methods , vol. 14, no. 1, pp. 71–73, Jan. 2017, doi: 10.1038/nmeth.4067. T. Darden, D. York, and L. Pedersen, “Particle mesh Ewald: An N ⋅log( N ) method for Ewald sums in large systems,” The Journal of Chemical Physics , vol. 98, no. 12, pp. 10089–10092, Jun. 1993, doi: 10.1063/1.464397. P. V. Klimovich, M. R. Shirts, and D. L. Mobley, “Guidelines for the analysis of free energy calculations,” J Comput Aided Mol Des , vol. 29, no. 5, pp. 397–411, May 2015, doi: 10.1007/s10822-015-9840-9. A. Daina, O. Michielin, and V. Zoete, “SwissTargetPrediction: updated data and new features for efficient prediction of protein targets of small molecules,” Nucleic Acids Research , vol. 47, no. W1, pp. W357–W364, Jul. 2019, doi: 10.1093/nar/gkz382. D. Szklarczyk et al. , “The STRING database in 2021: customizable protein–protein networks, and functional characterization of user-uploaded gene/measurement sets,” Nucleic Acids Research , vol. 49, no. D1, pp. D605–D612, Jan. 2021, doi: 10.1093/nar/gkaa1074. P. Shannon et al. , “Cytoscape: A Software Environment for Integrated Models of Biomolecular Interaction Networks,” Genome Res. , vol. 13, no. 11, pp. 2498–2504, Nov. 2003, doi: 10.1101/gr.1239303. H. Lu et al. , “Recent advances in the development of protein–protein interactions modulators: mechanisms and clinical trials,” Sig Transduct Target Ther , vol. 5, no. 1, p. 213, Sep. 2020, doi: 10.1038/s41392-020-00315-3. H. Jubb, T. L. Blundell, and D. B. Ascher, “Flexibility and small pockets at protein–protein interfaces: New insights into druggability,” Progress in Biophysics and Molecular Biology , vol. 119, no. 1, pp. 2–9, Oct. 2015, doi: 10.1016/j.pbiomolbio.2015.01.009. Q. Wu, L. Jiang, S. Li, Q. He, B. Yang, and J. Cao, “Small molecule inhibitors targeting the PD-1/PD-L1 signaling pathway,” Acta Pharmacol Sin , vol. 42, no. 1, pp. 1–9, Jan. 2021, doi: 10.1038/s41401-020-0366-x. S. A. Hollingsworth and R. O. Dror, “Molecular Dynamics Simulation for All,” Neuron , vol. 99, no. 6, pp. 1129–1143, Sep. 2018, doi: 10.1016/j.neuron.2018.08.011. D. S. Chen and I. Mellman, “Elements of cancer immunity and the cancer–immune set point,” Nature , vol. 541, no. 7637, pp. 321–330, Jan. 2017, doi: 10.1038/nature21349. E. I. Buchbinder and A. Desai, “CTLA-4 and PD-1 Pathways: Similarities, Differences, and Implications of Their Inhibition,” American Journal of Clinical Oncology , vol. 39, no. 1, pp. 98–106, Feb. 2016, doi: 10.1097/COC.0000000000000239. V. A. Boussiotis, “Molecular and Biochemical Aspects of the PD-1 Checkpoint Pathway,” N Engl J Med , vol. 375, no. 18, pp. 1767–1778, Nov. 2016, doi: 10.1056/NEJMra1514296. C. Feng et al. , “Discovery of Small-Molecule PD-L1 Inhibitors via Virtual Screening and Their Immune-Mediated Anti-Tumor Effects,” Pharmaceuticals , vol. 18, no. 8, p. 1209, Aug. 2025, doi: 10.3390/ph18081209. Additional Declarations No competing interests reported. Supplementary Files OnlineResource1.docx Online Resource 1 Supplementary methods, results, tables, figures, and detailed computational environment information supporting the results of this study. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8841071","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":602591900,"identity":"acac2eaf-ad4a-4409-a181-ff635504793e","order_by":0,"name":"Muhammad Rafif Pratama Putra Shanizal","email":"data:image/png;base64,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","orcid":"","institution":"Universitas Gadjah Mada Yogyakarta","correspondingAuthor":true,"prefix":"","firstName":"Muhammad","middleName":"Rafif Pratama Putra","lastName":"Shanizal","suffix":""}],"badges":[],"createdAt":"2026-02-10 12:38:33","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8841071/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8841071/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104441556,"identity":"f620bd1b-f687-4f76-a487-1d8970e8f95a","added_by":"auto","created_at":"2026-03-11 18:32:42","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":192465,"visible":true,"origin":"","legend":"\u003cp\u003eTwo-dimensional interaction maps of representative polyphenols docked to CTLA-4. (A) Gnetin C, (B) Quercetin, (C) Resveratrol. Hydrogen bonds, π-based interactions, and hydrophobic contacts within the predicted binding pocket are shown.\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8841071/v1/ddc9b5f4ba8aafcdaa776129.jpeg"},{"id":104441554,"identity":"53910286-023d-4c62-95f9-34b61e44b74b","added_by":"auto","created_at":"2026-03-11 18:32:42","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":74300,"visible":true,"origin":"","legend":"\u003cp\u003eBinary residue–ligand footprint heatmap illustrating conserved interface contact patterns across representative polyphenolic scaffolds. Rows represent interface residues, and columns represent ligands. Filled cells indicate residue–ligand contacts derived from the top-ranked docking pose (mode-1). The observed clustering highlights conserved interaction hotspots at the PD-L1 dimer interface and lower but non-random convergence at the CTLA-4 interface-adjacent surface.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8841071/v1/b2e0432ec16b6d61830ac6c4.png"},{"id":104780181,"identity":"24319efd-8792-492f-96b4-573f78e14279","added_by":"auto","created_at":"2026-03-17 07:51:19","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":206764,"visible":true,"origin":"","legend":"\u003cp\u003eTwo-dimensional interaction maps of representative polyphenols docked to PD-L1 dimer. (A) Gnetin C, (B) Quercetin, (C) Resveratrol. Hydrogen bonds, π-based interactions, and hydrophobic contacts within the predicted binding pocket are shown.\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8841071/v1/7839e48792b654092390f90f.jpeg"},{"id":104441557,"identity":"e98315b9-fa4a-41e3-aec2-2a739d09e3d0","added_by":"auto","created_at":"2026-03-11 18:32:42","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":22105,"visible":true,"origin":"","legend":"\u003cp\u003eMolecular dynamics characterization of the PD-L1 dimer interface over a 10-ns production trajectory. (A) Backbone RMSF profiles of protomers A and B. (B) Time evolution of inter-protomer contact counts using a 0.35-nm cutoff. (C) Minimum inter-protomer distance over time. (D) Number of inter-protomer hydrogen bonds formed during the simulation. All metrics were computed over the full 10-ns trajectory. These analyses describe intrinsic dimer interface behavior and do not assess ligand binding stability or binding free energy.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8841071/v1/066b331bf6fa8fdecb909f55.png"},{"id":104867605,"identity":"64b019ee-d967-483b-9958-512e4d40c5d1","added_by":"auto","created_at":"2026-03-18 07:13:23","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1869408,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8841071/v1/441f6594-1bb3-4246-995c-1259bac3ff42.pdf"},{"id":104441558,"identity":"400fa69c-67ad-4670-864a-d3734dd22dcd","added_by":"auto","created_at":"2026-03-11 18:32:42","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1144675,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eOnline Resource 1\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSupplementary methods, results, tables, figures, and detailed computational environment information supporting the results of this study.\u003c/p\u003e","description":"","filename":"OnlineResource1.docx","url":"https://assets-eu.researchsquare.com/files/rs-8841071/v1/7f738abce291b2f96cfabb86.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Structural and systems-level analysis of polyphenolic scaffold compatibility at immune checkpoint interfaces: a PD-L1 dimer–focused in silico study","fulltext":[{"header":"1. INTRODUCTION","content":"\u003cdiv id=\"Sec2\" class=\"Section2\"\u003e \u003ch2\u003e1.1 Immune checkpoints in cancer immunotherapy\u003c/h2\u003e \u003cp\u003eImmune checkpoint pathways regulate the magnitude and duration of T-cell activation, maintaining immune homeostasis and preventing excessive immune responses that may lead to tissue damage or autoimmunity. Among the best-characterized immune checkpoints, cytotoxic T-lymphocyte\u0026ndash;associated protein 4 (CTLA-4) and programmed death-ligand 1 (PD-L1) act as dominant regulators of antitumor immunity through distinct yet complementary mechanisms.[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e] Despite their biological relevance, immune checkpoint protein\u0026ndash;protein interaction interfaces pose persistent challenges for molecular modeling due to shallow topology, conformational flexibility, and limited canonical binding pockets.\u003c/p\u003e \u003cp\u003eCTLA-4 primarily regulates early T-cell priming within secondary lymphoid organs by competing with CD28 for the costimulatory ligands CD80 and CD86, thereby attenuating initial activation signals. CTLA-4 was therefore included not as a primary target, but as a structurally distinct immune checkpoint interface to enable comparative assessment of scaffold compatibility across divergent protein\u0026ndash;protein interaction architectures. In contrast, PD-L1 suppresses effector T-cell function in peripheral tissues through engagement with programmed death-1 (PD-1), contributing to immune tolerance within the tumor microenvironment.[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e], [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eTherapeutic blockade of CTLA-4 and the PD-1/PD-L1 axis using monoclonal antibodies has fundamentally altered cancer treatment, yielding durable clinical responses across multiple malignancies.[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e] Nevertheless, antibody-based immune checkpoint inhibitors are frequently associated with immune-related adverse events, high production costs, and limited tissue penetration. These challenges have motivated exploration of complementary strategies that expand immune checkpoint modulation beyond biologic therapeutics.[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e1.2 Small-molecule modulation of immune checkpoints\u003c/h2\u003e \u003cp\u003eSmall molecules represent an attractive alternative or adjunct to antibody-based therapies due to their oral bioavailability, scalable synthesis, and tunable pharmacokinetic profiles. However, immune checkpoint proteins pose substantial challenges for small-molecule targeting because their functional regions are dominated by relatively flat and solvent-exposed protein\u0026ndash;protein interaction interfaces rather than deep catalytic pockets.[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eAdvances in protein\u0026ndash;protein interaction drug discovery have demonstrated that such interfaces can be engaged by small molecules through distributed binding mechanisms involving aromatic stacking, hydrophobic surface complementarity, and electrostatic anchoring rather than classical lock-and-key interactions.[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] Within this framework, polyphenolic compounds have emerged as structurally relevant chemical scaffolds, as their extended aromatic systems and multiple hydrogen-bonding functionalities are well suited for interaction with shallow protein surfaces rather than classical active-site pockets.[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e], [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] In this context, polyphenols are considered as chemically diverse scaffolds for probing interface permissiveness rather than as candidate therapeutic inhibitors.\u003c/p\u003e \u003cp\u003eEmerging structural and computational studies further indicate that PD-L1 can adopt oligomeric assemblies that expose ligand-accessible interface pockets not apparent in monomeric conformations. Small molecules binding at the PD-L1 dimer interface have been shown to stabilize alternative protein arrangements and indirectly interfere with PD-1 recognition through steric or conformational mechanisms.[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e], [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e], [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eFrom a molecular modeling perspective, the PD-L1 dimer interface remains comparatively underexplored relative to monomeric binding sites, particularly with respect to scaffold-level interface compatibility. These findings highlight receptor oligomeric state as a critical determinant of interface-level ligand accommodation and underscore the importance of preserving native quaternary structure in computational evaluation of immune checkpoint interactions.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e1.3 Rationale and objective of the study\u003c/h2\u003e \u003cp\u003eDespite increasing interest in small-molecule engagement of immune checkpoint interfaces, many computational studies remain focused on single targets or monomeric protein structures, potentially overlooking interface-specific binding opportunities relevant to immune regulation. In particular, systematic evaluation of compounds capable of engaging both CTLA-4 and PD-L1 within a unified structural and systems-level framework remains limited. Unlike many prior studies that focus on monomeric PD-L1 structures or single-target affinity ranking, this work preserves the PD-L1 dimer architecture and evaluates interface-level compatibility across two immune checkpoints within a unified framework.\u003c/p\u003e \u003cp\u003eThe present study addresses this gap by applying an integrated in silico strategy to examine selected polyphenolic compounds as interface-compatible scaffolds in relation to two major immune checkpoints. Molecular docking was performed against CTLA-4 and the PD-L1 homodimer to assess ligand accommodation at structurally relevant surface cavities. Molecular dynamics\u0026ndash;based structural characterization was incorporated to describe the intrinsic integrity of the PD-L1 dimer interface, providing dynamic context for interface-focused docking interpretation. In parallel, network-based analyses were conducted to examine convergence of predicted molecular targets on immune-related signaling pathways.[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e], [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eDespite extensive computational exploration of immune checkpoint-related ligands, few studies have explicitly addressed interface-preserving modeling strategies that prioritize spatial compatibility, residue-level footprint convergence, and intrinsic interface stability over affinity-based ranking. As a result, scaffold-level permissiveness at immune checkpoint protein\u0026ndash;protein interfaces remains poorly defined from a structural modeling perspective.\u003c/p\u003e \u003cp\u003eSpecifically, this study aims to\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003echaracterize the docking behavior of selected polyphenols toward CTLA-4 and PD-L1,\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eassess the influence of PD-L1 dimeric architecture on interface-level ligand accommodation, and\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eintegrate structure-based docking results with network analyses to explore pathway-level relevance of dual immune checkpoint engagement.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eThrough this approach, the study provides a structural and systems-level rationale for dual immune checkpoint interface compatibility while clearly distinguishing computational insight from functional or therapeutic validation.\u003c/p\u003e \u003c/div\u003e"},{"header":"2. MATERIALS AND METHODS","content":"\u003cp\u003eAll computational analyses were conducted using a combination of Windows 11 and Ubuntu Linux environments. Molecular docking was performed using AutoDock Vina via PyRx, molecular dynamics simulations were carried out using GROMACS, and network analyses were performed using STRING and Cytoscape. Detailed software versions and computational environment specifications are provided in Online Resource 1.\u003c/p\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Protein structure preparation\u003c/h2\u003e \u003cp\u003eThree-dimensional structures of immune checkpoint proteins were obtained from the Protein Data Bank (PDB), a curated repository of experimentally determined biomolecular structures.[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] The extracellular domain of cytotoxic T-lymphocyte\u0026ndash;associated protein 4 (CTLA-4) was retrieved under PDB ID 1I85 at a resolution of 2.3 \u0026Aring;, while programmed death-ligand 1 (PD-L1) was obtained under PDB ID 5J89 at a resolution of 2.45 \u0026Aring;.\u003c/p\u003e \u003cp\u003eBoth proteins crystallize as homodimers. For CTLA-4, chain A was selected for docking analysis, as the reported ligand-accessible surface is contained within a single protomer. In contrast, PD-L1 chains A and B were retained to preserve the dimeric architecture, which exposes an inter-protomer interface pocket previously reported to accommodate small-molecule binding.[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e], [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eCTLA-4 was included as an architectural comparator rather than as a primary inhibitory target. Docking to CTLA-4 was performed to assess whether polyphenolic scaffolds capable of accommodating the PD-L1 dimer interface also exhibit compatibility with a second immune checkpoint surface. The analyzed CTLA-4 site represents an interface-adjacent permissive surface, enabling cross-interface comparison without implying direct blockade of the B7-1 binding site.[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eProtein preparation was performed using AutoDockTools following standard protocols.[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] Detailed preparation steps are provided in Online Resource 1.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Ligand preparation\u003c/h2\u003e \u003cp\u003eSix polyphenolic compounds were selected based on structural diversity and reported immunomodulatory relevance: gnetin C, quercetin, resveratrol, kaempferol, epigallocatechin gallate (EGCG), and curcumin. Ligand structures were retrieved from the PubChem database in SDF format.[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eLigand structures were prepared and energy-minimized using standard workflows. Full ligand preparation procedures are described in Online Resource 1.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Binding pocket identification and molecular docking\u003c/h2\u003e \u003cp\u003ePotential ligand-binding cavities were identified using CB-Dock2, an automated docking platform that detects protein cavities and estimates optimal grid centers based on structural topology.[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] For both CTLA-4 and PD-L1, the highest-ranked cavity was selected for subsequent docking analysis.\u003c/p\u003e \u003cp\u003eDocking simulations were performed using AutoDock Vina implemented through PyRx.[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eDocking parameters were set as follows: exhaustiveness\u0026thinsp;=\u0026thinsp;24, number of output binding modes\u0026thinsp;=\u0026thinsp;9, and the AutoDock Vina v1.2.3 scoring function applied. All docking calculations were performed using a fixed random seed to ensure reproducibility across independent docking runs.[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e], [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eGrid centers and dimensions for CTLA-4 and PD-L1 docking were defined based on CB-Dock2 cavity detection and are provided in Online Resource 1.\u003c/p\u003e \u003cp\u003eFor robustness assessment, each ligand was re-docked using five independent random seeds while keeping the interface-focused grid and all other docking parameters unchanged.\u003c/p\u003e \u003cp\u003eFor each ligand, mode-1, corresponding to the lowest predicted binding free energy, was selected for further analysis.\u003c/p\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.3.1 Ligand selection for comparative analysis\u003c/h2\u003e \u003cp\u003eLigand prioritization was guided by docking-derived evidence rather than binding affinity alone. All six polyphenols were initially docked against CTLA-4 and PD-L1 using identical protocols. Gnetin C consistently exhibited the most favorable binding energies and reproducible top-ranked docking poses across both targets.\u003c/p\u003e \u003cp\u003eQuercetin and resveratrol were retained for comparative analysis based on reproducible docking poses and consistent occupation of the primary binding pocket across independent docking runs. Importantly, these ligands occupied overlapping regions of the CTLA-4 and PD-L1 binding cavities, supporting comparable binding modes rather than nonspecific or peripheral interactions.[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e], [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eFor each ligand\u0026ndash;target pair, multiple docking poses were generated in a single docking run, and all output poses were retained for comparative analysis. Binding energies reported in the main text correspond to the top-ranked pose (mode-1), while mean binding energies across all generated poses are provided in Online Resource 1 to reflect overall pose stability rather than single-score optimization.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Docking result validation\u003c/h2\u003e \u003cp\u003eDocking validation was performed using a combination of pose reproducibility analysis, benchmark redocking of a co-crystallized PD-L1 interface ligand, and decoy-based enrichment testing to assess scoring robustness.\u003c/p\u003e \u003cp\u003ePose reproducibility was evaluated across independent docking runs using RMSD-based comparison. Detailed procedures are described in Online Resource 1. Given the shallow and solvent-exposed nature of protein\u0026ndash;protein interaction interfaces, RMSD values in the range of ~\u0026thinsp;2\u0026ndash;3 \u0026Aring; were considered acceptable.[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e], [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section3\"\u003e \u003ch2\u003e2.4.1 Interpretation boundaries of docking validation\u003c/h2\u003e \u003cp\u003eAll docking validation procedures were intended to support structural plausibility and reproducibility of docking placement rather than prediction of binding affinity, binding persistence, or functional inhibition.[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e], [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Interaction analysis\u003c/h2\u003e \u003cp\u003eLigand\u0026ndash;protein interactions were analyzed using standard visualization tools. Detailed interaction classification criteria are provided in Online Resource 1.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e2.6 Residue footprint and hotspot consensus analysis\u003c/h2\u003e \u003cp\u003eResidue\u0026ndash;ligand contact sets were extracted from the top-ranked docking pose (mode-1) using a distance-based criterion consistent with protein\u0026ndash;protein interface analyses (heavy-atom distance\u0026thinsp;\u0026le;\u0026thinsp;4.0 \u0026Aring;). Binary residue footprints were constructed for each ligand\u0026ndash;target pair based on the presence or absence of residue-level contacts. Pairwise Jaccard similarity indices were computed to quantify overlap between ligand interaction footprints. Consensus hotspot residues were defined as residues contacted by at least two ligands within the same target interface.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e2.7 Structural mapping of CTLA-4 docking site relative to the B7-1 interface\u003c/h2\u003e \u003cp\u003eTo assess the spatial relationship between the predicted CTLA-4 docking site and the physiological B7-1 binding interface, the CTLA-4/B7-1 complex structure was used as a reference.[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e] Docking-interacting residues identified from CTLA-4 docking analysis were mapped onto the CTLA-4 structure and visually compared with the B7-1 interaction surface using molecular visualization tools. This analysis was performed to determine whether the docking site overlapped with, or was adjacent to, the B7-1 interface and to classify the site as an exploratory interface-compatible surface rather than a direct blockade site.[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e], [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e2.8 Molecular dynamics simulation of the PD-L1 dimer\u003c/h2\u003e \u003cp\u003eTo assess the intrinsic structural integrity of the PD-L1 dimer interface, an apo molecular dynamics (MD) simulation was performed. The PD-L1 dimer structure (PDB ID: 5J89), retaining both protomers, was used as the initial model to preserve the native dimeric interface architecture. The simulation was designed to characterize inter-protomer behavior and interface stability rather than ligand dynamics or binding persistence.\u003c/p\u003e \u003cp\u003eMD simulations were carried out using GROMACS, a widely used molecular simulation engine optimized for biomolecular systems.[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e] The protein was parameterized using the CHARMM36m force field, which provides improved accuracy for folded and flexible protein regions.[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eAn unrestrained production MD simulation was performed for 10 ns with a 2 fs integration time step. Long-range electrostatic interactions were treated using the particle mesh Ewald (PME) method,[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e] and all covalent bonds involving hydrogen atoms were constrained using the LINCS algorithm. Interface-level analyses reported in this manuscript were computed from the available trajectory segment generated during post-processing. The 10-ns simulation length was selected to characterize short-timescale interface integrity and inter-protomer behavior rather than long-term conformational transitions or ligand residence dynamics.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e2.9 Molecular dynamics trajectory analysis\u003c/h2\u003e \u003cp\u003eTrajectory analyses were performed using standard GROMACS analysis utilities. Structural stability of the PD-L1 dimer was evaluated by calculating the root-mean-square fluctuation (RMSF) of backbone atoms, computed separately for each protomer to assess potential asymmetry in flexibility.\u003c/p\u003e \u003cp\u003eHydrogen bond analysis was performed to determine the number and temporal persistence of inter-protomer hydrogen bonds during the simulation, providing further insight into interface stabilization. All reported inter-protomer interface metrics (contacts, minimum distance, and hydrogen bonds) were computed over the current analyzed segment. All molecular dynamics analyses were computed over the full 10-ns production trajectory.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e2.10 Docking\u0026ndash;MD structural integration\u003c/h2\u003e \u003cp\u003eTo provide structural context for docking predictions, the top-ranked docking pose for PD-L1 was overlaid onto a representative MD snapshot of the PD-L1 dimer. Backbone alignment was performed to enable qualitative comparison between the docking-predicted ligand placement and the MD-derived dimer interface geometry.\u003c/p\u003e \u003cp\u003eThis integration step was intended for structural interpretation at the interface level and does not constitute validation of ligand binding stability, binding persistence, or inhibitory activity.[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e] No ligand-bound MD simulations or binding free energy calculations were performed. The representative MD snapshot was selected from the equilibrated portion of the analyzed segment to support qualitative interface-level interpretation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e2.11 Target prediction and network construction\u003c/h2\u003e \u003cp\u003ePutative molecular targets of the selected polyphenolic compounds were predicted using SwissTargetPrediction,[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e] with analyses restricted to Homo sapiens. For each ligand, the top-ranked predicted targets were retained and merged into a unified target set for subsequent network analysis.\u003c/p\u003e \u003cp\u003eProtein\u0026ndash;protein interaction (PPI) networks were constructed using STRING,[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e] limited to Homo sapiens and first-shell interactors. A minimum confidence score of 0.4 was applied, incorporating experimentally validated interactions, curated database evidence, and co-expression data, while excluding text-mining\u0026ndash;derived associations. CTLA-4 and PD-L1 were retained as seed proteins to anchor the network around immune checkpoint\u0026ndash;related interactions.\u003c/p\u003e \u003cp\u003ePutative molecular targets were predicted using SwissTargetPrediction, a ligand-based target inference platform. Predictions were restricted to \u003cem\u003eHomo sapiens\u003c/em\u003e. Approximately 10\u0026ndash;20 top-ranked targets per ligand were retained and combined as a union set for network analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e2.12 Network topology and functional enrichment analysis\u003c/h2\u003e \u003cp\u003eNetwork topology analysis was performed using STRING and Cytoscape,[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e] with node degree used as the primary centrality metric to identify highly connected hub proteins. The ten nodes with the highest degree values were designated as hub nodes for downstream interpretation.\u003c/p\u003e \u003cp\u003eFunctional enrichment analysis was conducted using the STRING enrichment module, including Gene Ontology Biological Process, KEGG, and Reactome pathway analyses. Enrichment results were evaluated against the Homo sapiens genome background, with multiple testing correction applied using the Benjamini\u0026ndash;Hochberg method and a false discovery rate threshold of \u0026lt;\u0026thinsp;0.05. Enrichment analyses were intended to provide pathway-level context for predicted target convergence rather than to infer direct biological modulation.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. RESULTS","content":"\u003cp\u003eDocking- and structure-derived observations reported in this section describe interface-level spatial compatibility and reproducibility within the applied computational framework and do not constitute evidence of immune checkpoint inhibition or functional immunomodulation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.1 Docking performance of polyphenols on CTLA-4\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMode-1 docking affinities for CTLA-4 spanned a narrow range (\u0026minus;8.0 to \u0026minus;6.1 kcal/mol), enabling consistent ranking across ligands (Table 1). Gnetin C ranked most favorably, followed by a middle tier including curcumin and epigallocatechin gallate, while resveratrol, quercetin, and kaempferol exhibited lower affinities. Top-ranked poses were reproducibly identified across repeated docking runs without rank inversion, indicating stable docking behavior for CTLA-4.\u003c/p\u003e\n\u003cp\u003eThe predicted CTLA-4 binding site was subsequently mapped against the physiological B7-1 interaction surface to contextualize the docking site within known structural constraints (see Section 3.2.1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1.\u003c/strong\u003e Mode-1 docking affinities of polyphenolic compounds toward CTLA-4.\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLigand\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBinding affinity (kcal/mol)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eGnetin C\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;8.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eEGCG\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;6.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eCurcumin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;6.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eResveratrol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;6.3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eQuercetin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;6.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eKaempferol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;6.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003e3.2 Interaction profiling of CTLA-4\u0026ndash;ligand complexes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll three representative ligands occupied the same CTLA-4 binding pocket with overlapping orientations. Interaction profiles converged on a shared set of interface-adjacent residues, including LYS56, LYS78, ASP34, and ASP79, indicating consistent pocket usage rather than alternative-site binding. Differences in interaction density among ligands reflected relative docking ranks but did not alter overall binding location (Figure 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.2.1 Structural positioning of the CTLA-4 docking site relative to the B7-1 interface\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMapping of docking-interacting residues onto the CTLA-4/B7-1 complex structure revealed that the predicted ligand-binding cavity is spatially adjacent to, but does not fully overlap with, the B7-1 binding interface. Key residues recurrently involved in docking interactions (LYS56, LYS78, ASP34, ASP79, TYR83) localized near the interface boundary rather than the central B7-1 contact region. This spatial relationship supports classification of the docking site as an interface-compatible exploratory surface rather than a direct blockade site.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.3 Residue-level footprint overlap and hotspot consensus analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eResidue-level footprint analysis was performed to quantify ligand interaction convergence at the CTLA-4 and PD-L1 interfaces using docking-derived contact sets. At the PD-L1 dimer interface, quantifiable overlap was observed among ligand interaction profiles. Pairwise Jaccard similarity indices ranged from 0.43 to 0.80, indicating convergence toward a shared interaction region rather than diffuse surface association (Table 2).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2.\u003c/strong\u003e Pairwise Jaccard similarity indices of ligand interaction footprints.\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTarget\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLigand pair\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eJaccard index\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003ePD-L1 dimer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eGnetin C \u0026ndash; Quercetin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003ePD-L1 dimer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eGnetin C \u0026ndash; Resveratrol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e0.43\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003ePD-L1 dimer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eQuercetin \u0026ndash; Resveratrol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eCTLA-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eGnetin C \u0026ndash; Quercetin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e0.25\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eCTLA-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eGnetin C \u0026ndash; Resveratrol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e0.50\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eCTLA-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eQuercetin \u0026ndash; Resveratrol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e0.33\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eConsensus hotspot analysis further identified a subset of recurrent interface residues (Table 3). These findings provide quantitative support for interface-level permissiveness and suggest dual-interface compatibility of polyphenolic scaffolds. The observed clustering highlights conserved interaction hotspots at the PD-L1 dimer interface and lower but non-random convergence at the CTLA-4 interface-adjacent surface (Figure 2).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3.\u003c/strong\u003e Consensus hotspot residues identified across representative ligands.\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTarget\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eResidue\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLigands involved\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFrequency\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003ePD-L1 dimer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eTyr56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e3/3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003ePD-L1 dimer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eAla121\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e3/3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003ePD-L1 dimer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eMet115\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e3/3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003ePD-L1 dimer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eAsp122\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e2/3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e67%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003ePD-L1 dimer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eTyr123\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e2/3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e67%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eCTLA-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eLys78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e3/3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e100%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eCTLA-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eLys56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e2/3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e67%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eCTLA-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eAsp34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e2/3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e67%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eCTLA-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eTyr83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e2/3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e67%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eCTLA-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003eAsp79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e2/3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 132px;\"\u003e\n \u003cp\u003e67%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003e3.4 Docking performance of ligands on PD-L1 dimer\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll ligands converged to the same PD-L1 dimer interface pocket across independent docking runs, consistent with benchmark redocking and pose distribution analyses as shown in Online Resource 1. Gnetin C ranked most favorably within this interface, followed by resveratrol and quercetin, while the remaining polyphenols clustered at lower ranks (Table 4). Preservation of the dimeric architecture altered ligand accommodation relative to monomeric PD-L1, highlighting the structural relevance of the inter-protomer interface.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 4\u003c/strong\u003e. Mode-1 binding affinities of all screened polyphenols against the PD-L1 dimer.\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLigand\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBinding affinity (kcal/mol)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eGnetin C\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;8.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eKaempferol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;7.9\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eResveratrol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;7.5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eEGCG\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;7.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eCurcumin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;7.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003eQuercetin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 264px;\"\u003e\n \u003cp\u003e\u0026minus;7.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003e3.5 Interaction profiling of PD-L1 dimer\u0026ndash;ligand complexes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eInteraction profiling revealed that all representative ligands occupied a conserved hotspot at the PD-L1 dimer interface, engaging residues contributed by both protomers. Recurrent contacts involved Tyr56, Asp122, Ala121, and Met115, supporting a shared binding mechanism consistent with validated docking placement (Figure 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.6 Molecular dynamics characterization of the PD-L1 dimer interface\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo characterize the intrinsic structural behavior of the PD-L1 dimer interface independent of ligand binding, an apo molecular dynamics simulation was performed over a 10-ns production trajectory. Structural stability was evaluated using backbone flexibility, inter-protomer contact persistence, minimum inter-protomer distance, and hydrogen bond analysis (Figure 4).\u003c/p\u003e\n\u003cp\u003eBackbone root-mean-square fluctuation analysis revealed comparable flexibility profiles between protomers A and B, with low fluctuations observed across interface-forming residues relative to solvent-exposed regions. Inter-protomer contact analysis demonstrated sustained interface engagement throughout the simulation, with only minor temporal variation. Consistently, the minimum inter-protomer distance remained within a narrow range, indicating maintained proximity between the two chains.\u003c/p\u003e\n\u003cp\u003eHydrogen bond analysis further supported interface stability, with a small but persistent number of inter-protomer hydrogen bonds observed throughout the trajectory. No interface separation events or large-scale disruption were detected during the simulated timescale. Collectively, these observations indicate that the PD-L1 dimer interface remains structurally cohesive under apo conditions, providing a stable architectural context for interface-focused docking interpretation rather than evidence of ligand binding stability.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.7 Structural integration of docking and molecular dynamics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo provide structural context for docking predictions, the top-ranked docking pose of gnetin C was overlaid onto a representative snapshot extracted from the equilibrated molecular dynamics trajectory of the PD-L1 dimer. Backbone alignment was performed to enable qualitative comparison between docking-predicted ligand placement and the geometry of the MD-derived dimer interface.\u003c/p\u003e\n\u003cp\u003eThe overlay indicated that the docking-predicted ligand placement spatially overlapped with an interface region characterized by low backbone fluctuation and persistent inter-protomer contacts in the apo MD simulation. This observation suggests that docking convergence occurred within a pre-organized interface region rather than a highly flexible surface groove. This qualitative overlay is consistent with the benchmark redocking results, which independently confirmed convergence of docking poses to the same PD-L1 interface region (Online Resource 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.8 Predicted molecular targets of representative polyphenols\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePredicted molecular targets for gnetin C, quercetin, and resveratrol were obtained using SwissTargetPrediction and used to support subsequent network-based analyses. For each ligand, top-ranked predicted targets were retained based on probability scores and merged for comparative evaluation.\u003c/p\u003e\n\u003cp\u003eAcross the three ligands, target prediction revealed partial overlap, with several targets shared between at least two compounds and a smaller subset common to all three. Despite differences in individual target composition, the predicted target sets exhibited comparable distributions across major protein classes, including enzymes, kinases, and nuclear receptors (Table 5).\u003c/p\u003e\n\u003cp\u003eA subset of predicted targets was associated with immune-related signaling and regulatory pathways based on functional annotation. These shared targets provided a rationale for protein\u0026ndash;protein interaction network construction to examine connectivity patterns in relation to immune checkpoint proteins, without implying direct biological modulation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 5.\u003c/strong\u003e Predicted molecular target classes of gnetin C, quercetin, and resveratrol\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLigand\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDominant target classes\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eRepresentative targets\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eGnetin C\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eNuclear receptors, kinases, enzymes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eESR1, ESR2, SRC, MAPK1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eQuercetin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eEnzymes, nuclear receptors, kinases\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003ePTGS2, ESR1, MAPK1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eResveratrol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eNuclear receptors, enzymes, kinases\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 176px;\"\u003e\n \u003cp\u003eESR1, SIRT1, MAPK1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u003cstrong\u003e3.9 Protein\u0026ndash;protein interaction network analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eProtein\u0026ndash;protein interaction network analysis was performed to explore connectivity among predicted molecular targets in relation to immune checkpoint proteins. Separate networks were constructed for CTLA-4\u0026ndash;centered and PD-L1\u0026ndash;centered target sets using STRING, with CTLA-4 and CD274 (PD-L1) included as seed nodes.\u003c/p\u003e\n\u003cp\u003eBoth networks exhibited statistically significant enrichment of protein\u0026ndash;protein interactions relative to random expectation, indicating non-random connectivity within the predicted target space. In each network, the immune checkpoint protein occupied the dominant hub position, with secondary connectivity observed among immune-related signaling components.\u003c/p\u003e\n\u003cp\u003eThe PD-L1\u0026ndash;centered network displayed higher overall connectivity compared with the CTLA-4\u0026ndash;centered network, consistent with broader interaction context associated with the PD-L1 dimer interface. These network characteristics provide systems-level context for the convergence of predicted targets around immune checkpoint\u0026ndash;associated interaction landscapes and are not intended to infer functional immune modulation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.10 Functional enrichment results\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFunctional enrichment analysis was conducted separately for CTLA-4\u0026ndash;centered and PD-L1\u0026ndash;centered protein sets to provide pathway-level context for predicted target connectivity. Enrichment significance was assessed using Benjamini\u0026ndash;Hochberg correction with a false discovery rate threshold of \u0026lt; 0.05.\u003c/p\u003e\n\u003cp\u003eFor the CTLA-4\u0026ndash;centered network, Gene Ontology Biological Process analysis identified enrichment of immune-related processes associated with T-cell signaling and costimulatory regulation. KEGG and Reactome pathway analyses further highlighted pathways involved in immune communication and lymphocyte-associated signaling. These enriched terms reflected functional coherence among CTLA-4 and its directly connected interactors rather than isolated pathway associations.\u003c/p\u003e\n\u003cp\u003eFunctional enrichment of the PD-L1\u0026ndash;centered network revealed a partially overlapping but broader enrichment profile. Gene Ontology terms were enriched for immune signaling and receptor-mediated processes, while KEGG and Reactome analyses identified multiple signaling pathways consistent with the denser connectivity observed in the PD-L1\u0026ndash;centered network. The higher number of significantly enriched pathways reflected the increased interaction density within this network rather than differential biological effect.\u003c/p\u003e\n\u003cp\u003eAcross both networks, enrichment patterns supported non-random functional clustering of predicted targets around immune checkpoint\u0026ndash;associated signaling landscapes. These results provide systems-level context for the observed network structures and are not intended to imply direct modulation of immune pathways or therapeutic activity.\u003c/p\u003e"},{"header":"4. DISCUSSION","content":"\u003cp\u003eMost computational studies of small-molecule immune checkpoint modulation remain single-target and monomer-centric, primarily focusing on affinity ranking rather than interface-level permissiveness.\u003c/p\u003e \u003cp\u003eIn contrast, the present study reframes small-molecule engagement as an interface-compatibility problem across immune checkpoints rather than a target-specific inhibition problem. By jointly evaluating CTLA-4 and the PD-L1 dimer interface within a unified structural and systems-level framework, this work emphasizes spatial convergence, interface reuse, and architectural permissiveness rather than optimization of docking scores.\u003c/p\u003e \u003cp\u003eThis dual-interface perspective does not aim to identify potent inhibitors but instead proposes a structural hypothesis in which certain chemical scaffolds are inherently compatible with multiple immune checkpoint interfaces. Such compatibility may represent a prerequisite for rational interface-targeted ligand optimization rather than a surrogate for biological activity.\u003c/p\u003e \u003cdiv id=\"Sec33\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Binding behavior of polyphenols on immune checkpoint proteins\u003c/h2\u003e \u003cp\u003eThis study shows that selected polyphenolic compounds exhibit reproducible docking compatibility with two immune checkpoint proteins, CTLA-4 and PD-L1, within the applied computational framework. Across both targets, docking results consistently ranked gnetin C as the most favorable ligand, while quercetin and resveratrol displayed intermediate or architecture-dependent docking performance.\u003c/p\u003e \u003cp\u003eAlthough CTLA-4 is classically discussed in the context of its interaction with B7 family ligands, the present analysis does not aim to infer direct functional blockade. Instead, CTLA-4 is considered here as a structurally distinct immune checkpoint surface used for cross-interface comparison. From a structural perspective, immune checkpoint proteins expose multiple solvent-accessible regions beyond their canonical ligand-binding interfaces, allowing small molecules to engage interface-adjacent or permissive surfaces without necessarily disrupting physiological protein\u0026ndash;protein interactions. In this context, the convergence of similar polyphenolic scaffolds on interface-adjacent regions of both PD-L1 and CTLA-4 supports a model of scaffold-level interface compatibility rather than target-specific inhibition.[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e], [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eImportantly, the reported docking affinities should not be interpreted as absolute predictors of biological activity. Rather, they reflect relative compatibility between ligand physicochemical features and receptor surface topology. Polyphenols are characterized by extended aromatic systems and multiple hydrogen bond donors and acceptors, properties that favor interactions with shallow, solvent-exposed protein surfaces rather than deep catalytic pockets. Immune checkpoint proteins, which function primarily through protein\u0026ndash;protein interactions, present such interfaces as potential sites for small-molecule engagement.\u003c/p\u003e \u003cp\u003eThis binding paradigm aligns with growing evidence that modulation of protein\u0026ndash;protein interfaces by small molecules is feasible when ligands exploit distributed hydrophobic contacts, aromatic stacking, and electrostatic anchoring instead of classical lock-and-key complementarity.[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e], [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] Several recent studies have highlighted that docking against protein\u0026ndash;protein interaction surfaces yields narrower affinity ranges and higher pose degeneracy compared with enzyme active sites, emphasizing the importance of reproducibility and spatial convergence rather than absolute docking scores.[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e], [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec34\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Structural determinants of PD-L1 dimer interface engagement\u003c/h2\u003e \u003cp\u003eA central finding of this work is that preservation of the PD-L1 dimer architecture reveals a conserved interface pocket that is preferentially occupied by all representative ligands in docking analyses. Docking consistently identified residues such as Tyr56, Asp122, Ala121, and Met115 as recurrent contributors to ligand accommodation at the dimer interface, consistent with the presence of a structurally relevant interaction hotspot rather than a docking artifact.\u003c/p\u003e \u003cp\u003eMolecular dynamics analysis further indicated that this interface region constitutes a structurally maintained element of the PD-L1 dimer within the simulated timescale. Reduced backbone flexibility at the interface relative to terminal and solvent-exposed regions, together with persistent inter-protomer contacts and hydrogen bonds, suggests that the identified pocket is pre-organized and structurally maintained under the simulated conditions. The recurrence of identical interface residues across chemically distinct ligands further argues against stochastic surface docking and supports the presence of a structurally permissive interface hotspot.\u003c/p\u003e \u003cp\u003ePrevious structural and biophysical studies have demonstrated that PD-L1 dimerization can be stabilized by small molecules and represents a functionally distinct structural state compared with the monomer.[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e], [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e] Targeting this interface has been proposed as a potential indirect strategy to influence PD-1 recognition through steric or conformational effects, as suggested by prior structural studies. The improved docking rank of resveratrol in the dimeric PD-L1 context compared with monomer-based docking further underscores the sensitivity of ligand accommodation to receptor oligomeric state.\u003c/p\u003e \u003cp\u003eAmong the investigated ligands, gnetin C formed the most extensive interaction network spanning both protomers, consistent with its favorable docking rank within the applied protocol. Quercetin and resveratrol exhibited progressively simpler interaction profiles, suggesting that molecular size, aromatic density, and spatial reach influence the capacity of polyphenols to bridge the PD-L1 dimer interface effectively.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec35\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Integration of docking and molecular dynamics insights\u003c/h2\u003e \u003cp\u003eThe molecular dynamics analysis was designed to provide descriptive characterization of PD-L1 dimer interface integrity over the simulated timescale rather than statistical estimation across independent replicas. Although only a single apo trajectory was analyzed, multiple independent interface metrics exhibited consistent behavior across consecutive simulation windows with available data.\u003c/p\u003e \u003cp\u003eWithin these constraints, the MD results support interface-level interpretation of docking predictions by defining the structural environment of the receptor rather than validating ligand binding stability or functional modulation.[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e] While longer and replicated simulations would be required to capture rare conformational events, the present timescale is sufficient to assess baseline interface cohesion and structural maintenance.\u003c/p\u003e \u003cp\u003eImportantly, quantitative footprint overlap and hotspot consensus analyses indicate that ligand accommodation across CTLA-4 and PD-L1 interfaces is driven by shared residue hotspots rather than stochastic docking placement. The higher degree of footprint convergence observed at the PD-L1 dimer interface relative to CTLA-4 further supports the role of receptor architecture in shaping interface permissiveness, consistent with the conceptual framing of scaffold-level compatibility rather than direct functional inhibition.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec36\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Network-level interpretation of dual immune checkpoint modulation\u003c/h2\u003e \u003cp\u003eBeyond direct protein\u0026ndash;ligand interactions, network analysis revealed that predicted molecular targets of the investigated polyphenols converge on immune-related signaling nodes associated with T-cell regulation, costimulatory signaling, and immune checkpoint pathways. In both CTLA-4\u0026ndash; and PD-L1\u0026ndash;centered networks, the immune checkpoint proteins emerged as dominant hubs, reinforcing their relevance as focal points for multi-target modulation.\u003c/p\u003e \u003cp\u003eThe convergence of docking-supported targets with network-identified central nodes supports a systems-level interpretation in which predicted ligand targets map onto interconnected immune-related pathways rather than isolated proteins. These network features reflect topological properties derived from the selected interaction database and do not imply direct modulation of all connected nodes by the investigated compounds.\u003c/p\u003e \u003cp\u003eWithin this framework, dual consideration of CTLA-4\u0026ndash; and PD-L1\u0026ndash;associated networks provides a conceptual basis for exploring pathway-level relationships between early T-cell priming and peripheral immune regulation, offering a mechanistic rationale for the dual-target hypothesis explored in this study.[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e], [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eNetwork-level analysis was applied as a contextualization layer to relate structurally compatible scaffolds to pathway-adjacent protein sets, rather than as evidence of direct multi-target modulation. Within this constraint, network convergence highlights biological neighborhoods potentially compatible with the identified scaffold architectures.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec37\" class=\"Section2\"\u003e \u003ch2\u003e4.5 Implications for immunotherapy development\u003c/h2\u003e \u003cp\u003eCurrent immune checkpoint therapies rely predominantly on monoclonal antibodies, which offer high specificity but are associated with limitations such as immune-related adverse events, high production costs, and restricted tissue penetration. In particular, antibody-based checkpoint blockade primarily targets extracellular interactions, whereas small molecules may enable alternative modes of interface engagement or allosteric modulation. Small-molecule modulators represent a complementary strategy, potentially enabling oral administration, tunable pharmacokinetics, and combinatorial use with antibody-based therapies.[\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e], [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eAlthough the present study does not claim direct inhibitory or therapeutic activity, the structural evidence presented here supports the feasibility of small-molecule engagement of immune checkpoint interfaces. Polyphenolic scaffolds, in particular, may serve as starting points for rational optimization rather than final drug candidates. Their interaction patterns highlight interface-compatible structural motifs that could be refined to enhance interface compatibility and interaction selectivity.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec38\" class=\"Section2\"\u003e \u003ch2\u003e4.6 Study limitations and future directions\u003c/h2\u003e \u003cp\u003eSeveral limitations should be acknowledged. First, while molecular dynamics analysis was performed to assess intrinsic dimer stability, ligand-bound simulations and binding free energy calculations were not conducted. Consequently, conclusions regarding ligand stability over time remain inferential rather than quantitative. The observed consistency across independent docking seeds further supports that the identified PD-L1 dimer interface interactions are not artifacts of stochastic sampling.\u003c/p\u003e \u003cp\u003eSecond, all findings are derived from computational analyses and do not account for cellular context, protein expression levels, or downstream signaling effects. Docking scores and interaction profiles alone cannot predict immunomodulatory efficacy or safety.\u003c/p\u003e \u003cp\u003eLigand-bound molecular dynamics simulations were not performed in this study by design. The primary objective was to assess interface permissiveness and structural accommodation rather than ligand residence time, binding persistence, or inhibitory behavior.\u003c/p\u003e \u003cp\u003eBy decoupling interface stability from ligand dynamics, this study focuses on defining whether the PD-L1 dimer interface constitutes a pre-organized structural feature capable of accommodating chemically diverse scaffolds. Ligand-bound simulations and free energy calculations are necessary to evaluate binding stability and functional relevance but fall outside the scope of the present interface-level analysis.\u003c/p\u003e \u003cp\u003eFuture studies should incorporate ligand-bound molecular dynamics simulations, free energy estimations, and experimental validation using biochemical and cellular assays. Such efforts will be essential to determine whether interface-binding polyphenols can modulate immune checkpoint signaling in biologically meaningful ways. Accordingly, the findings presented here should be interpreted as structural and network-level hypotheses that require experimental validation to establish biological relevance.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. CONCLUSION","content":"\u003cp\u003eThis study presents an integrated in silico framework to examine polyphenolic scaffolds in relation to two immune checkpoint proteins, CTLA-4 and PD-L1, by combining molecular docking, molecular dynamics\u0026ndash;based structural characterization, and network-level analysis.\u003c/p\u003e \u003cp\u003eDocking analyses demonstrated reproducible and spatially consistent interface-level accommodation of gnetin C, quercetin, and resveratrol across both targets, with gnetin C consistently ranked most favorably within the applied protocol. Preservation of the PD-L1 dimer architecture revealed a conserved inter-protomer interface pocket shared across ligands, highlighting the importance of receptor oligomeric state in shaping small-molecule interface compatibility.\u003c/p\u003e \u003cp\u003eApo molecular dynamics analysis was used to characterize intrinsic PD-L1 dimer interface integrity rather than ligand behavior. Persistent inter-protomer contacts, maintained hydrogen bonding, and reduced backbone flexibility at the interface support its classification as a structurally maintained and pre-organized architectural feature, providing context for interface-focused docking interpretation without assessing ligand binding stability or functional modulation.\u003c/p\u003e \u003cp\u003eAt the systems level, network analysis indicated convergence of predicted molecular targets on immune-related signaling nodes, with CTLA-4 and PD-L1 emerging as central hubs. This observation supports a dual-target concept framed in terms of structural compatibility and pathway connectivity rather than direct inhibitory or immunomodulatory activity.\u003c/p\u003e \u003cp\u003eOverall, this work defines interface-level structural compatibility of selected polyphenolic scaffolds with immune checkpoint proteins and provides a structure-informed, hypothesis-generating framework to guide future ligand optimization and experimental validation, without implying immune checkpoint inhibition or therapeutic efficacy.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eCompeting Interests\u003c/h2\u003e\n\u003cp\u003eThe authors have no relevant financial or non-financial interests to disclose.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eThe authors declare that no funds, grants, or other support were received during the preparation of this manuscript.\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\n\u003cp\u003eMRPPS conceived the study, designed the computational workflow, performed data collection and analysis, and wrote the manuscript.\u003c/p\u003e\n\u003ch2\u003eAcknowledgments\u003c/h2\u003e\n\u003cp\u003eThe authors thank colleagues for constructive discussions and technical assistance related to computational analysis.\u003c/p\u003e\n\u003ch2\u003eData Availability\u003c/h2\u003e\n\u003cp\u003eThe datasets generated during and/or analysed during the current study, including molecular docking outputs, molecular dynamics trajectories, and processed analysis files, are available from the corresponding author on reasonable request. Publicly available protein structures used in this study were obtained from the Protein Data Bank.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eS. C. Wei, C. R. Duffy, and J. P. Allison, \u0026ldquo;Fundamental Mechanisms of Immune Checkpoint Blockade Therapy,\u0026rdquo; \u003cem\u003eCancer Discovery\u003c/em\u003e, vol. 8, no. 9, pp. 1069\u0026ndash;1086, Sep. 2018, doi: 10.1158/2159-8290.CD-18-0367.\u003c/li\u003e\n\u003cli\u003eA. Ribas and J. D. Wolchok, \u0026ldquo;Cancer immunotherapy using checkpoint blockade,\u0026rdquo; \u003cem\u003eScience\u003c/em\u003e, vol. 359, no. 6382, pp. 1350\u0026ndash;1355, Mar. 2018, doi: 10.1126/science.aar4060.\u003c/li\u003e\n\u003cli\u003eD. M. Pardoll, \u0026ldquo;The blockade of immune checkpoints in cancer immunotherapy,\u0026rdquo; \u003cem\u003eNat Rev Cancer\u003c/em\u003e, vol. 12, no. 4, pp. 252\u0026ndash;264, Apr. 2012, doi: 10.1038/nrc3239.\u003c/li\u003e\n\u003cli\u003eP. Sharma, S. Hu-Lieskovan, J. A. Wargo, and A. Ribas, \u0026ldquo;Primary, Adaptive, and Acquired Resistance to Cancer Immunotherapy,\u0026rdquo; \u003cem\u003eCell\u003c/em\u003e, vol. 168, no. 4, pp. 707\u0026ndash;723, Feb. 2017, doi: 10.1016/j.cell.2017.01.017.\u003c/li\u003e\n\u003cli\u003eM. A. Postow, R. Sidlow, and M. D. Hellmann, \u0026ldquo;Immune-Related Adverse Events Associated with Immune Checkpoint Blockade,\u0026rdquo; \u003cem\u003eN Engl J Med\u003c/em\u003e, vol. 378, no. 2, pp. 158\u0026ndash;168, Jan. 2018, doi: 10.1056/NEJMra1703481.\u003c/li\u003e\n\u003cli\u003eF. Martins \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Adverse effects of immune-checkpoint inhibitors: epidemiology, management and surveillance,\u0026rdquo; \u003cem\u003eNat Rev Clin Oncol\u003c/em\u003e, vol. 16, no. 9, pp. 563\u0026ndash;580, Sep. 2019, doi: 10.1038/s41571-019-0218-0.\u003c/li\u003e\n\u003cli\u003eM. R. Arkin, Y. Tang, and J. A. Wells, \u0026ldquo;Small-Molecule Inhibitors of Protein-Protein Interactions: Progressing toward the Reality,\u0026rdquo; \u003cem\u003eChemistry \u0026amp; Biology\u003c/em\u003e, vol. 21, no. 9, pp. 1102\u0026ndash;1114, Sep. 2014, doi: 10.1016/j.chembiol.2014.09.001.\u003c/li\u003e\n\u003cli\u003eD. E. Scott, A. R. Bayly, C. Abell, and J. Skidmore, \u0026ldquo;Small molecules, big targets: drug discovery faces the protein\u0026ndash;protein interaction challenge,\u0026rdquo; \u003cem\u003eNat Rev Drug Discov\u003c/em\u003e, vol. 15, no. 8, pp. 533\u0026ndash;550, Aug. 2016, doi: 10.1038/nrd.2016.29.\u003c/li\u003e\n\u003cli\u003eA. Mullard, \u0026ldquo;Protein\u0026ndash;protein interaction inhibitors get into the groove,\u0026rdquo; \u003cem\u003eNat Rev Drug Discov\u003c/em\u003e, vol. 11, no. 3, pp. 173\u0026ndash;175, Mar. 2012, doi: 10.1038/nrd3680.\u003c/li\u003e\n\u003cli\u003eH. Cory, S. Passarelli, J. Szeto, M. Tamez, and J. Mattei, \u0026ldquo;The Role of Polyphenols in Human Health and Food Systems: A Mini-Review,\u0026rdquo; \u003cem\u003eFront. Nutr.\u003c/em\u003e, vol. 5, p. 87, Sep. 2018, doi: 10.3389/fnut.2018.00087.\u003c/li\u003e\n\u003cli\u003eC. G. Fraga, K. D. Croft, D. O. Kennedy, and F. A. Tom\u0026aacute;s-Barber\u0026aacute;n, \u0026ldquo;The effects of polyphenols and other bioactives on human health,\u0026rdquo; \u003cem\u003eFood Funct.\u003c/em\u003e, vol. 10, no. 2, pp. 514\u0026ndash;528, 2019, doi: 10.1039/C8FO01997E.\u003c/li\u003e\n\u003cli\u003eK. M. Zak \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Structure of the Complex of Human Programmed Death 1, PD-1, and Its Ligand PD-L1,\u0026rdquo; \u003cem\u003eStructure\u003c/em\u003e, vol. 23, no. 12, pp. 2341\u0026ndash;2348, Dec. 2015, doi: 10.1016/j.str.2015.09.010.\u003c/li\u003e\n\u003cli\u003eK. Guzik \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Development of the Inhibitors That Target the PD-1/PD-L1 Interaction\u0026mdash;A Brief Look at Progress on Small Molecules, Peptides and Macrocycles,\u0026rdquo; \u003cem\u003eMolecules\u003c/em\u003e, vol. 24, no. 11, p. 2071, May 2019, doi: 10.3390/molecules24112071.\u003c/li\u003e\n\u003cli\u003eL. Skalniak \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Small-molecule inhibitors of PD-1/PD-L1 immune checkpoint alleviate the PD-L1-induced exhaustion of T-cells,\u0026rdquo; \u003cem\u003eOncotarget\u003c/em\u003e, vol. 8, no. 42, pp. 72167\u0026ndash;72181, Sep. 2017, doi: 10.18632/oncotarget.20050.\u003c/li\u003e\n\u003cli\u003eA. L. Hopkins, \u0026ldquo;Network pharmacology: the next paradigm in drug discovery,\u0026rdquo; \u003cem\u003eNat Chem Biol\u003c/em\u003e, vol. 4, no. 11, pp. 682\u0026ndash;690, Nov. 2008, doi: 10.1038/nchembio.118.\u003c/li\u003e\n\u003cli\u003eU. Chandran, N. Mehendale, S. Patil, R. Chaguturu, and B. Patwardhan, \u0026ldquo;Network Pharmacology,\u0026rdquo; in \u003cem\u003eInnovative Approaches in Drug Discovery\u003c/em\u003e, Elsevier, 2017, pp. 127\u0026ndash;164. doi: 10.1016/B978-0-12-801814-9.00005-2.\u003c/li\u003e\n\u003cli\u003eH. M. Berman \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;The Protein Data Bank,\u0026rdquo; \u003cem\u003eNucleic Acids Research\u003c/em\u003e, vol. 28, no. 1, pp. 235\u0026ndash;242, Jan. 2000, doi: 10.1093/nar/28.1.235.\u003c/li\u003e\n\u003cli\u003eG. M. Morris \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;AutoDock4 and AutoDockTools4: Automated docking with selective receptor flexibility,\u0026rdquo; \u003cem\u003eJ Comput Chem\u003c/em\u003e, vol. 30, no. 16, pp. 2785\u0026ndash;2791, Dec. 2009, doi: 10.1002/jcc.21256.\u003c/li\u003e\n\u003cli\u003eS. Kim \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;PubChem 2023 update,\u0026rdquo; \u003cem\u003eNucleic Acids Research\u003c/em\u003e, vol. 51, no. D1, pp. D1373\u0026ndash;D1380, Jan. 2023, doi: 10.1093/nar/gkac956.\u003c/li\u003e\n\u003cli\u003eY. Liu, X. Yang, J. Gan, S. Chen, Z.-X. Xiao, and Y. Cao, \u0026ldquo;CB-Dock2: improved protein\u0026ndash;ligand blind docking by integrating cavity detection, docking and homologous template fitting,\u0026rdquo; \u003cem\u003eNucleic Acids Research\u003c/em\u003e, vol. 50, no. W1, pp. W159\u0026ndash;W164, Jul. 2022, doi: 10.1093/nar/gkac394.\u003c/li\u003e\n\u003cli\u003eO. Trott and A. J. Olson, \u0026ldquo;AutoDock Vina: Improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading,\u0026rdquo; \u003cem\u003eJ Comput Chem\u003c/em\u003e, vol. 31, no. 2, pp. 455\u0026ndash;461, Jan. 2010, doi: 10.1002/jcc.21334.\u003c/li\u003e\n\u003cli\u003eN. S. Pagadala, K. Syed, and J. Tuszynski, \u0026ldquo;Software for molecular docking: a review,\u0026rdquo; \u003cem\u003eBiophys Rev\u003c/em\u003e, vol. 9, no. 2, pp. 91\u0026ndash;102, Apr. 2017, doi: 10.1007/s12551-016-0247-1.\u003c/li\u003e\n\u003cli\u003eD. Kozakov \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;The FTMap family of web servers for determining and characterizing ligand-binding hot spots of proteins,\u0026rdquo; \u003cem\u003eNat Protoc\u003c/em\u003e, vol. 10, no. 5, pp. 733\u0026ndash;755, May 2015, doi: 10.1038/nprot.2015.043.\u003c/li\u003e\n\u003cli\u003eJ. A. Wells and C. L. McClendon, \u0026ldquo;Reaching for high-hanging fruit in drug discovery at protein\u0026ndash;protein interfaces,\u0026rdquo; \u003cem\u003eNature\u003c/em\u003e, vol. 450, no. 7172, pp. 1001\u0026ndash;1009, Dec. 2007, doi: 10.1038/nature06526.\u003c/li\u003e\n\u003cli\u003eC. C. Stamper \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Crystal structure of the B7-1/CTLA-4 complex that inhibits human immune responses,\u0026rdquo; \u003cem\u003eNature\u003c/em\u003e, vol. 410, no. 6828, pp. 608\u0026ndash;611, Mar. 2001, doi: 10.1038/35069118.\u003c/li\u003e\n\u003cli\u003eM. J. Abraham \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers,\u0026rdquo; \u003cem\u003eSoftwareX\u003c/em\u003e, vol. 1\u0026ndash;2, pp. 19\u0026ndash;25, Sep. 2015, doi: 10.1016/j.softx.2015.06.001.\u003c/li\u003e\n\u003cli\u003eJ. Huang \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;CHARMM36m: an improved force field for folded and intrinsically disordered proteins,\u0026rdquo; \u003cem\u003eNat Methods\u003c/em\u003e, vol. 14, no. 1, pp. 71\u0026ndash;73, Jan. 2017, doi: 10.1038/nmeth.4067.\u003c/li\u003e\n\u003cli\u003eT. Darden, D. York, and L. Pedersen, \u0026ldquo;Particle mesh Ewald: An \u003cem\u003eN\u003c/em\u003e \u0026sdot;log( \u003cem\u003eN\u003c/em\u003e ) method for Ewald sums in large systems,\u0026rdquo; \u003cem\u003eThe Journal of Chemical Physics\u003c/em\u003e, vol. 98, no. 12, pp. 10089\u0026ndash;10092, Jun. 1993, doi: 10.1063/1.464397.\u003c/li\u003e\n\u003cli\u003eP. V. Klimovich, M. R. Shirts, and D. L. Mobley, \u0026ldquo;Guidelines for the analysis of free energy calculations,\u0026rdquo; \u003cem\u003eJ Comput Aided Mol Des\u003c/em\u003e, vol. 29, no. 5, pp. 397\u0026ndash;411, May 2015, doi: 10.1007/s10822-015-9840-9.\u003c/li\u003e\n\u003cli\u003eA. Daina, O. Michielin, and V. Zoete, \u0026ldquo;SwissTargetPrediction: updated data and new features for efficient prediction of protein targets of small molecules,\u0026rdquo; \u003cem\u003eNucleic Acids Research\u003c/em\u003e, vol. 47, no. W1, pp. W357\u0026ndash;W364, Jul. 2019, doi: 10.1093/nar/gkz382.\u003c/li\u003e\n\u003cli\u003eD. Szklarczyk \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;The STRING database in 2021: customizable protein\u0026ndash;protein networks, and functional characterization of user-uploaded gene/measurement sets,\u0026rdquo; \u003cem\u003eNucleic Acids Research\u003c/em\u003e, vol. 49, no. D1, pp. D605\u0026ndash;D612, Jan. 2021, doi: 10.1093/nar/gkaa1074.\u003c/li\u003e\n\u003cli\u003eP. Shannon \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Cytoscape: A Software Environment for Integrated Models of Biomolecular Interaction Networks,\u0026rdquo; \u003cem\u003eGenome Res.\u003c/em\u003e, vol. 13, no. 11, pp. 2498\u0026ndash;2504, Nov. 2003, doi: 10.1101/gr.1239303.\u003c/li\u003e\n\u003cli\u003eH. Lu \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Recent advances in the development of protein\u0026ndash;protein interactions modulators: mechanisms and clinical trials,\u0026rdquo; \u003cem\u003eSig Transduct Target Ther\u003c/em\u003e, vol. 5, no. 1, p. 213, Sep. 2020, doi: 10.1038/s41392-020-00315-3.\u003c/li\u003e\n\u003cli\u003eH. Jubb, T. L. Blundell, and D. B. Ascher, \u0026ldquo;Flexibility and small pockets at protein\u0026ndash;protein interfaces: New insights into druggability,\u0026rdquo; \u003cem\u003eProgress in Biophysics and Molecular Biology\u003c/em\u003e, vol. 119, no. 1, pp. 2\u0026ndash;9, Oct. 2015, doi: 10.1016/j.pbiomolbio.2015.01.009.\u003c/li\u003e\n\u003cli\u003eQ. Wu, L. Jiang, S. Li, Q. He, B. Yang, and J. Cao, \u0026ldquo;Small molecule inhibitors targeting the PD-1/PD-L1 signaling pathway,\u0026rdquo; \u003cem\u003eActa Pharmacol Sin\u003c/em\u003e, vol. 42, no. 1, pp. 1\u0026ndash;9, Jan. 2021, doi: 10.1038/s41401-020-0366-x.\u003c/li\u003e\n\u003cli\u003eS. A. Hollingsworth and R. O. Dror, \u0026ldquo;Molecular Dynamics Simulation for All,\u0026rdquo; \u003cem\u003eNeuron\u003c/em\u003e, vol. 99, no. 6, pp. 1129\u0026ndash;1143, Sep. 2018, doi: 10.1016/j.neuron.2018.08.011.\u003c/li\u003e\n\u003cli\u003eD. S. Chen and I. Mellman, \u0026ldquo;Elements of cancer immunity and the cancer\u0026ndash;immune set point,\u0026rdquo; \u003cem\u003eNature\u003c/em\u003e, vol. 541, no. 7637, pp. 321\u0026ndash;330, Jan. 2017, doi: 10.1038/nature21349.\u003c/li\u003e\n\u003cli\u003eE. I. Buchbinder and A. Desai, \u0026ldquo;CTLA-4 and PD-1 Pathways: Similarities, Differences, and Implications of Their Inhibition,\u0026rdquo; \u003cem\u003eAmerican Journal of Clinical Oncology\u003c/em\u003e, vol. 39, no. 1, pp. 98\u0026ndash;106, Feb. 2016, doi: 10.1097/COC.0000000000000239.\u003c/li\u003e\n\u003cli\u003eV. A. Boussiotis, \u0026ldquo;Molecular and Biochemical Aspects of the PD-1 Checkpoint Pathway,\u0026rdquo; \u003cem\u003eN Engl J Med\u003c/em\u003e, vol. 375, no. 18, pp. 1767\u0026ndash;1778, Nov. 2016, doi: 10.1056/NEJMra1514296.\u003c/li\u003e\n\u003cli\u003eC. Feng \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Discovery of Small-Molecule PD-L1 Inhibitors via Virtual Screening and Their Immune-Mediated Anti-Tumor Effects,\u0026rdquo; \u003cem\u003ePharmaceuticals\u003c/em\u003e, vol. 18, no. 8, p. 1209, Aug. 2025, doi: 10.3390/ph18081209.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"PD-L1 dimer, Polyphenols, Network pharmacology, protein–protein interaction docking, interface-focused docking, scaffold compatibility","lastPublishedDoi":"10.21203/rs.3.rs-8841071/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8841071/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eContext\u003c/h2\u003e \u003cp\u003eProtein\u0026ndash;protein interaction interfaces of immune checkpoints present persistent challenges for molecular modeling due to shallow topology, limited pocket definition, and conformational flexibility. Although polyphenolic compounds have been widely explored in immune-related computational studies, their structural compatibility with immune checkpoint interface architectures remains poorly defined at the scaffold level. In this study, we investigated interface-focused structural compatibility of representative polyphenolic scaffolds at immune checkpoint proteins, with primary emphasis on the programmed death-ligand 1 (PD-L1) dimer interface and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) included as a structural comparator. The analysis reveals reproducible pose convergence and non-random residue footprint overlap at the PD-L1 dimer interface, supported by stable intrinsic interface architecture during apo-state molecular dynamics simulations. Network-based analysis further situates predicted ligand-associated targets within immune-related interaction neighborhoods, providing systems-level context consistent with a structurally permissive signaling environment. These findings characterize interface-level structural compatibility rather than functional immune checkpoint inhibition and generate testable hypotheses for subsequent experimental studies.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eInterface-preserving molecular docking was performed using AutoDock Vina with multi-seed sampling to assess spatial compatibility, pose convergence, and residue-level footprint overlap at immune checkpoint interfaces. Docking validation included redocking benchmarks and decoy-based evaluation. Molecular dynamics simulations of the apo PD-L1 dimer were conducted using GROMACS with a classical all-atom force field to characterize intrinsic interface stability, residue flexibility, interfacial contacts, and hydrogen-bond persistence. Predicted molecular targets of representative polyphenols were identified using similarity-based target prediction tools and analyzed through protein\u0026ndash;protein interaction network construction and topological analysis using Cytoscape. All computational workflows were executed using standard molecular modeling and network analysis software, with full methodological details provided in the main text and Online Resource 1.\u003c/p\u003e","manuscriptTitle":"Structural and systems-level analysis of polyphenolic scaffold compatibility at immune checkpoint interfaces: a PD-L1 dimer–focused in silico study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-11 18:32:35","doi":"10.21203/rs.3.rs-8841071/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"96018945-994d-45a0-a4ad-6d21ab97f662","owner":[],"postedDate":"March 11th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":64122960,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":64122961,"name":"Biological sciences/Drug discovery"}],"tags":[],"updatedAt":"2026-03-18T07:12:09+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-11 18:32:35","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8841071","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8841071","identity":"rs-8841071","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}