The protein journal最新文献

{"title":"Rationalization of Solvent-Mediated Enthalpy-Entropy Compensation with the Gibbs-Helmholtz Equation.","authors":"David J Scott, Donald J Winzor","doi":"10.1007/s10930-026-10358-0","DOIUrl":"https://doi.org/10.1007/s10930-026-10358-0","url":null,"abstract":"<p><p>This review examines the effect of solvent involvement in protein-ligand interactions on the extent of conformity with the Gibbs-Helmholtz equation, which was based on assumed inertness of the solvent-an assumption that rendered possible the evaluation of ΔH^o and TΔS^o by employing the integrated form of the van't Hoff isochore. Conformity with the Gibbs-Helmholtz equation is retained in situations where there is temperature dependence of the calorimetrically measured standard enthalpy change despite no change in standard free energy change (ΔΔG^o ≈ 0). Before the advent of isothermal titration calorimetry this situation was taken to signify that ΔH^o = 0 rather than (ΔH^o)_vH = 0, whereupon (ΔG^o)_vH = (TΔS^o)_vH. In isothermal titration calorimetry the measured standard enthalpy change (ΔH^o)_cal is a thermodynamic parameter for the whole system, whereupon values of (TΔS^o)_cal calculated as (ΔH^o)_cal-(ΔG^o)_vH include the van't Hoff enthalpy contribution (ΔH^o)_vH. A sufficiently large magnitude for that van't Hoff contribution leads to a Gibbs-Helmholtz plot with a slope greater than unity. A similar situation arises in systems exhibiting a slight but significant dependence of (ΔG^o)_vH upon (ΔH^o)_cal -examples of systems exhibiting substantial rather than extreme enthalpy-entropy compensation.</p>","PeriodicalId":94249,"journal":{"name":"The protein journal","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149303485","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hirudin Extraction of Iranian Leech Saliva by Freezing and Lyophilization Methods.","authors":"Pardis Khodakarami, Fariba Soltanolkottabi, Hojat Jafari","doi":"10.1007/s10930-026-10361-5","DOIUrl":"https://doi.org/10.1007/s10930-026-10361-5","url":null,"abstract":"<p><p>Leech saliva contains a diverse array of bioactive compounds, most notably hirudin, a potent natural thrombin inhibitor. Despite the clinical utility of hirudotherapy, inherent risks such as localized infection, uncontrolled hemorrhage, and leech migration limit its widespread application. To overcome these limitations, this study presents a refined extraction, partial purification, and stabilization protocol for hirudin-containing salivary extract derived from the Iranian leech, Hirudo orientalis. Saliva was harvested using 0.001 M arginine in 0.15 M saline under controlled temperatures (23 °C vs. 37-40 °C) and subsequently stored at 4 °C. The crude extract underwent fractionation/purification via centrifugation, followed by stabilization with sucrose, fructose, and mannitol, and final lyophilization (freeze-drying). Optimal extraction efficiency was observed at 37-40 °C. The short-term stability (44 days at 4 °C) and antibacterial efficacy of the resulting lyophilizate were evaluated using High-Performance Liquid Chromatography (HPLC) against authentic reference parameters and Brain Heart Infusion (BHI) broth assays. The findings demonstrate that this stabilization matrix preserves the short-term chromatographic profile and in-vitro biological activity of the salivary protein extract, providing a preliminary foundation for future in-vivo evaluation and development of standardized leech-derived biotherapeutics.</p>","PeriodicalId":94249,"journal":{"name":"The protein journal","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-10-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149279240","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Inhibition Kinetics of Protein Inhibitors of Carboxypeptidase-A and Cysteine Protease Isolated from Non-Host Plants of Helicoverpa armigera Against Gut Proteases.","authors":"Lunminlal Kipgen, Sabita Rangra, Kamal Krishan Aggarwal","doi":"10.1007/s10930-026-10362-4","DOIUrl":"https://doi.org/10.1007/s10930-026-10362-4","url":null,"abstract":"<p><p>Plant protease inhibitors have been suggested as one of the pest control strategies to control Helicoverpa armigera. To isolate protease inhibitors that are effective against H. armigera gut proteases, plants were screened for the presence of potent protease inhibitors after segregating them into host and non-host depending on whether H. armigera was reported to feed on it or not. A carboxypeptidase inhibitor and a cysteine protease inhibitor were isolated from non-host plants Oxalis corniculata (OcPI) and Withania somnifera (WsPI) respectively. Both inhibitors have K_i in the micromolar range, active between pH 2.6 to pH 10.6, and are stable up to 75 °C. At 3 inhibitor units per gram of diet (IU/g) each, OcPI caused higher mortality (24%) than WsPI (12%) or Bowman Birk (20%) serine protease inhibitor (BBI). However, higher mortality (40%) was observed when the three inhibitors were combined than when used individually suggesting an additive effect. Also, a similar additive effect in lengthening the duration of the larval stage for surviving larvae was observed.</p>","PeriodicalId":94249,"journal":{"name":"The protein journal","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-10-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149279282","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Interactions of Veterinary Antibiotics with Purified Bovine Milk Alkaline Phosphatase: A Combined In Vitro Kinetic and Theoretical DFT/HF/M06-2X Study.","authors":"Çiğdem Bilen, Burak Tüzün, Nahit Gencer","doi":"10.1007/s10930-026-10360-6","DOIUrl":"https://doi.org/10.1007/s10930-026-10360-6","url":null,"abstract":"<p><p>Alkaline phosphatase (AP, EC 3.1.3.1) is a key metalloenzyme and has been widely used as a reliable marker for milk pasteurization. Milk containing antibiotic residues tends to inhibit AP function, which can result in faulty pasteurization determination. AP from bovine milk was purified by a purification strategy involving butanol homogenization, rennet precipitation, and Sephadex G-100 gel filtration chromatography. The enzyme was purified 53.43-fold with a specific activity of 7516.40 EU/mg. Kinetic characterization showed a K_M value of 2.27 mM (V_max value: 555.56 EU/min); substrate was p-nitrophenyl phosphate. Electrophoretic studies verified the purity and molecular nature of the product enzyme. To functionally validate the identity of the purified enzyme, tetramisole hydrochloride was evaluated as a standard reference inhibitor, yielding a potent IC₅₀ value of 0.045 mM. Fifteen veterinary drugs and antibiotics were used to test the inhibitory effects of purified milk AP in vitro. Among the antibiotic compounds tested, Danofloxacin Mesylate induced the strongest antibiotic inhibition (IC₅₀: 3.36 mM). Gentamycin Sulfate, Metamizole Sodium, Enrofloxacin, Kanamycin Sulfate, Sulfadimidine Sodium, and Neomycin Sulfate had IC₅₀ values ranging between 22.89 and 312.13 mM of inhibition in AP, while Dexamethasone, Oxytetracycline Hydrochloride, Marbofloxacin, Tilmicosin, Doxycycline Monohydrate, Tylosin, Florfenicol and Flunixin Meglumine made no detected inhibition in the presence of 0.5 mM p-nitrophenyl phosphate. The molecular docking studies were additionally performed that the density functional theory (DFT), Hartree-Fock (HF), and M06-2X calculations were used to validate the experimental results with the 6-31++G(d,p) basis set. The theoretical results also reported that Oxytetracycline Hydrochloride, Tilmicosin, and Doxycycline Monohydrate displayed superior electronic reactivity, lower energy gaps between HOMO-LUMO, and more electrophilic character compared with the other compounds. These results are the first complete assessment of the impact of veterinary antibiotics in milk AP and underscore the possible role of antibiotic residues in establishing the reliability of pasteurization control measures.</p>","PeriodicalId":94249,"journal":{"name":"The protein journal","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-10-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149279303","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Computational Identification of IL-17 a Small-Molecule Inhibitors as Potential Therapeutics for Plaque Psoriasis Through e-Pharmacophore Screening, Molecular Docking, and Molecular Dynamics Simulation.","authors":"Selvakumar Rajendran, Anantha Krishnan Dhanabalan, Sriram Srinivasa Raghavan, Jebiti Haribabu, Arunachalam Arulraj, Varaprasad Kokkarachedu, Gunasekaran Krishnasamy","doi":"10.1007/s10930-026-10359-z","DOIUrl":"https://doi.org/10.1007/s10930-026-10359-z","url":null,"abstract":"<p><p>Plaque psoriasis is a long-term skin disease driven by the immune system, impacting around 1-3% of people worldwide. Despite the availability of various treatment options, the limitations in effectiveness, side effects, and long-term response underscore the necessity for enhanced therapeutic strategies. Focusing on the interleukin-17 (IL-17) pathway offers a promising strategy for treating plaque psoriasis. In this research, established co-crystallized peptide-derived inhibitors (PDB IDs: 5HI3, 5HI4, and 5HI5) were utilized to create energetically optimized pharmacophore models, which were then used for virtual screening of chemical libraries, resulting in 4,923 potential hit compounds. Molecular docking based on structure and induced-fit docking were subsequently conducted on IL-17 A to pinpoint potential lead compounds exhibiting advantageous binding orientations and interactions with essential active-site residues. The selected protein-ligand complexes' stability and dynamic behavior were examined through molecular dynamics (MD) simulations, which were complemented by root mean square deviation (RMSD), radius of gyration (RoG), principal component analysis (PCA), and dynamic cross-correlation matrix (DCCM) analyses. The MD simulations revealed that the selected leads retained consistent interactions in the IL-17 A binding pocket, with C2 and C3 demonstrating especially advantageous protein structural stability, while C3 displayed the most stable ligand behavior. PCA uncovered unique collective movements within the complexes, while DCCM analysis showed specific correlated and anti-correlated residue movements linked to protein conformational dynamics. Ongoing protein-ligand interactions further confirmed the stable placement of the chosen compounds within the binding site. In summary, the combined pharmacophore-based virtual screening, molecular docking, MD, PCA, and DCCM analyses discovered promising lead compounds exhibiting advantageous binding stability and dynamic behavior, offering a structural foundation for their subsequent experimental validation and potential advancement as therapeutic candidates for plaque psoriasis.</p>","PeriodicalId":94249,"journal":{"name":"The protein journal","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149169933","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From Helicobacter pylori CagA to Cross-Kingdom Amyloid Control: Emerging Anti-amyloid Biology, Mechanistic Paradoxes, and Translational Opportunities.","authors":"Yasser H Habib, Yasmine A Barakat","doi":"10.1007/s10930-026-10355-3","DOIUrl":"https://doi.org/10.1007/s10930-026-10355-3","url":null,"abstract":"<p><p>Amyloid formation contributes to a broad range of biological and pathological processes, from bacterial biofilm organization to human neurodegenerative and metabolic diseases. Consequently, molecules capable of inhibiting amyloid assembly across structurally unrelated proteins are exceptionally uncommon and mechanistically significant. Recent findings identifying the Helicobacter pylori virulence factor CagA as a potent broad-spectrum amyloid inhibitor establish an unexpected connection between microbial pathogenicity and protein misfolding biology. Experimental studies demonstrated that CagA and its active derivative, CagAN, suppress aggregation of both bacterial functional amyloids and multiple disease-associated human amyloid proteins, including amyloid-beta, tau, alpha-synuclein, and islet amyloid polypeptide. Mechanistic analyses further indicate that CagA interferes with different stages of amyloid assembly depending on the target substrate, including primary nucleation, secondary nucleation, and fibril elongation. Despite this promising anti-amyloid activity, CagA remains a pathogenic bacterial effector associated with epithelial disruption, oncogenic signaling, and gastric carcinogenesis. Therefore, its translational value likely lies not in direct therapeutic use, but in identifying and engineering safer anti-aggregation motifs derived from its structure. This review examines the dual biological identity of CagA, summarizes current mechanistic evidence supporting its anti-amyloid activity, and discusses future opportunities for developing detoxified peptides, biologics, and AI-guided anti-aggregation therapeutics inspired by CagA biology.</p>","PeriodicalId":94249,"journal":{"name":"The protein journal","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-10-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149121322","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Computational Validation of Tecomella undulata Bark as a Source of Anti-Filarial Leads Targeting UDP-Galactopyranose Mutase.","authors":"Shyam Baboo Prasad, Sk Aftabul Alam, Shrawan Kumar Sahu, Nikhilesh Joardar, Animesh Sen","doi":"10.1007/s10930-026-10357-1","DOIUrl":"https://doi.org/10.1007/s10930-026-10357-1","url":null,"abstract":"<p><p>Lymphatic filariasis remains a major neglected tropical disease and there is an urgent need for parasite-specific therapeutic targets. The enzyme UDP-galactopyranose mutase (UGM) is essential for galactofuranose biosynthesis in Brugia malayi and is absent in humans, which makes it a promising drug target. In this study, we investigated phytochemicals from the bark of Tecomella undulata to find potential drug candidates against B. malayi UGM using an in silico approach. The structure of the target protein was predicted using the Alphafold model, and the active site was predicted using the combination of PrankWeb, DoGSiteScorer, and FTMap servers and further supported by superposition with a homologous experimentally determined UGM structure. Forty phytochemicals of T. undulata bark were screened by molecular docking using AutoDock Tools. Docking analysis showed Dehydrotectol (- 11.73 kcal/mol), Daucosterol (- 11.35 kcal/mol), urs-12-ene (- 10.87 kcal/mol), β-sitosterol (- 10.72 kcal/mol), and betulinic acid (- 10.47 kcal/mol) are the top-scoring ligands. Based on predicted binding free energy and interaction with key catalytic residues of UGM, Daucosterol was selected for molecular dynamics simulation using the GROMACS package. MD analysis and MM/PBSA calculations supported the thermodynamic stability of the docked complex. ADMET predictions indicated acceptable drug-likeness and low toxicity of several phytoconstituents of T. undulata bark. This study provides computational evidence supporting T. undulata bark phytochemicals can be a new drug formulation against B. malayi.</p>","PeriodicalId":94249,"journal":{"name":"The protein journal","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149066932","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Integrated Structural and Dynamic Analysis Reveals Destabilizing Effects of KRAS Missense Variants Associated with Lung Cancer.","authors":"Md Sohel Mia, Bimal Kumar Datta, Khadija Akter, Tomal Krishno Das Topu, Jamima Mahzabin Shawon, Faisul Hossen, Abdur Rahman Pranto, Md Saklain Tanver Shadhin, Nowmee Tahsin, Safa H Qahl, Tahani Bakhsh, Md Enamul Kabir Talukder","doi":"10.1007/s10930-026-10356-2","DOIUrl":"https://doi.org/10.1007/s10930-026-10356-2","url":null,"abstract":"<p><p>Given the established oncogenic role of KRAS in lung cancer and its predicted regulation by hsa-miR-134 and transcription factors (NCOA4, NR3C2, ETS2), we systematically catalogued genetic variation across the KRAS gene. In this study, we identified 21,598 SNPs, of which 475 were missense (2.19%), 580 were synonymous (2.68%), and 18,093 were intronic (83.77%). In silico prioritization combined with cross-validation in TCGA-LUAD and MSK-IMPACT cohorts identified four high-confidence deleterious variants (G13C, G13D, G60S, G60V), confirming the recurrence of G13C/G13D and highlighting the fully conserved switch II residue G60 as a structurally distinctive candidate. In MD simulations, G60S showed the largest conformational displacement (RMSD ≈ 8-9 Å), elevated per-residue flexibility (residues 28-37: 1.90 → 2.99 Å), increased terminal Rg, and reduced hydrogen-bond occupancy, suggesting localized, mutation-specific perturbations within the G-domain. In contrast, G13D maintained near-wild-type dynamics, indicating that functional impairment may not require global structural destabilization. Consistent ensemble analyses (PCA, DCCM, FEL) revealed a dominant single-mode rearrangement for G60S (PC1 ≈ 50.8%), a heterogeneous multi-state landscape for G60V, and a restricted low-energy ensemble for G13D, supporting distinct, position-specific mechanisms. Gene-gene and protein-protein interaction analyses linked KRAS to multiple signaling partners, GO and KEGG enrichment analyses associated KRAS with GTPase activity, Rac protein signaling, and the RAS-MAPK and PI3K-AKT pathways implicated in lung cancer progression. Overall, these findings identify G13C, G13D, G60S, and G60V as high-priority candidates for experimental validation and may inform future strategies for genomic risk assessment and mutation-specific therapeutic intervention in lung cancer.</p>","PeriodicalId":94249,"journal":{"name":"The protein journal","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149066981","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Quantum-Classical Analysis of Variability in Peptide Energies and Conformations.","authors":"Ayushi, Madhvi Shakya","doi":"10.1007/s10930-026-10352-6","DOIUrl":"https://doi.org/10.1007/s10930-026-10352-6","url":null,"abstract":"<p><p>The exploration of protein structure and folding mechanisms is a cornerstone of molecular biology, with significant implications for understanding biological functions and advancing drug discovery. Traditional computational methods, such as AlphaFold, RoseTTAFold and I-TASSER, have been instrumental in this field. However, these methods are often constrained by extensive challenges in predicting structures for proteins lacking significant homologous sequences in the training data, impacting accuracy in novel or less well-characterized proteins. In contrast, quantum computing offers a promising avenue for overcoming these limitations through algorithms like the Variational Quantum Eigensolver (VQE). Quantum computing is in an early stage for optimization problems, particularly for protein folding. This study harnesses the power of VQE to explore the folding behavior and low-energy conformational landscape of 15 peptides, each comprising eight amino acids, selected from the RCSB Protein Data Bank (PDB). Classical 100-ns molecular dynamics (MD) simulations are used to characterize the structural stability and relative energetic trends of the corresponding atomistic conformations. These results are then contrasted with energies obtained from a lattice-based Hamiltonian optimization using the VQE algorithm, executed on the IBM quantum simulator. Although the absolute energy scales differ between MD and lattice Hamiltonian models, this comparison enables a correlation-based evaluation of how effectively VQE identifies physically plausible low-energy folded states. The results demonstrate that VQE consistently converges to competitive low-energy conformations. Further investigation was carried out by converting the PDB structure into a lattice model using the LatFit tool for visual comparison with the VQE-predicted structure. In addition, quantitative evaluation was performed by comparing the VQE-predicted structure with the original PDB structure using Cα Root Mean Square Deviation (RMSD) and Template Modeling (TM) score. Furthermore, the TM-scores obtained from the VQE predictions were compared with those obtained from the AlphaFold predictions. This work highlights the benefits of utilizing quantum algorithms to examine protein folding and also paves the road for future quantum biology studies.</p>","PeriodicalId":94249,"journal":{"name":"The protein journal","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149032358","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"NRB-FusePS: A Protein Language Model-Based Multimodal Framework for Predicting Ligand-Binding Residues in Nuclear Receptors.","authors":"TianXing Qiao, YongE Feng","doi":"10.1007/s10930-026-10354-4","DOIUrl":"https://doi.org/10.1007/s10930-026-10354-4","url":null,"abstract":"<p><p>Nuclear receptors (NRs) are ligand-dependent transcriptional regulators whose ligand-binding residues are crucial for receptor activation and drug-target analysis. In this study, we developed NRB-FusePS, a protein language model-based multimodal framework for predicting BioLiP-defined ligand-binding residues in NRs. Nuclear receptor ligand-binding information was collected from the BioLiP database, and residue-level samples were generated using 21-residue sliding windows. A protein-level independent test strategy was used to reduce information leakage caused by overlapping windows from the same protein. Sequence composition, amino acid pair, structural, evolutionary, and ESM2 protein language model descriptors were extracted and evaluated using conventional machine learning models. Based on these results, a three-branch deep learning model called NRB-Fuse was constructed to integrate sequence statistical descriptors, structural descriptors, and ESM2 embeddings, and protein-level probability smoothing was introduced to obtain NRB-FusePS. On the independent test set, NRB-FusePS-Full achieved an area under the receiver operating characteristic curve (AUC) of 0.814, a Matthews correlation coefficient (MCC) of 0.406, and a sensitivity of 0.964. Shapley additive explanations (SHAP) analysis showed that ESM2 was the main information source, and the structural case of O00482 supported the predictive utility. The NRBsite web server provides three prediction modes for different levels of input information and is freely available at http://47.95.115.29:8501/ .</p>","PeriodicalId":94249,"journal":{"name":"The protein journal","volume":" ","pages":""},"PeriodicalIF":1.4,"publicationDate":"2026-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149032439","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}