Computational and Machine Learning Approaches for Optimizing Anti-CD3ε Nanobody: Humanization and Characterization for Enhanced Therapeutic Efficacy

Computational and Machine Learning Approaches for Optimizing Anti-CD3ε Nanobody: Humanization and Characterization for Enhanced Therapeutic Efficacy | 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 Computational and Machine Learning Approaches for Optimizing Anti-CD3ε Nanobody: Humanization and Characterization for Enhanced Therapeutic Efficacy Ali Rahmati Bonab, Hannaneh Jalilzadeh Ghahi, Mahmoud Hassani, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5769566/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 This study presents a comprehensive machine-learning-driven approach for the in silico humanization and characterization of anti-CD3ε nanobodies. Nanobodies, single-domain antibodies derived from camelids, hold immense therapeutic potential due to their small size, high solubility, and exceptional stability. However, their camelid origin necessitates humanization to minimize immunogenicity in therapeutic applications. Using state-of-the-art computational tools such as NanoNet, RoseTTAFold, and PyDock, we modeled and analyzed both wild type and humanized anti-CD3ε nanobody variants. Key metrics, including structural stability, binding efficiency, thermal stability, and aggregation propensity, were evaluated. Humanization achieved enhanced humanness scores, increased thermal stability, and retained strong binding interactions with CD3ε while preserving the nanobody’s structural integrity. Molecular dynamics simulations confirmed minimal deviations in structural flexibility and binding-site compatibility post-humanization. These findings support the efficacy of computational methods in optimizing nanobody therapeutics for clinical applications, paving the way for advanced immunotherapy strategies targeting immune-related disorders. The results demonstrate that the humanized anti-CD3ε nanobody exhibits enhanced thermal stability, reduced aggregation propensity, improved humanness scores, and comparable binding efficiency to the wild type nanobody, making it a promising therapeutic candidate. Biological sciences/Biotechnology Biological sciences/Computational biology and bioinformatics Biological sciences/Drug discovery Biological sciences/Immunology Biological sciences/Molecular biology Biological sciences/Structural biology Anti-CD3ε nanobody Humanization immunogenicity Structural modeling nanobody engineering Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 Figure 16 Figure 17 Figure 18 Figure 19 Figure 20 Figure 21 Introduction Nanobodies, derived from camelid heavy-chain antibodies, are small (~ 15 kDa), stable, and soluble, with demonstrated potential in treating cancer, infectious diseases, neurodegenerative disorders, and autoimmune conditions. Their small size enhances tissue penetration, while production in microbial systems is cost-effective [ 1 , 2 ]. Examples include Caplacizumab, an FDA-approved nanobody for aTTP [ 3 ]. Nanobodies surpass scFvs in structural stability, solubility, and ease microbial production, offering superior tissue penetration [ 4 ]. Nanobody applications extend to neutralizing agents [ 5 ], receptor blockers[ 6 ], and drug-delivery vehicles [ 7 ], One notable application involves anti-CD3 nanobodies, which specifically bind the CD3ε chain to activate T cells. The CD3ε chain plays a critical role in TCR signaling, traditionally activated through TCR-MHC interactions. However, recent studies have revealed MHC-independent activation mechanisms involving structural rearrangements or direct molecular interactions, enabling immune responses without antigen presentation [ 8 , 9 ]. This is particularly advantageous in environments with MHC downregulation, enhancing immune surveillance and offering promising potential for immunotherapy [ 10 ].Activation through anti-CD3ε nanobodies promotes T-cell proliferation, cytokine production (e.g., IL-2 and IFNγ), and a Th1-dominant immune response while simultaneously suppressing tolerogenic cytokines [ 11 ]. Preclinical models demonstrate their efficacy in tumor suppression and enhanced immune surveillance, with clear advantages in stability, size, and immune activation compared to monoclonal antibodies [ 12 ]. Despite their therapeutic versatility, anti-CD3 nanobodies require humanization to address their lower intrinsic humanness scores, thereby minimizing immunogenicity in clinical applications [ 13 ]. To address this challenge, computational tools such as machine learning and deep learning have emerged as transformative approaches. These methods have demonstrated their ability to efficiently humanize nanobodies, ensuring compatibility with the human immune system while preserving their functional properties [ 14 , 15 ]. By leveraging these advanced tools, researchers can overcome immunogenicity barriers, enhancing the therapeutic potential of nanobody-based treatments In this study, we leverage a suite of advanced computational methods to enhance the modeling, humanization, and characterization of nanobody structures. Deep learning (DL) and machine learning (ML) approaches, including AGGRESCAN3D [ 16 ] and DeepSTABp [ 17 ], are utilized to evaluate aggregation propensity [ 18 ] and stability of nanobody structures. For structure prediction specific to nanobodies, NanoNet [ 19 ] is employed, demonstrating exceptional accuracy and outperforming traditional tools such as AlphaFold [ 20 ], ESMFold [ 21 ], and Yang-Server [ 22 ]. These methods enable precise identification of structural features, such as complementarity-determining regions (CDRs), framework regions (FWRs) and aggregation hotspots, which are critical for designing stable and functional nanobody therapeutics[ 23 ]. To address the immunogenicity associated with nanobodies derived from camelid antibodies, we incorporate ML-based humanization tools, such as Llamanade [ 24 , 25 ] and BioPhi [ 26 ]. These tools allowed us to optimize sequence similarity to the human antibody repertoire while maintaining the binding affinity and structural stability of the nanobody. By combining these state-of-the-art methods, we conducted a comprehensive comparison between humanized and wild type nanobodies. This analysis includes evaluations of thermal stability, aggregation propensity, and binding efficiency, alongside other critical metrics that influence their therapeutic potential. The integration of DL and ML-driven tools throughout this process not only ensures enhanced accuracy and efficiency but also underscores the transformative potential of computational approaches in nanobody engineering and optimization. Methods Sequence Acquisition for Anti-CD3ε Nanobody and CD3ε The sequences for the anti-CD3ε nanobody and the CD3ε were obtained from the patent WO 2010/037838 A2 [27] shown in Table 1. Nanobody Structural Prediction, Visualization, Validation, and Comparison The anti-CD3ε nanobodies (wild type and humanized) were modeled using NanoNet [28], a specialized tool for nanobody structural prediction. To analyze and visualize the resulting nanobody structures, PyMOL[29] and BioLuminate [30] were employed for detailed 3D modeling. The structural integrity of the humanized and wild type nanobody models was validated using Ramachandran plots, generated with the RamPlot [31], to assess residue conformational quality [32]. Comparative analysis of the nanobody structures was conducted using Template Modeling-align (TM-align) from the Zhang Lab [33], which evaluates structural similarity through metrics such as Template Modeling score (TM-score), Root Mean Square Deviation (RMSD), and the percentage of aligned residues [34]. These metrics provide a quantitative basis for assessing the impact of humanization on the structural integrity and alignment of the nanobody framework, ensuring a thorough evaluation of the modeled structures. The structural comparison was visualized using the RCSB PDB 3D viewer [35] which effectively illustrated the minimal differences between the two nanobody variants. Structural Prediction and Energy Minimization of CD3ε The CD3ε protein structure was predicted using the RoseTTAFold All-Atom model [36], a machine learning tool for high-accuracy structural predictions. The amino acid sequence of CD3ε was used as input, and the resulting model was evaluated for confidence and alignment quality using predicted Local Distance Difference Test (pLDDT) and Predicted Aligned Error (PAE) metrics [37]. The structure was refined using molecular dynamics simulations via the AMBER Relaxation protocol [38] , with energy values recorded to assess stability improvements. Finally, the accuracy of the refined model was evaluated by comparing it to the experimental 1XIW structure [39], focusing on framework regions and flexible loops for alignment. CDR and FWR Identification The anti-CD3ε nanobody sequence was analyzed using ANARCI [40], employing the North/Aho numbering scheme for accurate delineation of CDRs and FWRs [41, 42]. Humanization via CDR Grafting CDR grafting has been employed to humanize nanobodies by transferring CDRs onto human antibody frameworks [43] while preserving critical Vernier zone residues to ensure structural integrity and binding affinity [44]. The process was optimized using Llamanade and evaluated with BioPhi to measure sequence similarity to human antibody repertoires, ensuring minimal disruption of CDR conformation. Additionally, the T20 Score Scale has been used to quantify the degree of humanization and predict immunogenicity risks [45], thereby enhancing therapeutic compatibility. Docking and Binding Analysis Docking studies evaluated interactions between the CD3ε antigen and two nanobody variants: humanized and wild type. A two-step docking approach was utilized [46], beginning with global docking using BioLuminate, which sampled 70,000 rotational poses to ensure comprehensive coverage of binding conformations, with parameters applied for energy minimization and analyzing interactions within an 8 Å radius, identifying significant binding contributions from the CDRs, mutation for humanization candidate residues, and Vernier zone residues in the nanobody FWR. This was followed by Site-Specific docking using PyDock [47, 48], specifically targeting amino acids 1–27 of CD3ε, informed by patent data indicating the nanobody binds this segment. The resulting docking conformations were analyzed and visualized using PDBsum [49] to evaluate the structural features of the docked complexes. Surface Conservation Analysis The ConSurf web server [50] was utilized to assess surface residue conservation by generating an evolutionary conservation profile of the nanobody sequence. This analysis involved aligning the nanobody sequence against homologous sequences retrieved from the UniProt database to construct a multiple sequence alignment (MSA). Evolutionary conservation scores were subsequently calculated using the Bayesian inference method [51], as described in the ConSurf methodology. Residues with higher conservation scores were identified as critical for structural and functional integrity [52]. Nanobody Thermostability and Physicochemical Property Analysis The thermal stability of wild type and humanized nanobody variants was analyzed using DeepSTABp, which predicts ΔTm values based on amino acid sequences under lysate-based and cellular environments at 36.9°C [17]. This machine-learning tool evaluates thermal stability using sequence-derived features without requiring structural data. The pKa values of ionizable residues were predicted with SystemBuilder [53], highlighting shifts in ionization behavior post-humanization[54]. Protein-Sol [55] and NetSurfP [56, 57] were employed to analyze solubility, charge, and structural propensities. Hydrophobicity was assessed using two tools: SODOPE [58, 59], which calculates the GRAVY index, and Protein-Sol, which applies the Kyte-Doolittle scale.[60] Ion mobility collision cross-section (CCS) prediction CCS, a measure of an ion's effective surface area, was calculated using Rosetta integrated with the PARCS algorithm for accurate and rapid estimations directly from protein structures [61, 62]. CCS analysis of wild type and humanized nanobody structures highlighted differences in surface area, folding dynamics, and molecular compactness introduced by humanization, providing valuable insights into their stability and functionality [63, 64]. Aggregation Propensity Prediction AGGRESCAN3D was used to predict aggregation hotspots in nanobodies. This machine-learning-based tool evaluates the aggregation propensity of 3D protein models, identifying destabilizing regions that could lead to aggregation. Toxicity and Allergenicity Prediction Nanobody toxicity was evaluated using ToxinPred [65], a computational tool that employs metrics such as the Motif-Extraction-Reliable Classification Indices (MERCI) Score to detect toxicity-prone sequence motifs and Hybrid Scores, which combine predictive parameters for a comprehensive toxicity assessment [66]. Allergenicity was assessed using AlgPred, which integrates allergenic motif databases, sequence features, and Support Vector Machine (SVM) based predictions [67] to generate a robust allergenicity profile [68]. Molecular Dynamics Simulations The system was solvated using the TIP3P water model in a cubic box with a 1.0 nm solute-box edge distance and neutralized with Na⁺ and Cl⁻ ions. Steepest descent minimization optimized ion placement and energy, with long-range electrostatics and Van der Waals interactions handled via PME and a 1.0 nm cut-off. Production Molecular Dynamics (MD) was conducted for 100 ps (50,000 steps) with a 2 fs time step, using the leap-frog integrator, V-rescale thermostat (300 K), and Parrinello-Rahman barostat (1.0 bar) [69]. Data was recorded at set intervals with automatic checkpointing. Analysis focused on C-alpha atoms to study the structural dynamics and stability of wild type and humanized nanobodies. Results Structural Prediction, Humanization, and Docking Analysis of the Wild Type Nanobody The 3D structure of the wild type nanobody was predicted using NanoNet [28] and validated through a Ramachandran plot (Figures 1 and 2), which revealed a PB scale score of 0.927, indicating that most residues were in favored regions and confirming the structure's quality and nanobody sequence was analyzed for North/Aho numbering and accurate delineation of CDRs and FWRs. The results are summarized in Table 2: Upon applying the Llamanade humanization process, six mutations were identified as necessary for humanization using the CDR grafting method (Table 3) and BioPhi suggested seven mutations, including an additional mutation at position 75 (A → S) (Table 4). Despite BioPhi's recommendation, residue frequency analysis indicated that the frequency of position 75 increased only modestly from 38% to 59% (a 21% rise), which is a minor change compared to other positions. Furthermore, conservation analysis (Figure 3) classified position 75 as highly conserved, with a conservation scale score of 8 (Figure 4), suggesting potential stability and binding alterations if mutated. Given its high conservation and negligible impact on humanness, the mutation at position 75 was not implemented, leaving this residue unchanged. The CD3ε protein structure was predicted using the RoseTTAFold All-Atom model achieving a mean pLDDT score of 73.674, indicative of moderate to high confidence, with higher precision observed in conserved regions. The model demonstrated a mean PAE of 8.307, indicating a good residue-residue alignment. Following this, energy minimization, this process improved the model's stability and accuracy, with the computed energy values decreasing from 3696.127 kJ/mol (initial) to -1177.263 kJ/mol (final). In the final structure, the framework regions demonstrated strong alignment with the 1XIW template, while the flexible loops exhibited moderate improvements in both confidence and alignment. Following global docking of the wild type nanobody with CD3ε, all six defined residues were mutated to assess their impact on binding affinity. BioLuminate analysis showed negligible changes in binding affinity (Table 5). The strongest interactions and closest distances remained localized in the CDR regions, where the nanobody binds the antigen. Following site-specific docking analysis using PyDock, the first model was selected for further investigation due to its lowest energy value (ΔG = -100.33 kcal/mol), indicative of strong binding affinity (Table 6). In this model, a key interaction was observed between tryptophan (W) at position 44 within the nanobody’s FWR2 region and specific amino acid residues of CD3ε (Figure 5). Tryptophan 44 exhibited substantial interactions, emphasizing its critical role in stabilizing the antigen-binding interface. Notably, in nine out of the top ten ranked docking poses, the interaction between W44 and CD3ε was consistently observed. However, detailed descriptions of these additional models were omitted for brevity. Given these findings, we opted to retain W44 in its native form while mutating the remaining five residues to achieve humanization, without compromising the structural integrity or functional performance of the nanobody. As shown in Figure 3, the conservation analysis of residues at positions 88 and 105 revealed moderate to low conservation scores. (Figure 4), and mutations at these positions significantly increased the human residue frequency, enhancing the nanobody’s "humanness" and reducing immunogenicity. In addition, residues at positions 76, 87, and 89 displayed above-average conservation scores, and mutations at these positions also significantly increased the human residue frequency, like positions 88 and 105 enhancing the nanobody’s humanness level and reducing immunogenicity. Further analysis across various aspects in the following sections of this article demonstrated that all five mutations (at positions 76, 87, 88, 89, and 105) had minimal impact on the nanobody's affinity, stability, or other functional properties, despite the conservation scores from Consurf, which may seem concerning. We predicted the 3D structure of this nanobody sequence using NanoNet, modeled similarly to the wild type structure. Validation was performed using a Ramachandran plot, which revealed that the humanized nanobody exhibited a higher proportion of residues in favored regions (96.3%) compared to the wild type (92.7%) (Figures 6 and 7). This difference (ΔPb = 3.6%) suggests that the humanization process enhanced structural regularity and optimization. These findings indicate that the humanized nanobody is slightly more conformationally favorable than the wild type counterpart (Figure 8). The structural comparison between the humanized and wild type nanobody revealed an RMSD of 0.25 Å, indicating minimal deviation, with 95.8% of residues aligning between the two structures. Furthermore, a TM score of 0.996 confirmed near-complete structural equivalence, highlighting that the humanized nanobody effectively retains the 3D conformation of its wild type counterpart. These results emphasize the successful preservation of the wild type nanobody’s structural integrity in its humanized version (Figure 9). All subsequent nanobody characterizations (wild type and humanized) discussed in the following sections are based on NanoNet-generated 3D structural PDB files. We conducted docking analysis again with PyDock, this time on the humanized nanobody, to evaluate its interaction with the target epitope of CD3ε (1-27 aa). Among the top 10 models with the lowest binding energies, the model with the lowest energy (-96.64 kJ/mol) was selected for further analysis (Table 7). The docking analysis revealed that tryptophan at position 44 (W44) in the FWR of the humanized nanobody maintains its interaction with residues of the CD3ε antigen, similar to the wild type nanobody(Figure 10). This conserved interaction underscores the structural and functional integrity of the humanized nanobody. The comparison of interface statistics between the wild type and humanized nanobody complexes with CD3ε highlights only slight differences, demonstrating that the humanization process successfully retained much of the nanobody's binding characteristics. In the wild type complex, the CD3ε chain (Chain A) has 18 interface residues with an interface area of 947 Ų, while the nanobody (Chain B) has 13 residues and an interface area of 791 Ų. In the humanized nanobody complex, the CD3ε interface residues decrease slightly to 14, and the interface area reduces modestly to 728 Ų, while the nanobody's interface area remains similar at 725 Ų. These small variations suggest that the humanized nanobody maintains strong structural compatibility with CD3ε. In terms of interactions, the wild type nanobody forms 5 hydrogen bonds and 200 non-bonded contacts with CD3ε, while the humanized nanobody has 1 hydrogen bond and 165 non-bonded contacts. While there is a minor reduction in polar and non-bonded interactions, the overall difference is not significant, indicating that the humanized nanobody still exhibits a robust binding potential. Additionally, we observed that there were no disulfide bonds formed either within the nanobody structures (intra-molecular) or between the nanobody and CD3ε (inter-molecular) in both the wild type and humanized models. We conducted a comparative analysis of docking models generated using PyDock for both wild type and humanized nanobodies. As shown in Figure11, the docking scores for the two nanobody types against CD3ε exhibit a similar overall distribution, indicating that the humanization process maintained the nanobody's binding characteristics. Both models show a wide energy range (-100 to +40), with the wild type nanobody displaying a slightly higher frequency in the most favorable scoring range. The humanized nanobody closely follows this pattern, with only minor differences in peak frequencies and scoring spread. These findings suggest negligible differences in the binding. Humanness Score Analysis The wild type nanobody achieved a T20 FWR score of 85.74 and an overall T20 score of 82.83. After humanization, the nanobody demonstrated improved metrics, with a T20 FWR score of 91.03 and an overall T20 score of 87.04. Analysis of Nanobody Thermostability and Physicochemical Properties The humanized nanobody exhibited superior thermal stability compared to the wild type, with higher melting temperatures (Tm) in both lysate (54.56°C vs. 51.09°C) and cell-based environments (53.19°C vs. 51.72°C), indicating enhanced robustness (Figure 12). Post-humanization, pKa values for ionizable residues showed minimal changes, preserving key chemical properties (Figure 13). Humanization also reduced the absolute charge, slightly increased hydrophobicity, and maintained stable beta-strand propensity (Figure 14). Solubility metrics, including Relative Solvent Accessibility (RSA) and Accessible Surface Area (ASA), revealed negligible differences (Cohen's d < 0.2), and both nanobodies were moderately hydrophilic, as reflected by Grand Average of Hydropathicity (GRAVY) indices of -0.40 for wild type and -0.32 for humanized nanobody (Figure 15). Ion mobility collision cross-section (CCS) prediction and Aggregation Score Analysis The humanized nanobody exhibited a CCS value of 1226.86 Ų, slightly smaller than the wild type nanobody’s CCS value of 1234.85 Ų. Aggregation Propensity Prediction As Figure 16 shows the humanized nanobody exhibited a slightly less negative mean aggregation score (-0.6447) compared to the wild type (-0.7001). Toxicity and Allergenicity Prediction Both the wild type and humanized nanobodies are predicted to be non-toxic, with no sequences classified as "Toxin." The humanized nanobody shows a slightly higher ML Score (0.21 vs. 0.175), while other toxicity-related scores, such as MERCI and Hybrid scores, remain identical between the two versions. Regarding allergenicity, the wild type nanobody has a score of -0.14507017, which is above the threshold of -0.4, classifying it as an allergen. In contrast, the humanized nanobody achieves a score of -0.52723881, below the threshold, and is therefore classified as a non-allergen. Molecular Dynamics Simulations The combined analyses of radius of gyration (Figure 17), RMSD (Figures 18, and 19), Root Mean Square Fluctuation (RMSF, Figure 20), and Neq (Figure 21) provide a comprehensive understanding of the structural dynamics of the wild type and humanized nanobody. RMSD analysis shows similar global structural deviations, stabilizing around ~0.06–0.08 nm, indicating preserved global stability after humanization. RMSF values remain low in the framework regions (~0.1–0.3 nm) for both structures, confirming the conserved scaffold's structural integrity. Region-specific differences were observed in the CDRs: in CDR1 (residues 23–35), RMSF increased (~0.25–0.4 nm to ~0.3–0.45 nm), as did Neq (~1.0 to ~1.1), reflecting enhanced flexibility. CDR2 (residues 50–59) showed minimal changes in RMSF (~0.2–0.35 nm) and Neq (~1.05 to ~1.1), indicating structural conservation. In contrast, CDR3 (residues 97–109) exhibited reduced RMSF (~0.35–0.6 nm to ~0.3–0.5 nm) and Neq (~1.3 to ~1.1), suggesting loop stabilization. The Rg analysis further supports these findings, with the humanized nanobody showing a mean Rg of 1.366 nm and the wild type 1.361 nm, reflecting comparable stability. Axis-specific Rg values suggest that the humanized nanobody exhibits more balanced stability along the X and Z axes (1.163 nm and 1.210 nm, respectively), compared to the wild type's anisotropy, more compact along the Y-axis (0.900 nm). These localized adjustments in flexibility and stabilization, particularly in the CDRs, suggest that humanization preserved global structural integrity while introducing changes that could influence antigen-binding dynamics. Data Visualization GraphPad Prism version 9.0 was utilized for generating comparative line graphs and statistical plots, while Python enabled the inclusion of advanced customizations such as shaded regions, annotations, and multi-layered data visualizations. Discussion The findings of this study highlight the transformative role of advanced computational methods in optimizing nanobody therapeutics. By comparing the wild type and humanized anti-CD3ε nanobodies, we demonstrate the advantages of a data-driven approach to enhance their structural, functional, and therapeutic properties. Using NanoNet, the structural modeling confirmed that the humanized nanobody retained the overall 3D conformation of the wild type, with an RMSD of 0.25 Å and a TM-score of 0.996. These metrics validate the structural preservation achieved through humanization. Furthermore, the humanized nanobody exhibited improved conformational variability, which is essential for maintaining dynamic adaptability and functionality in biological environments. Enhanced thermal stability was observed, with melting temperatures (Tm) increasing by 3.47°C in lysate and 1.47°C in cell-based environments. These results underscore the success of the humanization process in bolstering structural robustness while preserving flexibility. The functional importance of framework residues, particularly those in FWR2, has been well-documented [ 70 ]. Alterations in these residues can significantly influence the conformation of the complementarity-determining regions (CDRs), thereby affecting the nanobody's overall affinity and stability [ 71 – 73 ]. Previous studies have also demonstrated that framework region residues can affect CDR conformation and, consequently, the binding affinity of antibodies [ 74 ]. These insights underscore the necessity of accounting for FWR residues, particularly FWR2, during the humanization process to preserve nanobody functionality and stability. In this article, it is demonstrated that tryptophan at position 44 in the framework region (FWR) has a significant interaction with its antigen. This finding underscores the critical role of amino acids in the FWR, as they influence intermolecular interactions, stability, and the affinity of nanobodies. Consequently, during the mutation of residues in the FWR for humanization, this factor must be carefully taken into account to preserve nanobody functionality and stability. Docking analyses revealed comparable binding affinities for both variants, with the humanized nanobody retaining critical binding interactions with CD3ε. However, this reduction was negligible and did not significantly impact overall antigen-binding performance. Importantly, the improved conformational variability of the humanized nanobody likely offsets this minimal decrease, supporting its functional reliability in therapeutic contexts. The comparison of global and site-specific docking underscores the significance of site-specific docking methods in nanobody humanization. In global docking performed using BioLuminate software, no interaction was observed within 8 Å of W44. This outcome likely stems from the nature of global docking, where the nanobody may interact with other epitopes on the antigen that are not of therapeutic interest but, site-specific docking with PyDock revealed strong interactions between W44, located in the FWR2 region of the nanobody, and the CD3ε desired epitope. PyDock targeted the 1–27 amino acid region of CD3ε, the therapeutically relevant epitope critical for this nanobody's activity as outlined in patent WO 2010/037838 A2 [ 27 ]. AGGRESCAN3D analysis confirmed reduced aggregation propensity for the humanized nanobody, reflecting its enhanced stability and suitability for clinical use. Solubility metrics remained consistent between the two variants, indicating that humanization preserved key physicochemical properties. CCS analysis showed a slight reduction in molecular size for the humanized nanobody (1226.86 Ų vs. 1234.85 Ų for the wil type), suggesting improved molecular compactness and potential benefits for pharmacokinetics. The humanized nanobody exhibited a significant reduction in allergenicity score, transitioning from allergenic (wild type) to non-allergenic (humanized). Both nanobodies were classified as non-toxic, confirming their suitability for therapeutic applications. Molecular dynamics simulations further revealed that the humanized nanobody demonstrated improved conformational variability. While CDR1 exhibited increased flexibility, CDR3 showed stabilization, contributing to a balance between adaptability and structural integrity. The humanization process not only preserved the structural and functional integrity of the nanobody but also significantly enhanced its humanness, as evidenced by the T20 score results. The humanized nanobody achieved an overall T20 score of 87.04 compared to 82.83 for the wild type nanobody, with notable improvements in the framework region (91.03 vs. 85.74). These higher scores underscore the effectiveness of the humanization strategy in aligning the nanobody sequence closer to the human antibody repertoire. The combination of DL and ML tools such as BioPhi, NanoNet, LlamaNADE and AGGRESCAN3D enabled precise humanization and optimization of the nanobody. Improved humanness scores, reduced aggregation, enhanced conformational variability, and preserved antigen-binding capabilities establish the humanized nanobody as a superior therapeutic candidate. Conclusion This study demonstrates the potential of integrating advanced computational tools to design optimized nanobody therapeutics. The humanized anti-CD3ε nanobody exhibited enhanced thermal stability, reduced aggregation propensity, and improved conformational variability, making it well-suited for clinical applications. Despite a negligible decrease in CDR3 binding affinity, the nanobody's overall performance, including preserved binding efficiency and reduced immunogenicity, underscores the robustness of the humanization process. These findings validate the role of DL and ML techniques in advancing nanobody engineering, and provide a strong foundation for the development of next-generation biologics targeting immune-related and other diseases. This study highlights how computational strategies can bridge the gap between molecular design and therapeutic application. Despite the significant insights gained, several limitations must be acknowledged. The findings are primarily based on computational modeling, with no in vitro or in vivo validation, limiting their direct clinical applicability. Additionally, while aggregation scores were reduced, the long-term stability and solubility under varying physiological conditions remain unexplored. Building on these findings, future research should focus on conducting in vitro and in vivo experiments to validate the stability, binding efficiency, and therapeutic efficacy of the humanized nanobody. Furthermore, investigations into thermal stability, aggregation propensity, and solubility under diverse environmental and physiological conditions, such as varying pH or ionic strength, will provide a clearer understanding of its clinical potential. Developing pharmacokinetic and immunogenicity studies in animal models will further assess its compatibility and performance in therapeutic settings. Declarations Competing interests The authors declare that they have no known competing financial interests or personal Author Contribution Ali Rahmati Bonab and Hannaneh Jalilzadeh Ghahi contributed equally to this work. Both were involved in writing, review, and editing of the manuscript, original draft preparation, visualization, and validation of results. Mahmoud Hasani, Vahid Jajarmi, and Javad Ranjbari provided critical feedback, supervised the study, and contributed to data analysis and interpretation. All authors participated in conceptualization, approved the final manuscript, and agreed to be accountable for the work. Acknowledgement We acknowledge the use of OpenAI's ChatGPT for assistance in drafting and refining sections of this manuscript. ChatGPT was used to generate suggestions for text improvement and content clarity. Data Availability The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. References Vincke, C., et al., General strategy to humanize a camelid single-domain antibody and identification of a universal humanized nanobody scaffold. J Biol Chem, 2009. 284 (5): p. 3273-3284. Muyldermans, S., Nanobodies: natural single-domain antibodies. Annual review of biochemistry, 2013. 82 : p. 775-797. Scully, M., et al., Caplacizumab Treatment for Acquired Thrombotic Thrombocytopenic Purpura. N Engl J Med, 2019. 380 (4): p. 335-346. Panikar, S.S., et al., Nanobodies as efficient drug-carriers: Progress and trends in chemotherapy. Journal of Controlled Release, 2021. 334 : p. 389-412. Yang, J., et al., Development of a bispecific nanobody conjugate broadly neutralizes diverse SARS-CoV-2 variants and structural basis for its broad neutralization. PLOS Pathogens, 2023. 19 (11): p. e1011804. Omidfar, K., et al., Efficient growth inhibition of EGFR over-expressing tumor cells by an anti-EGFR nanobody. Mol Biol Rep, 2013. 40 (12): p. 6737-45. Abdolvahab, M.H., et al., Targeted drug delivery using nanobodies to deliver effective molecules to breast cancer cells: the most attractive application of nanobodies. Cancer Cell Int, 2024. 24 (1): p. 67. Rangarajan, S., et al., Peptide-MHC (pMHC) binding to a human antiviral T cell receptor induces long-range allosteric communication between pMHC- and CD3-binding sites. J Biol Chem, 2018. 293 (41): p. 15991-16005. He, Y., et al., Peptide-MHC Binding Reveals Conserved Allosteric Sites in MHC Class I- and Class II-Restricted T Cell Receptors (TCRs). J Mol Biol, 2020. 432 (24): p. 166697. Alba, J. and M. D'Abramo, The Full Model of the pMHC-TCR-CD3 Complex: A Structural and Dynamical Characterization of Bound and Unbound States. Cells, 2022. 11 (4). Moradi-Kalbolandi, S., et al., Evaluation the potential of recombinant anti-CD3 nanobody on immunomodulatory function. Mol Immunol, 2020. 118 : p. 174-181. Khatibi, A.S., et al., In vivo tumor-suppressing and anti-angiogenic activities of a recombinant anti-CD3ε nanobody in breast cancer mice model. Immunotherapy, 2019. 11 (18): p. 1555-1567. Gordon, G.L., et al., Prospects for the computational humanization of antibodies and nanobodies. Front Immunol, 2024. 15 : p. 1399438. Chinery, L., J.R. Jeliazkov, and C.M. Deane, Humatch - fast, gene-specific joint humanisation of antibody heavy and light chains. MAbs, 2024. 16 (1): p. 2434121. Choi, Y., et al., Computationally driven antibody engineering enables simultaneous humanization and thermostabilization. Protein Eng Des Sel, 2016. 29 (10): p. 419-426. Zambrano, R., et al., AGGRESCAN3D (A3D): server for prediction of aggregation properties of protein structures. Nucleic Acids Res, 2015. 43 (W1): p. W306-13. Jung, F., et al., DeepSTABp: A Deep Learning Approach for the Prediction of Thermal Protein Stability. Int J Mol Sci, 2023. 24 (8). Kuriata, A., et al., Aggrescan3D (A3D) 2.0: prediction and engineering of protein solubility. Nucleic Acids Research, 2019. 47 (W1): p. W300-W307. Cohen, T., M. Halfon, and D. Schneidman-Duhovny, NanoNet: Rapid and accurate end-to-end nanobody modeling by deep learning. Front Immunol, 2022. 13 : p. 958584. Jumper, J., et al., Highly accurate protein structure prediction with AlphaFold. Nature, 2021. 596 (7873): p. 583-589. Lin, Z., et al., Evolutionary-scale prediction of atomic-level protein structure with a language model. Science, 2023. 379 (6637): p. 1123-1130. Du, Z., et al., The trRosetta server for fast and accurate protein structure prediction. Nature Protocols, 2021. 16 (12): p. 5634-5651. Valdés-Tresanco, M.S., et al., Structural Modeling of Nanobodies: A Benchmark of State-of-the-Art Artificial Intelligence Programs. Molecules, 2023. 28 (10). Sang, Z., et al., Llamanade: An open-source computational pipeline for robust nanobody humanization. Structure, 2022. 30 (3): p. 418-429.e3. Sangzhe, Llamanade: A nanobody humanization tool . 2025. Prihoda, D., et al., BioPhi: A platform for antibody design, humanization, and humanness evaluation based on natural antibody repertoires and deep learning. MAbs, 2022. 14 (1): p. 2020203. Kufer, P. and T. Raum, Cross-species-specific single domain bispecific single chain antibody . 2010, World Intellectual Property Organization (WIPO). Dina, L., NanoNet: A computational tool for nanobody prediction and analysis . 2024, GitHub repository. Schrödinger, L.L.C., PyMOL Molecular Graphics System . 2020, Schrödinger: New York. Schrödinger, L.L.C., BioLuminate . 2021, Schrödinger: New York. RamPlot: A Webserver for Ramachandran Plot Analysis . Available from: https://ramplot.in/. Sobolev, O.V., et al., A Global Ramachandran Score Identifies Protein Structures with Unlikely Stereochemistry. Structure, 2020. 28 (11): p. 1249-1258.e2. Zhang, L. TM-align: A Protein Structure Alignment Tool . Available from: https://zhanggroup.org/TM-align/. Zhang, Y. and J. Skolnick, TM-align: a protein structure alignment algorithm based on the TM-score. Nucleic Acids Res, 2005. 33 (7): p. 2302-9. viewer, R.P., RCSB Pdb Viewer . 2024. Baker, L., RoseTTAFold-All-Atom: High-resolution protein structure modeling . 2024, GitHub repository. Anbo, H., et al., How AlphaFold2 Predicts Conditionally Folding Regions Annotated in an Intrinsically Disordered Protein Database, IDEAL. Biology (Basel), 2023. 12 (2). NeuroSnap, AMBER Relaxation Service . 2024, NeuroSnap AI. Rcsb, P.D.B., Structure 1XIW: Crystal structure of a protein . RCSB Protein Data Bank. Oxford Protein Informatics, G., Antibody Sequence Analysis using ANARCI . 2024. Dunbar, J. and C.M. Deane, ANARCI: antigen receptor numbering and receptor classification. Bioinformatics, 2016. 32 (2): p. 298-300. Dondelinger, M., et al., Understanding the Significance and Implications of Antibody Numbering and Antigen-Binding Surface/Residue Definition. Frontiers in Immunology, 2018. 9 . Williams, D.G., D.J. Matthews, and T. Jones, Humanising Antibodies by CDR Grafting , in Antibody Engineering , R. Kontermann and S. Dübel, Editors. 2010, Springer Berlin Heidelberg: Berlin, Heidelberg. p. 319-339. Makabe, K., et al., Thermodynamic consequences of mutations in vernier zone residues of a humanized anti-human epidermal growth factor receptor murine antibody, 528. J Biol Chem, 2008. 283 (2): p. 1156-66. Almagro, J.C. and J. Fransson, Humanization of antibodies. Front Biosci, 2008. 13 : p. 1619-33. Kong, R., et al., CoDockPP: A Multistage Approach for Global and Site-Specific Protein–Protein Docking. Journal of Chemical Information and Modeling, 2019. 59 (8): p. 3556-3564. Jiménez-García, B., C. Pons, and J. Fernández-Recio, pyDockWEB: a web server for rigid-body protein-protein docking using electrostatics and desolvation scoring. Bioinformatics, 2013. 29 (13): p. 1698-9. Cheng, T.M., T.L. Blundell, and J. Fernandez-Recio, pyDock: electrostatics and desolvation for effective scoring of rigid-body protein-protein docking. Proteins, 2007. 68 (2): p. 503-15. Laskowski, R.A., PDBsum: summaries and analyses of PDB structures. Nucleic Acids Res, 2001. 29 (1): p. 221-2. The ConSurf Project, T. ConSurf: A Webserver for Identifying Functional Regions in Proteins by Mapping Evolutionary Conservation . Available from: https://consurf.tau.ac.il/consurf_index.php. Taboga, M., Bayesian inference Lectures on probability theory and mathematical statistics . 2021: Kindle Direct Publishing. Ashkenazy, H., et al., ConSurf 2016: an improved methodology to estimate and visualize evolutionary conservation in macromolecules. Nucleic Acids Res, 2016. 44 (W1): p. W344-50. PlayMolecule Project, T. PlayMolecule System Builder: A Webserver for Molecular System Preparation and pKa Prediction . Available from: https://open.playmolecule.org/tools/systembuilder. Burger, S.K. and P.W. Ayers, Empirical prediction of protein pKa values with residue mutation. J Comput Chem, 2011. 32 (10): p. 2140-8. Hebditch, M., et al., Protein-Sol: a web tool for predicting protein solubility from sequence. Bioinformatics, 2017. 33 (19): p. 3098-3100. Eryk, NetSurfP-3.0: Predicting the structural properties of proteins. Høie, M.H., et al., NetSurfP-3.0: accurate and fast prediction of protein structural features by protein language models and deep learning. Nucleic Acids Research, 2022. 50 (W1): p. W510-W515. Bhandari, B.K., C.S. Lim, and P.P. Gardner, TISIGNER.com: web services for improving recombinant protein production. Nucleic Acids Research, 2021. 49 (W1): p. W654-W661. Bhandari, B.K., P.P. Gardner, and C.S. Lim, Solubility-Weighted Index: fast and accurate prediction of protein solubility. Bioinformatics, 2020. 36 (18): p. 4691-4698. Kyte, J. and R.F. Doolittle, A simple method for displaying the hydropathic character of a protein. J Mol Biol, 1982. 157 (1): p. 105-32. Turzo, S., et al., Predicting ion mobility collision cross sections using projection approximation with ROSIE-PARCS webserver. Brief Bioinform, 2023. 24 (5). Turzo, S., et al., Protein shape sampled by ion mobility mass spectrometry consistently improves protein structure prediction. Nat Commun, 2022. 13 (1): p. 4377. Williams, J.P., et al., Use of ion mobility mass spectrometry and a collision cross-section algorithm to study an organometallic ruthenium anticancer complex and its adducts with a DNA oligonucleotide. Rapid Commun Mass Spectrom, 2009. 23 (22): p. 3563-9. Covey, T. and D.J. Douglas, Collision cross sections for protein ions. J Am Soc Mass Spectrom, 1993. 4 (8): p. 616-23. Neurosnap.ai, ToxinPred: Peptide Toxicity Prediction Service. Rathore, A.S., et al., ToxinPred 3.0: An improved method for predicting the toxicity of peptides. Computers in Biology and Medicine, 2024. 179 : p. 108926. Cervantes, J., et al., A comprehensive survey on support vector machine classification: Applications, challenges and trends. Neurocomputing, 2020. 408 : p. 189-215. Saha, S. and G.P. Raghava, AlgPred: prediction of allergenic proteins and mapping of IgE epitopes. Nucleic Acids Res, 2006. 34 (Web Server issue): p. W202-9. Parrinello, M. and A. Rahman, Polymorphic transitions in single crystals: A new molecular dynamics method. Journal of Applied Physics, 1981. 52 (12): p. 7182-7190. Kazemi-Lomedasht, F., et al., Design of a humanized anti vascular endothelial growth factor nanobody and evaluation of its in vitro function. Iran J Basic Med Sci, 2018. 21 (3): p. 260-266. Xiang, J., et al., Framework residues 71 and 93 of the chimeric B72.3 antibody are major determinants of the conformation of heavy-chain hypervariable loops. J Mol Biol, 1995. 253 (3): p. 385-90. Wu, H., et al., Humanization of a murine monoclonal antibody by simultaneous optimization of framework and CDR residues. J Mol Biol, 1999. 294 (1): p. 151-62. Fernández-Quintero, M.L., M.C. Heiss, and K.R. Liedl, Antibody humanization-the Influence of the antibody framework on the CDR-H3 loop ensemble in solution. Protein Eng Des Sel, 2019. 32 (9): p. 411-422. Foote, J. and G. Winter, Antibody framework residues affecting the conformation of the hypervariable loops. J Mol Biol, 1992. 224 (2): p. 487-99. Tables Table 1. Amino acid sequences of the anti-CD3ε nanobody and CD3ε extracellular region. The single-letter codes represent the respective protein components Protein Component Amino Acid Sequence (Single Letter Code) Anti-CD3ε Nanobody EVQLVESGGGLVQPGGSLRLSCAASGFTFDDYGMSWVRQAPGKWLEWVSDISWNGGSTYYADSVKGRFTISRDNAENTLYLQMNSLKPDDT AVYYCAKMGEGGWGANDYWGQGTQVTVSS CD3ε (Extracellular) QDGNEEMGGITQTPYKVSISGTTVILTCPQYPGSEILWQHNDKNIGGDEDDKNIGSDEDHLSLKEFSELEQSGYYVCYPRGSKPEDANFYLYLRA RVCENCMEMD Table 2. detailing the amino acid sequence regions of the wild type nanobody structure. The framework regions (FWRs) and complementarity-determining regions (CDRs) are specified with their residue ranges and corresponding sequences.(Aminoacides depicted is single code format) Region Residue Range with Aho numbering Positions Sequence FWR1 1–23 EVQLVESGGGLVQPGGSLRLSC CDR1 24–42 AASGFTFDDYGMS FWR2 43–56 WVRQAPGKWLEWVS CDR2 57–68 DISWNGGSTY FWR3 69–106 YADSVKGRFTISRDNAVENTLYLQMNSLKPDDTAVYYC CDR3 107–138 AKMGEGGGWGANDY FWR4 139–149 WGQGTQVTVSS Table 3. Summarizing the amino acid substitutions introduced by Llamanade during the humanization of nanobody framework regions (FWRs). FWR Position Original Amino Acid Modified Amino Acid FWR 1 - No changes No changes FWR 2 44 Tryptophan (W) Glycine (G) FWR 3 76 Glutamic Acid (E) Serine (S) 86 Lysine (K) Arginine (R) 87 Proline (P) Alanine (A) 88 Aspartic Acid (D) Glutamic Acid (E) FWR 4 105 Glutamine (Q) Leucine (L) Table 4. Frequency shifts of amino acid residue substitutions in specific framework (FR) regions of a protein sequence. Substitutions are shown using the single-letter amino acid code in the format original residue → new residue, accompanied by their respective frequency changes (%).mutations are recommended by Biophi. Region position Aho numbering position Sequence Residue Frequency (human) FR2 44 H51 W → G 0.3% → 99% FR3 75 H85 A → S 38% → 59% 76 H86 E → K 2% → 92% 87 H97 K → R 9% → 86% 88 H98 P → A 3% → 75% 89 H99 D → E 5% → 93% FR4 115 H144 Q → L 0.8% → 68% Table 5. Impact of Specific Framework Mutations on Binding Affinity. Each mutation is presented with its corresponding effect on affinity (ΔAffinity) measured as the change in binding energy upon substitution. Positive values indicate improved binding, while negative values suggest a slight reduction. Notably, the mutation H:89 (ASH → GLU) exhibits the highest positive effect (+0.6617), significantly enhancing antigen interaction, whereas mutations such as H:115 (GLN → LEU) show minimal negative impact (-0.0222), indicating overall preserved binding performance. Mutation Δ Affinity (kJ/mol) H:44 (TRP à GLY) -0.1441 H:76 (GLU à ARG) 0.4925 H:87 (LYS à ARG) -0.0156 H:88 (PRO à ALA) 0.0073 H:89 (ASH à GLU) 0.6617 H:115 (GLN à LEU) -0.0222 Table 6. Ranking of docking conformations for wild-type nanobody-CD3ε interactions based on binding energy components. The table summarizes key parameters, including electrostatics, desolvation, van der Waals (VdW) forces, and relative restraint energy (relIRST), contributing to the total energy score for each docking conformation. Lower total energy values indicate more favorable binding poses. The top-ranked conformation (#1) exhibits the lowest total energy of -100.335, suggesting the most stable binding interaction (units: kJ/mol). Rank Electrostatics Desolvation VdW relIRST Total 1 -3.514 -53.73 50.571 48.148 -100.335 2 -1.82 -56.152 89.015 48.148 -97.219 3 -11.346 -30.378 23.102 51.852 -91.266 4 -2.606 -56.409 167.936 48.148 -90.369 5 -7.181 -38.622 76.498 48.148 -86.301 6 -1.701 -37.367 -16.937 48.408 -85.206 7 -3.023 -33.609 2.577 44.444 -80.818 8 -0.412 -29.265 45.805 55.556 -80.653 9 -2.109 -38.229 54.385 44.444 -79.344 10 -4.254 -39.853 70.826 40.741 -77.765 Table 7. Ranking of docking conformations of the humanized nanobody based on binding interaction energies. The table summarizes key interaction parameters, including electrostatics, desolvation, van der Waals (VdW) forces, and relative restraint energy (relRST), contributing to the total energy score for each docking conformation. Rankings are based on the total energy (lower values indicate better binding affinity and units are based on kJ/mol). Rank Electrostatics Desolvation VdW relRST Total 1 -2.542 -40.709 21.428 55.556 -96.664 2 -9.652 -32.039 25.908 51.852 -90.952 3 -2.841 -32.532 37.257 55.556 -87.203 4 -12.488 -2.541 -2.443 70.370 -85.643 5 -3.780 -28.727 33.879 55.556 -84.675 6 -4.418 -48.209 124.144 44.444 -84.657 7 -13.591 -8.354 43.740 66.667 -84.238 8 -6.096 -23.035 51.885 59.259 -83.201 9 -11.985 -6.719 67.806 70.370 -82.293 10 -0.440 -29.814 40.144 55.556 -81.796 Additional Declarations No competing interests reported. Supplementary Files table1.docx Table4.docx Table3.docx Table5.docx Table2.docx Table7.docx Table6.docx 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-5769566","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":399875773,"identity":"cc3d95c8-8162-4bdf-987b-194c95aaa993","order_by":0,"name":"Ali Rahmati Bonab","email":"","orcid":"","institution":"Shahid Beheshti University of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Ali","middleName":"Rahmati","lastName":"Bonab","suffix":""},{"id":399875776,"identity":"45985707-ad60-4053-a496-b5656a634afe","order_by":1,"name":"Hannaneh Jalilzadeh Ghahi","email":"","orcid":"","institution":"Shahid Beheshti University of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Hannaneh","middleName":"Jalilzadeh","lastName":"Ghahi","suffix":""},{"id":399875785,"identity":"b0a74073-6723-4f1c-988f-55beaf27df72","order_by":2,"name":"Mahmoud Hassani","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9UlEQVRIiWNgGAWjYFACHhDBzGAA4UnIQUVI0GJMshaGxAZCWnTbzx7d8LHNmsGc/fjjDz93WKRvOH724IMPDHZyug3YtZidyUu7ObMtncGyJ8dMsveMRO6GM3nJhjMYko3NDuDQciDH7DZv22EGgwM5bAy8bUAtQBFpHoYDidtwaTn/Bqrl/PPHH/+2SaQbAEXwa7kBs+VGgoE00JYEgxuEbLnxxuzmjHPpPAZAhrRsm4ThzBtvjA1nGODxy/kcsxsfyqzlDM6nP/74tq1Onu98juGDDxV2cri0wAAiLhTAKg1wKcQG5BtIUT0KRsEoGAUjAQAATrhgs0QbqXMAAAAASUVORK5CYII=","orcid":"","institution":"Shahid Beheshti University of Medical Sciences","correspondingAuthor":true,"prefix":"","firstName":"Mahmoud","middleName":"","lastName":"Hassani","suffix":""},{"id":399875793,"identity":"1c2349e3-1985-4221-862b-5d55a4f1b0eb","order_by":3,"name":"Vahid Jajarmi","email":"","orcid":"","institution":"Shahid Beheshti University of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Vahid","middleName":"","lastName":"Jajarmi","suffix":""},{"id":399875798,"identity":"69d5d654-f76b-4305-9fbb-3525f24898a3","order_by":4,"name":"Javad Ranjbari","email":"","orcid":"","institution":"Shahid Beheshti University of Medical Sciences","correspondingAuthor":false,"prefix":"","firstName":"Javad","middleName":"","lastName":"Ranjbari","suffix":""}],"badges":[],"createdAt":"2025-01-05 22:38:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5769566/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5769566/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":73619708,"identity":"fc6112b2-1e77-4f1e-969d-f39fe3a42dc6","added_by":"auto","created_at":"2025-01-13 04:05:54","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":206944,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eRamachandran plot for the wild type nanobody, showing the distribution of φ (phi) and ψ (psi) torsion angles. Most residues are located within the favored regions, highlighting the structural integrity and conformational quality of the model. Residues in disallowed regions are minimal, indicating minimal steric clashes and a reliable backbone geometry.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/d19c7211af3263580a60c7b2.png"},{"id":73619739,"identity":"03bef2a0-e13a-4f84-a9f5-15c0581cfae9","added_by":"auto","created_at":"2025-01-13 04:05:56","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":204332,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e3D Ramachandran plot for the wild type nanobody, illustrating the frequency distribution of φ (phi) and ψ (psi) torsion angles. Peaks indicate high occurrences of residues within favored regions, affirming structural stability, while minimal distribution is observed in disallowed regions, confirming backbone geometry reliability.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/7d150b9648048554166f735b.png"},{"id":73619703,"identity":"b300b574-57f0-4476-80c0-35921b6301ca","added_by":"auto","created_at":"2025-01-13 04:05:53","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":130774,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eSequence conservation analysis highlighting the variability and functional features of residues. Each amino acid is color-coded based on the conservation scale (1-9), where green represents variable residues and reddenotes highly conserved residues predicted by the neural network algorithm.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/16ff1edc4bbf44800d1fc8ac.png"},{"id":73620652,"identity":"8ec19552-2a24-4b24-ba79-dd950b90c59d","added_by":"auto","created_at":"2025-01-13 04:13:56","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":51579,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eSequence conservation analysis scale indicating residue variability. The scale ranges from 1 (green, variable) to 9 (red, conserved), with insufficient data represented by yellow. No insufficient data was recognized in this study.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/4213d80199e8ccd94c52d7a1.png"},{"id":73619715,"identity":"467b58e9-a74a-4a6f-a794-317dadbf4930","added_by":"auto","created_at":"2025-01-13 04:05:55","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":111983,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eResidue interactions across the interface between Chain A (CD3ε) and Chain B (wild type nanobody) in the complex. The schematic diagram illustrates residue interactions, categorized by residue type and color: positive residues (H, K, R) in blue, negative residues (D, E) in red, neutral residues (S, T, N, Q) in green, aromatic residues (F, Y, W) in magenta, aliphatic residues (A, V, L, I, M) in gray, and proline and glycine residues (P, G) in orange. Hydrogen bonds are represented as blue lines, salt bridges as red lines and non-bonded contacts as orange dashed lines. The thickness of the striped lines indicates the extent of atomic contacts.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/8f2aa915f8a59efb0c91f36f.png"},{"id":73619741,"identity":"5d718c88-c957-46ff-a08f-b35ca29b82d0","added_by":"auto","created_at":"2025-01-13 04:05:56","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":173771,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eRamachandran plot for the humanized nanobody, displaying φ (phi) and ψ (psi) torsion angle distributions. Most residues cluster in favored regions, indicating preservation of structural integrity post-humanization. A small number of residues appear in less favorable regions, suggesting minimal structural deviations.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/53fd33de7d07e363ae569223.png"},{"id":73619759,"identity":"d6c35e3a-b1a6-4a98-bd4e-303bbd519666","added_by":"auto","created_at":"2025-01-13 04:05:57","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":199780,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e3D Ramachandran plot for the humanized nanobody, illustrating the frequency distribution of φ (phi) and ψ (psi) torsion angles. Peaks indicate high occurrences of residues within favored regions, affirming structural stability, while minimal distribution is observed in disallowed regions, confirming backbone geometry reliability.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/8f232b1a5c792bc843cd9b39.png"},{"id":73619728,"identity":"2c19ba01-96d4-4d8d-8620-088ab3c4d0c3","added_by":"auto","created_at":"2025-01-13 04:05:55","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":178493,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eComparison of the Proportion of Residues in Favored Regions (pb) between Humanized and Wild Nanobodies. The humanized nanobody shows a higher proportion of residues in favored regions (pb = 0.963, blue) compared to the wild nanobody (pb = 0.927, green), with the red dashed line indicating the wild nanobody baseline.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure8.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/48cef518b94cb5f205fb92c8.png"},{"id":73620648,"identity":"f04eee88-4ec3-4ead-b599-e29fcaacb776","added_by":"auto","created_at":"2025-01-13 04:13:56","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":84008,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eStructural alignment of wild type and humanized nanobody, the green structure represents the wild type nanobody, while the orange structure corresponds to the humanized version. Both nanobody variants were analyzed using NanoNet for prediction, providing insights into their structural and functional characteristics.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure9.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/9adb810f1a432a8afd8df1d3.png"},{"id":73619788,"identity":"36743e1a-3385-4b49-abcc-66bd20b96178","added_by":"auto","created_at":"2025-01-13 04:05:58","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":97591,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eResidue interactions across the interface between Chain A (CD3ε) and Chain B (humanized nanobody) in the complex. The schematic diagram illustrates residue interactions, categorized by residue type and color: positive residues (H, K, R) in blue, negative residues (D, E) in red, neutral residues (S, T, N, Q) in green, aromatic residues (F, Y, W) in magenta, aliphatic residues (A, V, L, I, M) in gray, and proline and glycine residues (P, G) in orange. Hydrogen bonds are represented as blue lines, salt bridges as red lines and non-bonded contacts as orange dashed lines. The thickness of the striped lines indicates the extent of atomic contacts.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure10.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/d77778b8c016e0a3673276a9.png"},{"id":73619706,"identity":"1d9e5337-b209-4439-9c9a-4efaca72f57b","added_by":"auto","created_at":"2025-01-13 04:05:54","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":115592,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eComparison of docking scores between humanized and wild type nanobody libraries. The histogram depicts the scoring energy distribution, with humanized nanobodies shown in blue and wild type nanobodies in yellow. Lower docking scores indicate better binding affinities. (x axis units: kJ/mol).\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure11.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/3f4ff11ee7e6cceadd94b6f0.png"},{"id":73620639,"identity":"171c72cf-abc1-483c-9b09-88e44d5f3ec5","added_by":"auto","created_at":"2025-01-13 04:13:54","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":102783,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eComparison of thermal stability (Tm) for wild type and humanized nanobody in lysate and cell-based environments.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure12.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/fd8c31b8de82bd03900fb3f0.png"},{"id":73619782,"identity":"5debd458-945b-4f1d-be19-a1b5cae16ca0","added_by":"auto","created_at":"2025-01-13 04:05:58","extension":"png","order_by":13,"title":"Figure 13","display":"","copyAsset":false,"role":"figure","size":659506,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eComparison of pKa Values Between Wild type and Humanized Nanobody. The bar graph displays the pKa values for key residues in the Wild type (blue) and Humanized (orange) nanobody structures. Variations in pKa values are evident across several residues, reflecting differences in electrostatic properties introduced by humanization. Residues such as A-46-GLU and A-94-ASP exhibit notable shifts, emphasizing regions where humanization impacts biochemical properties.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure13.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/1c9590e55a4097e47dc0cf7f.png"},{"id":73619757,"identity":"a9e56685-8a1c-4249-be7d-5057b5616dde","added_by":"auto","created_at":"2025-01-13 04:05:57","extension":"png","order_by":14,"title":"Figure 14","display":"","copyAsset":false,"role":"figure","size":374183,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eComparison of Biochemical Properties Between Wild Type and Humanized Nanobody. The chart visualizes the differences in various biochemical properties, calculated as the values in the Humanized nanobody minus those in the Wild Type. Positive bars indicate increases in properties due to humanization, while negative bars represent decreases. Notable changes are seen in properties such as the number of specific amino acids (e.g., K-R, D-E), pI (isoelectric point), and hydrophobicity-related metrics (e.g., Kyte-Doolittle and fold propensity), reflecting the structural and functional impacts of the humanization process..\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure14.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/5f4fc1bc7cd8e93ff173abc2.png"},{"id":73619731,"identity":"c310fd2f-0452-409d-a657-0177dc55e798","added_by":"auto","created_at":"2025-01-13 04:05:55","extension":"png","order_by":15,"title":"Figure 15","display":"","copyAsset":false,"role":"figure","size":330337,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eBoxplots comparing the Relative Solvent Accessibility (RSA) and Accessible Surface Area (ASA) between wild type and humanized nanobody structures. The left plot shows RSA values, while the right plot displays ASA values, highlighting the similarity in structural exposure and surface area between the two versions. Outliers are marked for ASA values in the humanized nanobody.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure15.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/218b74947e50fc30124fa147.png"},{"id":73619768,"identity":"df06a779-041c-4d8e-96b1-61b8c6b2ffc8","added_by":"auto","created_at":"2025-01-13 04:05:57","extension":"png","order_by":16,"title":"Figure 16","display":"","copyAsset":false,"role":"figure","size":387149,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eThis line plot compares the residue-wise aggregation scores between wild type and humanized nanobodies.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure16.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/62a2afc4d9d9dd99e9c24882.png"},{"id":73619751,"identity":"849543d9-968b-458e-8dff-a2aa6618accf","added_by":"auto","created_at":"2025-01-13 04:05:56","extension":"png","order_by":17,"title":"Figure 17","display":"","copyAsset":false,"role":"figure","size":362774,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eLine graph showing the comparison of the Radius of Gyration (total) over time for wild type and humanized nanobody structures. The x-axis represents time in picoseconds (ps), while the y-axis indicates the Radius of Gyration in nanometers (nm). The plot highlights slight fluctuations and overall structural compactness differences between the two variants.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure17.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/2b09aa1238589c2d93b2bca6.png"},{"id":73619748,"identity":"008ae06e-98ea-41c8-a23a-40037acc3145","added_by":"auto","created_at":"2025-01-13 04:05:56","extension":"png","order_by":18,"title":"Figure 18","display":"","copyAsset":false,"role":"figure","size":303512,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eLine graph of Root Mean Square Deviation (RMSD) over time (ps) for humanized and wild-type nanobody structures, showing comparable structural stability with slightly higher deviations in the humanized variant.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure18.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/27c618e5f4342b55851d628b.png"},{"id":73620758,"identity":"df3f43d5-63e2-420e-9a1d-7cf5352edc88","added_by":"auto","created_at":"2025-01-13 04:21:56","extension":"png","order_by":19,"title":"Figure 19","display":"","copyAsset":false,"role":"figure","size":474021,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eThe graph shows the RMSD trends over time for CDR regions of humanized and wild nanobodies.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure19.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/a9be5ac5f73ee2e60d3d380d.png"},{"id":73620645,"identity":"56d076ce-732e-49b3-b643-74cbe57d11ae","added_by":"auto","created_at":"2025-01-13 04:13:55","extension":"png","order_by":20,"title":"Figure 20","display":"","copyAsset":false,"role":"figure","size":197618,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eRMSF Comparison between Wild-type and Humanized Structures Highlighting Mutations and CDR Regions. The Root Mean Square Fluctuation (RMSF) profiles of wild-type (blue line) and humanized (orange line) structures are compared across the residue index (1-120). Regions corresponding to the Complementarity Determining Regions (CDRs) are highlighted with yellow shading for CDR1, CDR2, and CDR3. Vertical dashed lines indicate the positions of specific mutations, while the horizontal red dashed line represents a reference RMSF threshold of 0.00 nm.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure20.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/594373ccc0d0d163046fd7d1.png"},{"id":73620658,"identity":"4fb42bbb-46b1-452e-b76c-38253988f2af","added_by":"auto","created_at":"2025-01-13 04:13:57","extension":"png","order_by":21,"title":"Figure 21","display":"","copyAsset":false,"role":"figure","size":212075,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eComparison of Neq​ values between Wild Type and Humanized Nanobody Variants across Residues 1–120. Highlighted regions represent the CDR1, CDR2, and CDR3, with vertical dashed lines indicating mutation sites.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"figure21.png","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/1493bec09341c78d4fc4cf53.png"},{"id":83347930,"identity":"05890446-43e0-406e-8887-98ea7b5b03cf","added_by":"auto","created_at":"2025-05-23 13:01:19","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6149710,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/edb9fed7-e9fd-40e6-95b2-a324c875b7b0.pdf"},{"id":73619702,"identity":"6cd8b791-719c-43fb-9137-85eeedf4be89","added_by":"auto","created_at":"2025-01-13 04:05:53","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":15407,"visible":true,"origin":"","legend":"","description":"","filename":"table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/0fa0d0f8887d050b9c9dfe5f.docx"},{"id":73619755,"identity":"6867d6fa-69d5-43ed-9621-a49df55e53af","added_by":"auto","created_at":"2025-01-13 04:05:57","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":16209,"visible":true,"origin":"","legend":"","description":"","filename":"Table4.docx","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/8027d8239fd2203683cb5f37.docx"},{"id":73619733,"identity":"9a737127-33fd-4de7-9a92-3cdd3f43a67e","added_by":"auto","created_at":"2025-01-13 04:05:56","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":15913,"visible":true,"origin":"","legend":"","description":"","filename":"Table3.docx","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/1b7629095f5944bb8f161c94.docx"},{"id":73620654,"identity":"3c0c4c01-14a1-4b32-8160-6c171dbb3a6a","added_by":"auto","created_at":"2025-01-13 04:13:57","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":16085,"visible":true,"origin":"","legend":"","description":"","filename":"Table5.docx","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/327c5e865bf38613331bba17.docx"},{"id":73620642,"identity":"d619c9fd-51d3-4150-98d1-237fee7ecca2","added_by":"auto","created_at":"2025-01-13 04:13:55","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":15785,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.docx","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/b63c60e1c026b9a9c31f999e.docx"},{"id":73620638,"identity":"c5e0151f-8c6d-4327-8555-c2c1d62bc6d6","added_by":"auto","created_at":"2025-01-13 04:13:54","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":16543,"visible":true,"origin":"","legend":"","description":"","filename":"Table7.docx","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/bc650c2917b2cd8448be2254.docx"},{"id":73620640,"identity":"65040282-5364-4c78-a1be-4a8aebe2ba7f","added_by":"auto","created_at":"2025-01-13 04:13:54","extension":"docx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":16960,"visible":true,"origin":"","legend":"","description":"","filename":"Table6.docx","url":"https://assets-eu.researchsquare.com/files/rs-5769566/v1/f422fcfce40ce549d9ee11e4.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Computational and Machine Learning Approaches for Optimizing Anti-CD3ε Nanobody: Humanization and Characterization for Enhanced Therapeutic Efficacy","fulltext":[{"header":"Introduction","content":"\u003cp\u003eNanobodies, derived from camelid heavy-chain antibodies, are small (~\u0026thinsp;15 kDa), stable, and soluble, with demonstrated potential in treating cancer, infectious diseases, neurodegenerative disorders, and autoimmune conditions. Their small size enhances tissue penetration, while production in microbial systems is cost-effective [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Examples include Caplacizumab, an FDA-approved nanobody for aTTP [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Nanobodies surpass scFvs in structural stability, solubility, and ease microbial production, offering superior tissue penetration [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Nanobody applications extend to neutralizing agents [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], receptor blockers[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], and drug-delivery vehicles [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e],\u003c/p\u003e \u003cp\u003eOne notable application involves anti-CD3 nanobodies, which specifically bind the CD3ε chain to activate T cells. The CD3ε chain plays a critical role in TCR signaling, traditionally activated through TCR-MHC interactions. However, recent studies have revealed MHC-independent activation mechanisms involving structural rearrangements or direct molecular interactions, enabling immune responses without antigen presentation [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. This is particularly advantageous in environments with MHC downregulation, enhancing immune surveillance and offering promising potential for immunotherapy [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].Activation through anti-CD3ε nanobodies promotes T-cell proliferation, cytokine production (e.g., IL-2 and IFNγ), and a Th1-dominant immune response while simultaneously suppressing tolerogenic cytokines [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Preclinical models demonstrate their efficacy in tumor suppression and enhanced immune surveillance, with clear advantages in stability, size, and immune activation compared to monoclonal antibodies [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Despite their therapeutic versatility, anti-CD3 nanobodies require humanization to address their lower intrinsic humanness scores, thereby minimizing immunogenicity in clinical applications [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eTo address this challenge, computational tools such as machine learning and deep learning have emerged as transformative approaches. These methods have demonstrated their ability to efficiently humanize nanobodies, ensuring compatibility with the human immune system while preserving their functional properties [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. By leveraging these advanced tools, researchers can overcome immunogenicity barriers, enhancing the therapeutic potential of nanobody-based treatments\u003c/p\u003e \u003cp\u003eIn this study, we leverage a suite of advanced computational methods to enhance the modeling, humanization, and characterization of nanobody structures. Deep learning (DL) and machine learning (ML) approaches, including AGGRESCAN3D [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] and DeepSTABp [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], are utilized to evaluate aggregation propensity [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] and stability of nanobody structures. For structure prediction specific to nanobodies, NanoNet [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] is employed, demonstrating exceptional accuracy and outperforming traditional tools such as AlphaFold [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e], ESMFold [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e], and Yang-Server [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. These methods enable precise identification of structural features, such as complementarity-determining regions (CDRs), framework regions (FWRs) and aggregation hotspots, which are critical for designing stable and functional nanobody therapeutics[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. To address the immunogenicity associated with nanobodies derived from camelid antibodies, we incorporate ML-based humanization tools, such as Llamanade [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e] and BioPhi [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. These tools allowed us to optimize sequence similarity to the human antibody repertoire while maintaining the binding affinity and structural stability of the nanobody. By combining these state-of-the-art methods, we conducted a comprehensive comparison between humanized and wild type nanobodies. This analysis includes evaluations of thermal stability, aggregation propensity, and binding efficiency, alongside other critical metrics that influence their therapeutic potential. The integration of DL and ML-driven tools throughout this process not only ensures enhanced accuracy and efficiency but also underscores the transformative potential of computational approaches in nanobody engineering and optimization.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eSequence Acquisition for Anti-CD3\u0026epsilon; Nanobody and CD3\u0026epsilon;\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe sequences for the anti-CD3\u0026epsilon; nanobody and the CD3\u0026epsilon; were obtained from the patent WO 2010/037838 A2 [27] shown in Table 1.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNanobody Structural Prediction, Visualization, Validation, and Comparison\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe anti-CD3\u0026epsilon; nanobodies (wild type and humanized) were modeled using NanoNet [28], a specialized tool for nanobody structural prediction. To analyze and visualize the resulting nanobody structures, PyMOL[29] and BioLuminate [30] were employed for detailed 3D modeling. The structural integrity of the humanized and wild type nanobody models was validated using Ramachandran plots, generated with the RamPlot [31], to assess residue conformational quality [32]. Comparative analysis of the nanobody structures was conducted using Template Modeling-align (TM-align) from the Zhang Lab [33], which evaluates structural similarity through metrics such as Template Modeling score (TM-score), Root Mean Square Deviation (RMSD), and the percentage of aligned residues [34]. These metrics provide a quantitative basis for assessing the impact of humanization on the structural integrity and alignment of the nanobody framework, ensuring a thorough evaluation of the modeled structures. The structural comparison was visualized using the RCSB PDB 3D viewer [35] which effectively illustrated the minimal differences between the two nanobody variants.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStructural Prediction and Energy Minimization of CD3\u0026epsilon;\u0026nbsp;\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe CD3\u0026epsilon; protein structure was predicted using the RoseTTAFold All-Atom model [36], a machine learning tool for high-accuracy structural predictions. The amino acid sequence of CD3\u0026epsilon; was used as input, and the resulting model was evaluated for confidence and alignment quality using predicted Local Distance Difference Test (pLDDT) and Predicted Aligned Error (PAE) metrics [37]. The structure was refined using molecular dynamics simulations via the AMBER Relaxation protocol [38] , with energy values recorded to assess stability improvements. Finally, the accuracy of the refined model was evaluated by comparing it to the experimental 1XIW structure [39], focusing on framework regions and flexible loops for alignment.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCDR and FWR Identification\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe anti-CD3\u0026epsilon; nanobody sequence was analyzed using ANARCI [40], employing the North/Aho numbering scheme for accurate delineation of CDRs and \u0026nbsp;FWRs [41, 42]. \u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHumanization via CDR Grafting\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCDR grafting has been employed to humanize nanobodies by transferring CDRs onto human antibody frameworks [43] while preserving critical Vernier zone residues to ensure structural integrity and binding affinity [44]. The process was optimized using Llamanade and evaluated with BioPhi to measure sequence similarity to human antibody repertoires, ensuring minimal disruption of CDR conformation. Additionally, the T20 Score Scale has been used to quantify the degree of humanization and predict immunogenicity risks [45], thereby enhancing therapeutic compatibility. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDocking and Binding Analysis\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDocking studies evaluated interactions between the CD3\u0026epsilon; antigen and two nanobody variants: humanized and wild type. A two-step docking approach was utilized [46], beginning with global docking using BioLuminate, which sampled 70,000 rotational poses to ensure comprehensive coverage of binding conformations, with parameters applied for energy minimization and analyzing interactions within an 8 \u0026Aring; radius, identifying significant binding contributions from the CDRs, mutation for humanization candidate residues, and Vernier zone residues in the nanobody FWR. This was followed by Site-Specific docking using PyDock [47, 48], specifically targeting amino acids 1\u0026ndash;27 of CD3\u0026epsilon;, informed by patent data indicating the nanobody binds this segment. The resulting docking conformations were analyzed and visualized using PDBsum [49] to evaluate the structural features of the docked complexes.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSurface Conservation Analysis\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe ConSurf web server [50] was utilized to assess surface residue conservation by generating an evolutionary conservation profile of the nanobody sequence. This analysis involved aligning the nanobody sequence against homologous sequences retrieved from the UniProt database to construct a multiple sequence alignment (MSA). Evolutionary conservation scores were subsequently calculated using the Bayesian inference method [51], as described in the ConSurf methodology. Residues with higher conservation scores were identified as critical for structural and functional integrity [52].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNanobody Thermostability and Physicochemical Property Analysis\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe thermal stability of wild type and humanized nanobody variants was analyzed using DeepSTABp, which predicts \u0026Delta;Tm values based on amino acid sequences under lysate-based and cellular environments at 36.9\u0026deg;C [17]. This machine-learning tool evaluates thermal stability using sequence-derived features without requiring structural data. The pKa values of ionizable residues were predicted with SystemBuilder [53], highlighting shifts in ionization behavior post-humanization[54]. Protein-Sol [55] and NetSurfP [56, 57] were employed to analyze solubility, charge, and structural propensities. Hydrophobicity was assessed using two tools: SODOPE [58, 59], which calculates the GRAVY index, and Protein-Sol, which applies the Kyte-Doolittle scale.[60]\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIon mobility collision cross-section (CCS) prediction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCCS, a measure of an ion\u0026apos;s effective surface area, was calculated using Rosetta integrated with the PARCS algorithm for accurate and rapid estimations directly from protein structures [61, 62]. CCS analysis of wild type and humanized nanobody structures highlighted differences in surface area, folding dynamics, and molecular compactness introduced by humanization, providing valuable insights into their stability and functionality [63, 64].\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAggregation Propensity Prediction\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAGGRESCAN3D was used to predict aggregation hotspots in nanobodies. This machine-learning-based tool evaluates the aggregation propensity of 3D protein models, identifying destabilizing regions that could lead to aggregation.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eToxicity and Allergenicity Prediction\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNanobody toxicity was evaluated using ToxinPred [65], a computational tool that employs metrics such as the Motif-Extraction-Reliable Classification Indices (MERCI) Score to detect toxicity-prone sequence motifs and Hybrid Scores, which combine predictive parameters for a comprehensive toxicity assessment [66]. Allergenicity was assessed using AlgPred, which integrates allergenic motif databases, sequence features, and Support Vector Machine (SVM) based predictions [67] to generate a robust allergenicity profile [68].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMolecular Dynamics\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eSimulations\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe system was solvated using the TIP3P water model in a cubic box with a 1.0 nm solute-box edge distance and neutralized with Na⁺ and Cl⁻ ions. Steepest descent minimization optimized ion placement and energy, with long-range electrostatics and Van der Waals interactions handled via PME and a 1.0 nm cut-off. Production Molecular Dynamics (MD) was conducted for 100 ps (50,000 steps) with a 2 fs time step, using the leap-frog integrator, V-rescale thermostat (300 K), and Parrinello-Rahman barostat (1.0 bar) [69]. Data was recorded at set intervals with automatic checkpointing. Analysis focused on C-alpha atoms to study the structural dynamics and stability of wild type and humanized nanobodies.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eStructural Prediction, Humanization, and Docking Analysis of the Wild Type Nanobody\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe 3D structure of the wild type nanobody was predicted using NanoNet [28] and validated through a Ramachandran plot (Figures 1 and 2), which revealed a PB scale score of 0.927, indicating that most residues were in favored regions and confirming the structure\u0026apos;s quality and nanobody sequence was analyzed for North/Aho numbering and accurate delineation of CDRs and FWRs. The results are summarized in Table 2:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eUpon applying the Llamanade humanization process, six mutations were identified as necessary for humanization using the CDR grafting method (Table 3) and BioPhi suggested seven mutations, including an additional mutation at position 75 (A \u0026rarr; S) (Table 4).\u003c/p\u003e\n\u003cp\u003eDespite BioPhi\u0026apos;s recommendation, residue frequency analysis indicated that the frequency of position 75 increased only modestly from 38% to 59% (a 21% rise), which is a minor change compared to other positions. Furthermore, conservation analysis (Figure 3) classified position 75 as highly conserved, with a conservation scale score of 8 (Figure 4), suggesting potential stability and binding alterations if mutated. Given its high conservation and negligible impact on humanness, the mutation at position 75 was not implemented, leaving this residue unchanged.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe CD3\u0026epsilon; protein structure was predicted using the RoseTTAFold All-Atom model achieving a mean pLDDT score of 73.674, indicative of moderate to high confidence, with higher precision observed in conserved regions. The model demonstrated a mean PAE of 8.307, indicating a good residue-residue alignment. Following this, energy minimization, this process improved the model\u0026apos;s stability and accuracy, with the computed energy values decreasing from 3696.127 kJ/mol (initial) to -1177.263 kJ/mol (final). In the final structure, the framework regions demonstrated strong alignment with the 1XIW template, while the flexible loops exhibited moderate improvements in both confidence and alignment.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFollowing global docking of the wild type nanobody with CD3\u0026epsilon;, all six defined residues were mutated to assess their impact on binding affinity. BioLuminate analysis showed negligible changes in binding affinity (Table 5). The strongest interactions and closest distances remained localized in the CDR regions, where the nanobody binds the antigen.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFollowing site-specific docking analysis using PyDock, the first model was selected for further investigation due to its lowest energy value (\u0026Delta;G = -100.33 kcal/mol), indicative of strong binding affinity (Table 6). In this model, a key interaction was observed between tryptophan (W) at position 44 within the nanobody\u0026rsquo;s FWR2 region and specific amino acid residues of CD3\u0026epsilon; (Figure 5).\u003c/p\u003e\n\u003cp\u003eTryptophan 44 exhibited substantial interactions, emphasizing its critical role in stabilizing the antigen-binding interface. Notably, in nine out of the top ten ranked docking poses, the interaction between W44 and CD3\u0026epsilon; was consistently observed. However, detailed descriptions of these additional models were omitted for brevity.\u003c/p\u003e\n\u003cp\u003eGiven these findings, we opted to retain W44 in its native form while mutating the remaining five residues to achieve humanization, without compromising the structural integrity or functional performance of the nanobody.\u003c/p\u003e\n\u003cp\u003eAs shown in Figure 3, the conservation analysis of residues at positions 88 and 105 revealed moderate to low conservation scores. (Figure 4), and mutations at these positions significantly increased the human residue frequency, enhancing the nanobody\u0026rsquo;s \u0026quot;humanness\u0026quot; and reducing immunogenicity. In addition, residues at positions 76, 87, and 89 displayed above-average conservation scores, and mutations at these positions also significantly increased the human residue frequency, like positions 88 and 105 enhancing the nanobody\u0026rsquo;s humanness level and reducing immunogenicity. Further analysis across various aspects in the following sections of this article demonstrated that all five mutations (at positions 76, 87, 88, 89, and 105) had minimal impact on the nanobody\u0026apos;s affinity, stability, or other functional properties, despite the conservation scores from Consurf, which may seem concerning.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe predicted the 3D structure of this nanobody sequence using NanoNet, modeled similarly to the wild type structure. Validation was performed using a Ramachandran plot, which revealed that the humanized nanobody exhibited a higher proportion of residues in favored regions (96.3%) compared to the wild type (92.7%) (Figures 6 and 7). This difference (\u0026Delta;Pb = 3.6%) suggests that the humanization process enhanced structural regularity and optimization. These findings indicate that the humanized nanobody is slightly more conformationally favorable than the wild type counterpart (Figure 8).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe structural comparison between the humanized and wild type nanobody revealed an RMSD of 0.25 \u0026Aring;, indicating minimal deviation, with 95.8% of residues aligning between the two structures. Furthermore, a TM score of 0.996 confirmed near-complete structural equivalence, highlighting that the humanized nanobody effectively retains the 3D conformation of its wild type counterpart. These results emphasize the successful preservation of the wild type nanobody\u0026rsquo;s structural integrity in its humanized version (Figure 9).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAll subsequent nanobody characterizations (wild type and humanized) discussed in the following sections are based on NanoNet-generated 3D structural PDB files.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe conducted docking analysis again with PyDock, this time on the humanized nanobody, to evaluate its interaction with the target epitope of \u0026nbsp;CD3\u0026epsilon; (1-27 aa). Among the top 10 models with the lowest binding energies, the model with the lowest energy (-96.64 kJ/mol) was selected for further analysis (Table 7). The docking analysis revealed that tryptophan at position 44 (W44) in the FWR of the humanized nanobody maintains its interaction with residues of the CD3\u0026epsilon; antigen, similar to the wild type nanobody(Figure 10). This conserved interaction underscores the structural and functional integrity of the humanized nanobody. The comparison of interface statistics between the wild type and humanized nanobody complexes with CD3\u0026epsilon; highlights only slight differences, demonstrating that the humanization process successfully retained much of the nanobody\u0026apos;s binding characteristics. In the wild type complex, the CD3\u0026epsilon; chain (Chain A) has 18 interface residues with an interface area of 947 \u0026Aring;\u0026sup2;, while the nanobody (Chain B) has 13 residues and an interface area of 791 \u0026Aring;\u0026sup2;. In the humanized nanobody complex, the CD3\u0026epsilon; interface residues decrease slightly to 14, and the interface area reduces modestly to 728 \u0026Aring;\u0026sup2;, while the nanobody\u0026apos;s interface area remains similar at 725 \u0026Aring;\u0026sup2;. These small variations suggest that the humanized nanobody maintains strong structural compatibility with CD3\u0026epsilon;.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn terms of interactions, the wild type nanobody forms 5 hydrogen bonds and 200 non-bonded contacts with CD3\u0026epsilon;, while the humanized nanobody has 1 hydrogen bond and 165 non-bonded contacts. While there is a minor reduction in polar and non-bonded interactions, the overall difference is not significant, indicating that the humanized nanobody still exhibits a robust binding potential. Additionally, we observed that there were no disulfide bonds formed either within the nanobody structures (intra-molecular) or between the nanobody and CD3\u0026epsilon; (inter-molecular) in both the wild type and humanized models.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWe conducted a comparative analysis of docking models generated using PyDock for both wild type and humanized nanobodies. As shown in Figure11, the docking scores for the two nanobody types against CD3\u0026epsilon; exhibit a similar overall distribution, indicating that the humanization process maintained the nanobody\u0026apos;s binding characteristics. Both models show a wide energy range (-100 to +40), with the wild type nanobody displaying a slightly higher frequency in the most favorable scoring range. The humanized nanobody closely follows this pattern, with only minor differences in peak frequencies and scoring spread. These findings suggest negligible differences in the binding. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHumanness Score Analysis\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe wild type nanobody achieved a T20\u003cspan dir=\"RTL\"\u003e\u0026nbsp;\u003c/span\u003eFWR score of 85.74 and an overall T20 score of 82.83. After humanization, the nanobody demonstrated improved metrics, with a T20 FWR score of 91.03 and an overall T20 score of 87.04.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnalysis of Nanobody Thermostability and Physicochemical Properties\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe humanized nanobody exhibited superior thermal stability compared to the wild type, with higher melting temperatures (Tm) in both lysate (54.56\u0026deg;C vs. 51.09\u0026deg;C) and cell-based environments (53.19\u0026deg;C vs. 51.72\u0026deg;C), indicating enhanced robustness (Figure 12). Post-humanization, pKa values for ionizable residues showed minimal changes, preserving key chemical properties (Figure 13). Humanization also reduced the absolute charge, slightly increased hydrophobicity, and maintained stable beta-strand propensity (Figure 14). Solubility metrics, including Relative Solvent Accessibility (RSA) and Accessible Surface Area (ASA), revealed negligible differences (Cohen\u0026apos;s d \u0026lt; 0.2), and both nanobodies were moderately hydrophilic, as reflected by Grand Average of Hydropathicity (GRAVY) indices of -0.40 for wild type and -0.32 for humanized nanobody (Figure 15).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIon mobility collision cross-section (CCS) prediction and Aggregation Score Analysis\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe humanized nanobody exhibited a CCS value of 1226.86 \u0026Aring;\u0026sup2;, slightly smaller than the wild type nanobody\u0026rsquo;s CCS value of 1234.85 \u0026Aring;\u0026sup2;. \u003cstrong\u003eAggregation Propensity Prediction\u003c/strong\u003e As Figure 16 shows the humanized nanobody exhibited a slightly less negative mean aggregation score (-0.6447) compared to the wild type (-0.7001).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eToxicity and Allergenicity Prediction\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBoth the wild type and humanized nanobodies are predicted to be non-toxic, with no sequences classified as \u0026quot;Toxin.\u0026quot; The humanized nanobody shows a slightly higher ML Score (0.21 vs. 0.175), while other toxicity-related scores, such as MERCI and Hybrid scores, remain identical between the two versions. Regarding allergenicity, the wild type nanobody has a score of -0.14507017, which is above the threshold of -0.4, classifying it as an allergen. In contrast, the humanized nanobody achieves a score of -0.52723881, below the threshold, and is therefore classified as a non-allergen.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMolecular Dynamics Simulations\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe combined analyses of radius of gyration (Figure 17), RMSD (Figures 18, and 19), Root Mean Square Fluctuation (RMSF, Figure 20), and Neq (Figure 21) provide a comprehensive understanding of the structural dynamics of the wild type and humanized nanobody. RMSD analysis shows similar global structural deviations, stabilizing around ~0.06\u0026ndash;0.08 nm, indicating preserved global stability after humanization. RMSF values remain low in the framework regions (~0.1\u0026ndash;0.3 nm) for both structures, confirming the conserved scaffold\u0026apos;s structural integrity. Region-specific differences were observed in the CDRs: in CDR1 (residues 23\u0026ndash;35), RMSF increased (~0.25\u0026ndash;0.4 nm to ~0.3\u0026ndash;0.45 nm), as did Neq (~1.0 to ~1.1), reflecting enhanced flexibility. CDR2 (residues 50\u0026ndash;59) showed minimal changes in RMSF (~0.2\u0026ndash;0.35 nm) and Neq (~1.05 to ~1.1), indicating structural conservation. In contrast, CDR3 (residues 97\u0026ndash;109) exhibited reduced RMSF (~0.35\u0026ndash;0.6 nm to ~0.3\u0026ndash;0.5 nm) and Neq (~1.3 to ~1.1), suggesting loop stabilization.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe Rg analysis further supports these findings, with the humanized nanobody showing a mean Rg of 1.366 nm and the wild type 1.361 nm, reflecting comparable stability. Axis-specific Rg values suggest that the humanized nanobody exhibits more balanced stability along the X and Z axes (1.163 nm and 1.210 nm, respectively), compared to the wild type\u0026apos;s anisotropy, more compact along the Y-axis (0.900 nm). These localized adjustments in flexibility and stabilization, particularly in the CDRs, suggest that humanization preserved global structural integrity while introducing changes that could influence antigen-binding dynamics.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Visualization\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGraphPad Prism version 9.0 was utilized for generating comparative line graphs and statistical plots, while Python enabled the inclusion of advanced customizations such as shaded regions, annotations, and multi-layered data visualizations.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe findings of this study highlight the transformative role of advanced computational methods in optimizing nanobody therapeutics. By comparing the wild type and humanized anti-CD3ε nanobodies, we demonstrate the advantages of a data-driven approach to enhance their structural, functional, and therapeutic properties. Using NanoNet, the structural modeling confirmed that the humanized nanobody retained the overall 3D conformation of the wild type, with an RMSD of 0.25 \u0026Aring; and a TM-score of 0.996. These metrics validate the structural preservation achieved through humanization. Furthermore, the humanized nanobody exhibited improved conformational variability, which is essential for maintaining dynamic adaptability and functionality in biological environments. Enhanced thermal stability was observed, with melting temperatures (Tm) increasing by 3.47\u0026deg;C in lysate and 1.47\u0026deg;C in cell-based environments. These results underscore the success of the humanization process in bolstering structural robustness while preserving flexibility.\u003c/p\u003e \u003cp\u003eThe functional importance of framework residues, particularly those in FWR2, has been well-documented [\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e]. Alterations in these residues can significantly influence the conformation of the complementarity-determining regions (CDRs), thereby affecting the nanobody's overall affinity and stability [\u003cspan additionalcitationids=\"CR72\" citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e]. Previous studies have also demonstrated that framework region residues can affect CDR conformation and, consequently, the binding affinity of antibodies [\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e]. These insights underscore the necessity of accounting for FWR residues, particularly FWR2, during the humanization process to preserve nanobody functionality and stability. In this article, it is demonstrated that tryptophan at position 44 in the framework region (FWR) has a significant interaction with its antigen. This finding underscores the critical role of amino acids in the FWR, as they influence intermolecular interactions, stability, and the affinity of nanobodies. Consequently, during the mutation of residues in the FWR for humanization, this factor must be carefully taken into account to preserve nanobody functionality and stability.\u003c/p\u003e \u003cp\u003eDocking analyses revealed comparable binding affinities for both variants, with the humanized nanobody retaining critical binding interactions with CD3ε. However, this reduction was negligible and did not significantly impact overall antigen-binding performance. Importantly, the improved conformational variability of the humanized nanobody likely offsets this minimal decrease, supporting its functional reliability in therapeutic contexts.\u003c/p\u003e \u003cp\u003eThe comparison of global and site-specific docking underscores the significance of site-specific docking methods in nanobody humanization. In global docking performed using BioLuminate software, no interaction was observed within 8 \u0026Aring; of W44. This outcome likely stems from the nature of global docking, where the nanobody may interact with other epitopes on the antigen that are not of therapeutic interest but, site-specific docking with PyDock revealed strong interactions between W44, located in the FWR2 region of the nanobody, and the CD3ε desired epitope. PyDock targeted the 1\u0026ndash;27 amino acid region of CD3ε, the therapeutically relevant epitope critical for this nanobody's activity as outlined in patent WO 2010/037838 A2 [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAGGRESCAN3D analysis confirmed reduced aggregation propensity for the humanized nanobody, reflecting its enhanced stability and suitability for clinical use. Solubility metrics remained consistent between the two variants, indicating that humanization preserved key physicochemical properties. CCS analysis showed a slight reduction in molecular size for the humanized nanobody (1226.86 \u0026Aring;\u0026sup2; vs. 1234.85 \u0026Aring;\u0026sup2; for the wil type), suggesting improved molecular compactness and potential benefits for pharmacokinetics.\u003c/p\u003e \u003cp\u003eThe humanized nanobody exhibited a significant reduction in allergenicity score, transitioning from allergenic (wild type) to non-allergenic (humanized). Both nanobodies were classified as non-toxic, confirming their suitability for therapeutic applications. Molecular dynamics simulations further revealed that the humanized nanobody demonstrated improved conformational variability. While CDR1 exhibited increased flexibility, CDR3 showed stabilization, contributing to a balance between adaptability and structural integrity. The humanization process not only preserved the structural and functional integrity of the nanobody but also significantly enhanced its humanness, as evidenced by the T20 score results. The humanized nanobody achieved an overall T20 score of 87.04 compared to 82.83 for the wild type nanobody, with notable improvements in the framework region (91.03 vs. 85.74). These higher scores underscore the effectiveness of the humanization strategy in aligning the nanobody sequence closer to the human antibody repertoire. The combination of DL and ML tools such as BioPhi, NanoNet, LlamaNADE and AGGRESCAN3D enabled precise humanization and optimization of the nanobody. Improved humanness scores, reduced aggregation, enhanced conformational variability, and preserved antigen-binding capabilities establish the humanized nanobody as a superior therapeutic candidate.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study demonstrates the potential of integrating advanced computational tools to design optimized nanobody therapeutics. The humanized anti-CD3ε nanobody exhibited enhanced thermal stability, reduced aggregation propensity, and improved conformational variability, making it well-suited for clinical applications. Despite a negligible decrease in CDR3 binding affinity, the nanobody's overall performance, including preserved binding efficiency and reduced immunogenicity, underscores the robustness of the humanization process. These findings validate the role of DL and ML techniques in advancing nanobody engineering, and provide a strong foundation for the development of next-generation biologics targeting immune-related and other diseases. This study highlights how computational strategies can bridge the gap between molecular design and therapeutic application.\u003c/p\u003e \u003cp\u003eDespite the significant insights gained, several limitations must be acknowledged. The findings are primarily based on computational modeling, with no in vitro or in vivo validation, limiting their direct clinical applicability. Additionally, while aggregation scores were reduced, the long-term stability and solubility under varying physiological conditions remain unexplored.\u003c/p\u003e \u003cp\u003eBuilding on these findings, future research should focus on conducting in vitro and in vivo experiments to validate the stability, binding efficiency, and therapeutic efficacy of the humanized nanobody. Furthermore, investigations into thermal stability, aggregation propensity, and solubility under diverse environmental and physiological conditions, such as varying pH or ionic strength, will provide a clearer understanding of its clinical potential. Developing pharmacokinetic and immunogenicity studies in animal models will further assess its compatibility and performance in therapeutic settings.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eCompeting interests\u003c/h2\u003e \u003cp\u003eThe authors declare that they have no known competing financial interests or personal\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAli Rahmati Bonab and Hannaneh Jalilzadeh Ghahi contributed equally to this work. Both were involved in writing, review, and editing of the manuscript, original draft preparation, visualization, and validation of results. Mahmoud Hasani, Vahid Jajarmi, and Javad Ranjbari provided critical feedback, supervised the study, and contributed to data analysis and interpretation. All authors participated in conceptualization, approved the final manuscript, and agreed to be accountable for the work.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eWe acknowledge the use of OpenAI's ChatGPT for assistance in drafting and refining sections of this manuscript. ChatGPT was used to generate suggestions for text improvement and content clarity.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eVincke, C., et al., \u003cem\u003eGeneral strategy to humanize a camelid single-domain antibody and identification of a universal humanized nanobody scaffold.\u003c/em\u003e J Biol Chem, 2009. \u003cstrong\u003e284\u003c/strong\u003e(5): p. 3273-3284.\u003c/li\u003e\n\u003cli\u003eMuyldermans, S., \u003cem\u003eNanobodies: natural single-domain antibodies.\u003c/em\u003e Annual review of biochemistry, 2013. \u003cstrong\u003e82\u003c/strong\u003e: p. 775-797.\u003c/li\u003e\n\u003cli\u003eScully, M., et al., \u003cem\u003eCaplacizumab Treatment for Acquired Thrombotic Thrombocytopenic Purpura.\u003c/em\u003e N Engl J Med, 2019. \u003cstrong\u003e380\u003c/strong\u003e(4): p. 335-346.\u003c/li\u003e\n\u003cli\u003ePanikar, S.S., et al., \u003cem\u003eNanobodies as efficient drug-carriers: Progress and trends in chemotherapy.\u003c/em\u003e Journal of Controlled Release, 2021. \u003cstrong\u003e334\u003c/strong\u003e: p. 389-412.\u003c/li\u003e\n\u003cli\u003eYang, J., et al., \u003cem\u003eDevelopment of a bispecific nanobody conjugate broadly neutralizes diverse SARS-CoV-2 variants and structural basis for its broad neutralization.\u003c/em\u003e PLOS Pathogens, 2023. \u003cstrong\u003e19\u003c/strong\u003e(11): p. e1011804.\u003c/li\u003e\n\u003cli\u003eOmidfar, K., et al., \u003cem\u003eEfficient growth inhibition of EGFR over-expressing tumor cells by an anti-EGFR nanobody.\u003c/em\u003e Mol Biol Rep, 2013. \u003cstrong\u003e40\u003c/strong\u003e(12): p. 6737-45.\u003c/li\u003e\n\u003cli\u003eAbdolvahab, M.H., et al., \u003cem\u003eTargeted drug delivery using nanobodies to deliver effective molecules to breast cancer cells: the most attractive application of nanobodies.\u003c/em\u003e Cancer Cell Int, 2024. \u003cstrong\u003e24\u003c/strong\u003e(1): p. 67.\u003c/li\u003e\n\u003cli\u003eRangarajan, S., et al., \u003cem\u003ePeptide-MHC (pMHC) binding to a human antiviral T cell receptor induces long-range allosteric communication between pMHC- and CD3-binding sites.\u003c/em\u003e J Biol Chem, 2018. \u003cstrong\u003e293\u003c/strong\u003e(41): p. 15991-16005.\u003c/li\u003e\n\u003cli\u003eHe, Y., et al., \u003cem\u003ePeptide-MHC Binding Reveals Conserved Allosteric Sites in MHC Class I- and Class II-Restricted T Cell Receptors (TCRs).\u003c/em\u003e J Mol Biol, 2020. \u003cstrong\u003e432\u003c/strong\u003e(24): p. 166697.\u003c/li\u003e\n\u003cli\u003eAlba, J. and M. D\u0026apos;Abramo, \u003cem\u003eThe Full Model of the pMHC-TCR-CD3 Complex: A Structural and Dynamical Characterization of Bound and Unbound States.\u003c/em\u003e Cells, 2022. \u003cstrong\u003e11\u003c/strong\u003e(4).\u003c/li\u003e\n\u003cli\u003eMoradi-Kalbolandi, S., et al., \u003cem\u003eEvaluation the potential of recombinant anti-CD3 nanobody on immunomodulatory function.\u003c/em\u003e Mol Immunol, 2020. \u003cstrong\u003e118\u003c/strong\u003e: p. 174-181.\u003c/li\u003e\n\u003cli\u003eKhatibi, A.S., et al., \u003cem\u003eIn vivo tumor-suppressing and anti-angiogenic activities of a recombinant anti-CD3\u0026epsilon; nanobody in breast cancer mice model.\u003c/em\u003e Immunotherapy, 2019. \u003cstrong\u003e11\u003c/strong\u003e(18): p. 1555-1567.\u003c/li\u003e\n\u003cli\u003eGordon, G.L., et al., \u003cem\u003eProspects for the computational humanization of antibodies and nanobodies.\u003c/em\u003e Front Immunol, 2024. \u003cstrong\u003e15\u003c/strong\u003e: p. 1399438.\u003c/li\u003e\n\u003cli\u003eChinery, L., J.R. Jeliazkov, and C.M. Deane, \u003cem\u003eHumatch - fast, gene-specific joint humanisation of antibody heavy and light chains.\u003c/em\u003e MAbs, 2024. \u003cstrong\u003e16\u003c/strong\u003e(1): p. 2434121.\u003c/li\u003e\n\u003cli\u003eChoi, Y., et al., \u003cem\u003eComputationally driven antibody engineering enables simultaneous humanization and thermostabilization.\u003c/em\u003e Protein Eng Des Sel, 2016. \u003cstrong\u003e29\u003c/strong\u003e(10): p. 419-426.\u003c/li\u003e\n\u003cli\u003eZambrano, R., et al., \u003cem\u003eAGGRESCAN3D (A3D): server for prediction of aggregation properties of protein structures.\u003c/em\u003e Nucleic Acids Res, 2015. \u003cstrong\u003e43\u003c/strong\u003e(W1): p. W306-13.\u003c/li\u003e\n\u003cli\u003eJung, F., et al., \u003cem\u003eDeepSTABp: A Deep Learning Approach for the Prediction of Thermal Protein Stability.\u003c/em\u003e Int J Mol Sci, 2023. \u003cstrong\u003e24\u003c/strong\u003e(8).\u003c/li\u003e\n\u003cli\u003eKuriata, A., et al., \u003cem\u003eAggrescan3D (A3D) 2.0: prediction and engineering of protein solubility.\u003c/em\u003e Nucleic Acids Research, 2019. \u003cstrong\u003e47\u003c/strong\u003e(W1): p. W300-W307.\u003c/li\u003e\n\u003cli\u003eCohen, T., M. Halfon, and D. Schneidman-Duhovny, \u003cem\u003eNanoNet: Rapid and accurate end-to-end nanobody modeling by deep learning.\u003c/em\u003e Front Immunol, 2022. \u003cstrong\u003e13\u003c/strong\u003e: p. 958584.\u003c/li\u003e\n\u003cli\u003eJumper, J., et al., \u003cem\u003eHighly accurate protein structure prediction with AlphaFold.\u003c/em\u003e Nature, 2021. \u003cstrong\u003e596\u003c/strong\u003e(7873): p. 583-589.\u003c/li\u003e\n\u003cli\u003eLin, Z., et al., \u003cem\u003eEvolutionary-scale prediction of atomic-level protein structure with a language model.\u003c/em\u003e Science, 2023. \u003cstrong\u003e379\u003c/strong\u003e(6637): p. 1123-1130.\u003c/li\u003e\n\u003cli\u003eDu, Z., et al., \u003cem\u003eThe trRosetta server for fast and accurate protein structure prediction.\u003c/em\u003e Nature Protocols, 2021. \u003cstrong\u003e16\u003c/strong\u003e(12): p. 5634-5651.\u003c/li\u003e\n\u003cli\u003eVald\u0026eacute;s-Tresanco, M.S., et al., \u003cem\u003eStructural Modeling of Nanobodies: A Benchmark of State-of-the-Art Artificial Intelligence Programs.\u003c/em\u003e Molecules, 2023. \u003cstrong\u003e28\u003c/strong\u003e(10).\u003c/li\u003e\n\u003cli\u003eSang, Z., et al., \u003cem\u003eLlamanade: An open-source computational pipeline for robust nanobody humanization.\u003c/em\u003e Structure, 2022. \u003cstrong\u003e30\u003c/strong\u003e(3): p. 418-429.e3.\u003c/li\u003e\n\u003cli\u003eSangzhe, \u003cem\u003eLlamanade: A nanobody humanization tool\u003c/em\u003e. 2025.\u003c/li\u003e\n\u003cli\u003ePrihoda, D., et al., \u003cem\u003eBioPhi: A platform for antibody design, humanization, and humanness evaluation based on natural antibody repertoires and deep learning.\u003c/em\u003e MAbs, 2022. \u003cstrong\u003e14\u003c/strong\u003e(1): p. 2020203.\u003c/li\u003e\n\u003cli\u003eKufer, P. and T. Raum, \u003cem\u003eCross-species-specific single domain bispecific single chain antibody\u003c/em\u003e. 2010, World Intellectual Property Organization (WIPO).\u003c/li\u003e\n\u003cli\u003eDina, L., \u003cem\u003eNanoNet: A computational tool for nanobody prediction and analysis\u003c/em\u003e. 2024, GitHub repository.\u003c/li\u003e\n\u003cli\u003eSchr\u0026ouml;dinger, L.L.C., \u003cem\u003ePyMOL Molecular Graphics System\u003c/em\u003e. 2020, Schr\u0026ouml;dinger: New York.\u003c/li\u003e\n\u003cli\u003eSchr\u0026ouml;dinger, L.L.C., \u003cem\u003eBioLuminate\u003c/em\u003e. 2021, Schr\u0026ouml;dinger: New York.\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eRamPlot: A Webserver for Ramachandran Plot Analysis\u003c/em\u003e. Available from: https://ramplot.in/.\u003c/li\u003e\n\u003cli\u003eSobolev, O.V., et al., \u003cem\u003eA Global Ramachandran Score Identifies Protein Structures with Unlikely Stereochemistry.\u003c/em\u003e Structure, 2020. \u003cstrong\u003e28\u003c/strong\u003e(11): p. 1249-1258.e2.\u003c/li\u003e\n\u003cli\u003eZhang, L. \u003cem\u003eTM-align: A Protein Structure Alignment Tool\u003c/em\u003e. Available from: https://zhanggroup.org/TM-align/.\u003c/li\u003e\n\u003cli\u003eZhang, Y. and J. Skolnick, \u003cem\u003eTM-align: a protein structure alignment algorithm based on the TM-score.\u003c/em\u003e Nucleic Acids Res, 2005. \u003cstrong\u003e33\u003c/strong\u003e(7): p. 2302-9.\u003c/li\u003e\n\u003cli\u003eviewer, R.P., \u003cem\u003eRCSB Pdb Viewer\u003c/em\u003e. 2024.\u003c/li\u003e\n\u003cli\u003eBaker, L., \u003cem\u003eRoseTTAFold-All-Atom: High-resolution protein structure modeling\u003c/em\u003e. 2024, GitHub repository.\u003c/li\u003e\n\u003cli\u003eAnbo, H., et al., \u003cem\u003eHow AlphaFold2 Predicts Conditionally Folding Regions Annotated in an Intrinsically Disordered Protein Database, IDEAL.\u003c/em\u003e Biology (Basel), 2023. \u003cstrong\u003e12\u003c/strong\u003e(2).\u003c/li\u003e\n\u003cli\u003eNeuroSnap, \u003cem\u003eAMBER Relaxation Service\u003c/em\u003e. 2024, NeuroSnap AI.\u003c/li\u003e\n\u003cli\u003eRcsb, P.D.B., \u003cem\u003eStructure 1XIW: Crystal structure of a protein\u003c/em\u003e. RCSB Protein Data Bank.\u003c/li\u003e\n\u003cli\u003eOxford Protein Informatics, G., \u003cem\u003eAntibody Sequence Analysis using ANARCI\u003c/em\u003e. 2024.\u003c/li\u003e\n\u003cli\u003eDunbar, J. and C.M. Deane, \u003cem\u003eANARCI: antigen receptor numbering and receptor classification.\u003c/em\u003e Bioinformatics, 2016. \u003cstrong\u003e32\u003c/strong\u003e(2): p. 298-300.\u003c/li\u003e\n\u003cli\u003eDondelinger, M., et al., \u003cem\u003eUnderstanding the Significance and Implications of Antibody Numbering and Antigen-Binding Surface/Residue Definition.\u003c/em\u003e Frontiers in Immunology, 2018. \u003cstrong\u003e9\u003c/strong\u003e.\u003c/li\u003e\n\u003cli\u003eWilliams, D.G., D.J. Matthews, and T. Jones, \u003cem\u003eHumanising Antibodies by CDR Grafting\u003c/em\u003e, in \u003cem\u003eAntibody Engineering\u003c/em\u003e, R. Kontermann and S. D\u0026uuml;bel, Editors. 2010, Springer Berlin Heidelberg: Berlin, Heidelberg. p. 319-339.\u003c/li\u003e\n\u003cli\u003eMakabe, K., et al., \u003cem\u003eThermodynamic consequences of mutations in vernier zone residues of a humanized anti-human epidermal growth factor receptor murine antibody, 528.\u003c/em\u003e J Biol Chem, 2008. \u003cstrong\u003e283\u003c/strong\u003e(2): p. 1156-66.\u003c/li\u003e\n\u003cli\u003eAlmagro, J.C. and J. Fransson, \u003cem\u003eHumanization of antibodies.\u003c/em\u003e Front Biosci, 2008. \u003cstrong\u003e13\u003c/strong\u003e: p. 1619-33.\u003c/li\u003e\n\u003cli\u003eKong, R., et al., \u003cem\u003eCoDockPP: A Multistage Approach for Global and Site-Specific Protein\u0026ndash;Protein Docking.\u003c/em\u003e Journal of Chemical Information and Modeling, 2019. \u003cstrong\u003e59\u003c/strong\u003e(8): p. 3556-3564.\u003c/li\u003e\n\u003cli\u003eJim\u0026eacute;nez-Garc\u0026iacute;a, B., C. Pons, and J. Fern\u0026aacute;ndez-Recio, \u003cem\u003epyDockWEB: a web server for rigid-body protein-protein docking using electrostatics and desolvation scoring.\u003c/em\u003e Bioinformatics, 2013. \u003cstrong\u003e29\u003c/strong\u003e(13): p. 1698-9.\u003c/li\u003e\n\u003cli\u003eCheng, T.M., T.L. Blundell, and J. Fernandez-Recio, \u003cem\u003epyDock: electrostatics and desolvation for effective scoring of rigid-body protein-protein docking.\u003c/em\u003e Proteins, 2007. \u003cstrong\u003e68\u003c/strong\u003e(2): p. 503-15.\u003c/li\u003e\n\u003cli\u003eLaskowski, R.A., \u003cem\u003ePDBsum: summaries and analyses of PDB structures.\u003c/em\u003e Nucleic Acids Res, 2001. \u003cstrong\u003e29\u003c/strong\u003e(1): p. 221-2.\u003c/li\u003e\n\u003cli\u003eThe ConSurf Project, T. \u003cem\u003eConSurf: A Webserver for Identifying Functional Regions in Proteins by Mapping Evolutionary Conservation\u003c/em\u003e. Available from: https://consurf.tau.ac.il/consurf_index.php.\u003c/li\u003e\n\u003cli\u003eTaboga, M., \u003cem\u003eBayesian inference \u003cem\u003eLectures on probability theory and mathematical statistics\u003c/em\u003e. 2021: Kindle Direct Publishing.\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eAshkenazy, H., et al., \u003cem\u003eConSurf 2016: an improved methodology to estimate and visualize evolutionary conservation in macromolecules.\u003c/em\u003e Nucleic Acids Res, 2016. \u003cstrong\u003e44\u003c/strong\u003e(W1): p. W344-50.\u003c/li\u003e\n\u003cli\u003ePlayMolecule Project, T. \u003cem\u003ePlayMolecule System Builder: A Webserver for Molecular System Preparation and pKa Prediction\u003c/em\u003e. Available from: https://open.playmolecule.org/tools/systembuilder.\u003c/li\u003e\n\u003cli\u003eBurger, S.K. and P.W. Ayers, \u003cem\u003eEmpirical prediction of protein pKa values with residue mutation.\u003c/em\u003e J Comput Chem, 2011. \u003cstrong\u003e32\u003c/strong\u003e(10): p. 2140-8.\u003c/li\u003e\n\u003cli\u003eHebditch, M., et al., \u003cem\u003eProtein-Sol: a web tool for predicting protein solubility from sequence.\u003c/em\u003e Bioinformatics, 2017. \u003cstrong\u003e33\u003c/strong\u003e(19): p. 3098-3100.\u003c/li\u003e\n\u003cli\u003eEryk, \u003cem\u003eNetSurfP-3.0: Predicting the structural properties of proteins.\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eH\u0026oslash;ie, M.H., et al., \u003cem\u003eNetSurfP-3.0: accurate and fast prediction of protein structural features by protein language models and deep learning.\u003c/em\u003e Nucleic Acids Research, 2022. \u003cstrong\u003e50\u003c/strong\u003e(W1): p. W510-W515.\u003c/li\u003e\n\u003cli\u003eBhandari, B.K., C.S. Lim, and P.P. Gardner, \u003cem\u003eTISIGNER.com: web services for improving recombinant protein production.\u003c/em\u003e Nucleic Acids Research, 2021. \u003cstrong\u003e49\u003c/strong\u003e(W1): p. W654-W661.\u003c/li\u003e\n\u003cli\u003eBhandari, B.K., P.P. Gardner, and C.S. Lim, \u003cem\u003eSolubility-Weighted Index: fast and accurate prediction of protein solubility.\u003c/em\u003e Bioinformatics, 2020. \u003cstrong\u003e36\u003c/strong\u003e(18): p. 4691-4698.\u003c/li\u003e\n\u003cli\u003eKyte, J. and R.F. Doolittle, \u003cem\u003eA simple method for displaying the hydropathic character of a protein.\u003c/em\u003e J Mol Biol, 1982. \u003cstrong\u003e157\u003c/strong\u003e(1): p. 105-32.\u003c/li\u003e\n\u003cli\u003eTurzo, S., et al., \u003cem\u003ePredicting ion mobility collision cross sections using projection approximation with ROSIE-PARCS webserver.\u003c/em\u003e Brief Bioinform, 2023. \u003cstrong\u003e24\u003c/strong\u003e(5).\u003c/li\u003e\n\u003cli\u003eTurzo, S., et al., \u003cem\u003eProtein shape sampled by ion mobility mass spectrometry consistently improves protein structure prediction.\u003c/em\u003e Nat Commun, 2022. \u003cstrong\u003e13\u003c/strong\u003e(1): p. 4377.\u003c/li\u003e\n\u003cli\u003eWilliams, J.P., et al., \u003cem\u003eUse of ion mobility mass spectrometry and a collision cross-section algorithm to study an organometallic ruthenium anticancer complex and its adducts with a DNA oligonucleotide.\u003c/em\u003e Rapid Commun Mass Spectrom, 2009. \u003cstrong\u003e23\u003c/strong\u003e(22): p. 3563-9.\u003c/li\u003e\n\u003cli\u003eCovey, T. and D.J. Douglas, \u003cem\u003eCollision cross sections for protein ions.\u003c/em\u003e J Am Soc Mass Spectrom, 1993. \u003cstrong\u003e4\u003c/strong\u003e(8): p. 616-23.\u003c/li\u003e\n\u003cli\u003eNeurosnap.ai, \u003cem\u003eToxinPred: Peptide Toxicity Prediction Service.\u003c/em\u003e\u003c/li\u003e\n\u003cli\u003eRathore, A.S., et al., \u003cem\u003eToxinPred 3.0: An improved method for predicting the toxicity of peptides.\u003c/em\u003e Computers in Biology and Medicine, 2024. \u003cstrong\u003e179\u003c/strong\u003e: p. 108926.\u003c/li\u003e\n\u003cli\u003eCervantes, J., et al., \u003cem\u003eA comprehensive survey on support vector machine classification: Applications, challenges and trends.\u003c/em\u003e Neurocomputing, 2020. \u003cstrong\u003e408\u003c/strong\u003e: p. 189-215.\u003c/li\u003e\n\u003cli\u003eSaha, S. and G.P. Raghava, \u003cem\u003eAlgPred: prediction of allergenic proteins and mapping of IgE epitopes.\u003c/em\u003e Nucleic Acids Res, 2006. \u003cstrong\u003e34\u003c/strong\u003e(Web Server issue): p. W202-9.\u003c/li\u003e\n\u003cli\u003eParrinello, M. and A. Rahman, \u003cem\u003ePolymorphic transitions in single crystals: A new molecular dynamics method.\u003c/em\u003e Journal of Applied Physics, 1981. \u003cstrong\u003e52\u003c/strong\u003e(12): p. 7182-7190.\u003c/li\u003e\n\u003cli\u003eKazemi-Lomedasht, F., et al., \u003cem\u003eDesign of a humanized anti vascular endothelial growth factor nanobody and evaluation of its in vitro function.\u003c/em\u003e Iran J Basic Med Sci, 2018. \u003cstrong\u003e21\u003c/strong\u003e(3): p. 260-266.\u003c/li\u003e\n\u003cli\u003eXiang, J., et al., \u003cem\u003eFramework residues 71 and 93 of the chimeric B72.3 antibody are major determinants of the conformation of heavy-chain hypervariable loops.\u003c/em\u003e J Mol Biol, 1995. \u003cstrong\u003e253\u003c/strong\u003e(3): p. 385-90.\u003c/li\u003e\n\u003cli\u003eWu, H., et al., \u003cem\u003eHumanization of a murine monoclonal antibody by simultaneous optimization of framework and CDR residues.\u003c/em\u003e J Mol Biol, 1999. \u003cstrong\u003e294\u003c/strong\u003e(1): p. 151-62.\u003c/li\u003e\n\u003cli\u003eFern\u0026aacute;ndez-Quintero, M.L., M.C. Heiss, and K.R. Liedl, \u003cem\u003eAntibody humanization-the Influence of the antibody framework on the CDR-H3 loop ensemble in solution.\u003c/em\u003e Protein Eng Des Sel, 2019. \u003cstrong\u003e32\u003c/strong\u003e(9): p. 411-422.\u003c/li\u003e\n\u003cli\u003eFoote, J. and G. Winter, \u003cem\u003eAntibody framework residues affecting the conformation of the hypervariable loops.\u003c/em\u003e J Mol Biol, 1992. \u003cstrong\u003e224\u003c/strong\u003e(2): p. 487-99.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u0026nbsp;Table 1. Amino acid sequences of the anti-CD3\u0026epsilon; nanobody and CD3\u0026epsilon; extracellular region. The single-letter codes represent the respective protein components\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" align=\"\n \" width=\"696\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 14.4907%;\"\u003e\n \u003cp\u003eProtein Component\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85.5093%;\"\u003e\n \u003cp\u003eAmino Acid Sequence (Single Letter Code)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 14.4907%;\"\u003e\n \u003cp\u003eAnti-CD3\u0026epsilon; Nanobody\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85.5093%;\"\u003e\n \u003cp\u003eEVQLVESGGGLVQPGGSLRLSCAASGFTFDDYGMSWVRQAPGKWLEWVSDISWNGGSTYYADSVKGRFTISRDNAENTLYLQMNSLKPDDT\u003c/p\u003e\n \u003cp\u003eAVYYCAKMGEGGWGANDYWGQGTQVTVSS\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 14.4907%;\"\u003e\n \u003cp\u003eCD3\u0026epsilon; (Extracellular)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85.5093%;\"\u003e\n \u003cp\u003eQDGNEEMGGITQTPYKVSISGTTVILTCPQYPGSEILWQHNDKNIGGDEDDKNIGSDEDHLSLKEFSELEQSGYYVCYPRGSKPEDANFYLYLRA\u003c/p\u003e\n \u003cp\u003eRVCENCMEMD\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 2.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003edetailing the amino acid sequence regions of the wild type nanobody structure. The framework regions (FWRs) and complementarity-determining regions (CDRs) are specified with their residue ranges and corresponding sequences.(Aminoacides depicted is single code format)\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"660\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.0606%;\"\u003e\n \u003cp\u003eRegion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 43.4848%;\"\u003e\n \u003cp\u003e\u0026nbsp;Residue Range with Aho numbering Positions\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 45.4545%;\"\u003e\n \u003cp\u003eSequence\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.0606%;\"\u003e\n \u003cp\u003eFWR1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 43.4848%;\"\u003e\n \u003cp\u003e1\u0026ndash;23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 45.4545%;\"\u003e\n \u003cp\u003eEVQLVESGGGLVQPGGSLRLSC\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.0606%;\"\u003e\n \u003cp\u003eCDR1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 43.4848%;\"\u003e\n \u003cp\u003e24\u0026ndash;42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 45.4545%;\"\u003e\n \u003cp\u003eAASGFTFDDYGMS\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.0606%;\"\u003e\n \u003cp\u003eFWR2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 43.4848%;\"\u003e\n \u003cp\u003e43\u0026ndash;56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 45.4545%;\"\u003e\n \u003cp\u003eWVRQAPGKWLEWVS\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.0606%;\"\u003e\n \u003cp\u003eCDR2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 43.4848%;\"\u003e\n \u003cp\u003e57\u0026ndash;68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 45.4545%;\"\u003e\n \u003cp\u003eDISWNGGSTY\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.0606%;\"\u003e\n \u003cp\u003eFWR3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 43.4848%;\"\u003e\n \u003cp\u003e69\u0026ndash;106\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 45.4545%;\"\u003e\n \u003cp\u003eYADSVKGRFTISRDNAVENTLYLQMNSLKPDDTAVYYC\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.0606%;\"\u003e\n \u003cp\u003eCDR3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 43.4848%;\"\u003e\n \u003cp\u003e107\u0026ndash;138\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 45.4545%;\"\u003e\n \u003cp\u003eAKMGEGGGWGANDY\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 11.0606%;\"\u003e\n \u003cp\u003eFWR4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 43.4848%;\"\u003e\n \u003cp\u003e139\u0026ndash;149\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 45.4545%;\"\u003e\n \u003cp\u003eWGQGTQVTVSS\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 3. Summarizing the amino acid substitutions introduced by Llamanade during the humanization of nanobody framework regions (FWRs).\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"612\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eFWR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003ePosition\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eOriginal Amino Acid\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003eModified Amino Acid\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eFWR 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eNo changes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003eNo changes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eFWR 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eTryptophan (W)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003eGlycine (G)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eFWR 3\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eGlutamic Acid (E)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003eSerine (S)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eLysine (K)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003eArginine (R)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eProline (P)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003eAlanine (A)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eAspartic Acid (D)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003eGlutamic Acid (E)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 204px;\"\u003e\n \u003cp\u003eFWR 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e105\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 156px;\"\u003e\n \u003cp\u003eGlutamine (Q)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003eLeucine (L)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 4.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eFrequency shifts of amino acid residue substitutions in specific framework (FR) regions of a protein sequence. Substitutions are shown using the single-letter amino acid code in the format original residue \u0026rarr; new residue, accompanied by their respective frequency changes (%).mutations are recommended by Biophi.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"642\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003eRegion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 116px;\"\u003e\n \u003cp\u003eposition\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 127px;\"\u003e\n \u003cp\u003eAho numbering position\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 135px;\"\u003e\n \u003cp\u003eSequence\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eResidue Frequency (human)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003eFR2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 116px;\"\u003e\n \u003cp\u003e44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 127px;\"\u003e\n \u003cp\u003eH51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 135px;\"\u003e\n \u003cp\u003eW \u0026rarr; G\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003e0.3% \u0026rarr; 99%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"5\" valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eFR3\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 116px;\"\u003e\n \u003cp\u003e75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 127px;\"\u003e\n \u003cp\u003eH85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 135px;\"\u003e\n \u003cp\u003eA \u0026rarr; S\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003e38% \u0026rarr; 59%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 116px;\"\u003e\n \u003cp\u003e76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 127px;\"\u003e\n \u003cp\u003eH86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 135px;\"\u003e\n \u003cp\u003eE \u0026rarr; K\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003e2% \u0026rarr; 92%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 116px;\"\u003e\n \u003cp\u003e87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 127px;\"\u003e\n \u003cp\u003eH97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 135px;\"\u003e\n \u003cp\u003eK \u0026rarr; R\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003e9% \u0026rarr; 86%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 116px;\"\u003e\n \u003cp\u003e88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 127px;\"\u003e\n \u003cp\u003eH98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 135px;\"\u003e\n \u003cp\u003eP \u0026rarr; A\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003e3% \u0026rarr; 75%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 116px;\"\u003e\n \u003cp\u003e89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 127px;\"\u003e\n \u003cp\u003eH99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 135px;\"\u003e\n \u003cp\u003eD \u0026rarr; E\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003e5% \u0026rarr; 93%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003eFR4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 116px;\"\u003e\n \u003cp\u003e115\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 127px;\"\u003e\n \u003cp\u003eH144\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 135px;\"\u003e\n \u003cp\u003eQ \u0026rarr; L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003e0.8% \u0026rarr; 68%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 5. Impact of Specific Framework Mutations on Binding Affinity. Each mutation is presented with its corresponding effect on affinity (\u0026Delta;Affinity) measured as the change in binding energy upon substitution. Positive values indicate improved binding, while negative values suggest a slight reduction. Notably, the mutation H:89 (ASH \u0026rarr; GLU) exhibits the highest positive effect (+0.6617), significantly enhancing antigen interaction, whereas mutations such as H:115 (GLN \u0026rarr; LEU) show minimal negative impact (-0.0222), indicating overall preserved binding performance.\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"635\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 360px;\"\u003e\n \u003cp\u003eMutation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 275px;\"\u003e\n \u003cp\u003e\u0026Delta; Affinity (kJ/mol)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 360px;\"\u003e\n \u003cp\u003eH:44 (TRP\u0026nbsp;\u0026agrave;\u0026nbsp;GLY)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 275px;\"\u003e\n \u003cp\u003e-0.1441\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 360px;\"\u003e\n \u003cp\u003eH:76 (GLU\u0026nbsp;\u0026agrave;\u0026nbsp;ARG)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 275px;\"\u003e\n \u003cp\u003e0.4925\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 360px;\"\u003e\n \u003cp\u003eH:87 (LYS\u0026nbsp;\u0026agrave;\u0026nbsp;ARG)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 275px;\"\u003e\n \u003cp\u003e-0.0156\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 360px;\"\u003e\n \u003cp\u003eH:88 (PRO\u0026nbsp;\u0026agrave;\u0026nbsp;ALA)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 275px;\"\u003e\n \u003cp\u003e0.0073\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 360px;\"\u003e\n \u003cp\u003eH:89 (ASH\u0026nbsp;\u0026agrave;\u0026nbsp;GLU)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 275px;\"\u003e\n \u003cp\u003e0.6617\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 360px;\"\u003e\n \u003cp\u003eH:115 (GLN\u0026nbsp;\u0026agrave;\u0026nbsp;LEU)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 275px;\"\u003e\n \u003cp\u003e-0.0222\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 6. Ranking of docking conformations for wild-type nanobody-CD3\u0026epsilon; interactions based on binding energy components. The table summarizes key parameters, including electrostatics, desolvation, van der Waals (VdW) forces, and relative restraint energy (relIRST), contributing to the total energy score for each docking conformation. Lower total energy values indicate more favorable binding poses. The top-ranked conformation (#1) exhibits the lowest total energy of -100.335, suggesting the most stable binding interaction (units: kJ/mol).\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"641\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003eRank\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003eElectrostatics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003eDesolvation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003eVdW\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003erelIRST\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003eTotal\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003e-3.514\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003e-53.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e50.571\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e48.148\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e-100.335\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003e-1.82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003e-56.152\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e89.015\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e48.148\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e-97.219\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003e-11.346\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003e-30.378\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e23.102\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e51.852\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e-91.266\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003e-2.606\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003e-56.409\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e167.936\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e48.148\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e-90.369\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003e-7.181\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003e-38.622\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e76.498\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e48.148\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e-86.301\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003e-1.701\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003e-37.367\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e-16.937\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e48.408\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e-85.206\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003e-3.023\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003e-33.609\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e2.577\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e44.444\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e-80.818\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003e-0.412\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003e-29.265\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e45.805\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e55.556\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e-80.653\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003e-2.109\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003e-38.229\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e54.385\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e44.444\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e-79.344\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 78px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 164px;\"\u003e\n \u003cp\u003e-4.254\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 125px;\"\u003e\n \u003cp\u003e-39.853\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\n \u003cp\u003e70.826\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003e40.741\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 103px;\"\u003e\n \u003cp\u003e-77.765\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 7. Ranking of docking conformations of the humanized nanobody based on binding interaction energies. The table summarizes key interaction parameters, including electrostatics, desolvation, van der Waals (VdW) forces, and relative restraint energy (relRST), contributing to the total energy score for each docking conformation. Rankings are based on the total energy (lower values indicate better binding affinity and units are based on kJ/mol).\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"650\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003eRank\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003eElectrostatics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003eDesolvation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003eVdW\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003erelRST\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003eTotal\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003e-2.542\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e-40.709\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003e21.428\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e55.556\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e-96.664\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003e-9.652\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e-32.039\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003e25.908\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e51.852\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e-90.952\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003e-2.841\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e-32.532\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003e37.257\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e55.556\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e-87.203\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003e-12.488\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e-2.541\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003e-2.443\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e70.370\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e-85.643\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003e-3.780\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e-28.727\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003e33.879\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e55.556\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e-84.675\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003e-4.418\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e-48.209\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003e124.144\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e44.444\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e-84.657\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003e-13.591\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e-8.354\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003e43.740\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e66.667\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e-84.238\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003e-6.096\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e-23.035\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003e51.885\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e59.259\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e-83.201\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003e-11.985\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e-6.719\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003e67.806\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e70.370\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e-82.293\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 98px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 147px;\"\u003e\n \u003cp\u003e-0.440\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e-29.814\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 101px;\"\u003e\n \u003cp\u003e40.144\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e55.556\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 100px;\"\u003e\n \u003cp\u003e-81.796\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\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":"Anti-CD3ε nanobody, Humanization, immunogenicity, Structural modeling, nanobody engineering","lastPublishedDoi":"10.21203/rs.3.rs-5769566/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5769566/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study presents a comprehensive machine-learning-driven approach for the in silico humanization and characterization of anti-CD3ε nanobodies. Nanobodies, single-domain antibodies derived from camelids, hold immense therapeutic potential due to their small size, high solubility, and exceptional stability. However, their camelid origin necessitates humanization to minimize immunogenicity in therapeutic applications. Using state-of-the-art computational tools such as NanoNet, RoseTTAFold, and PyDock, we modeled and analyzed both wild type and humanized anti-CD3ε nanobody variants. Key metrics, including structural stability, binding efficiency, thermal stability, and aggregation propensity, were evaluated. Humanization achieved enhanced humanness scores, increased thermal stability, and retained strong binding interactions with CD3ε while preserving the nanobody\u0026rsquo;s structural integrity. Molecular dynamics simulations confirmed minimal deviations in structural flexibility and binding-site compatibility post-humanization. These findings support the efficacy of computational methods in optimizing nanobody therapeutics for clinical applications, paving the way for advanced immunotherapy strategies targeting immune-related disorders. The results demonstrate that the humanized anti-CD3ε nanobody exhibits enhanced thermal stability, reduced aggregation propensity, improved humanness scores, and comparable binding efficiency to the wild type nanobody, making it a promising therapeutic candidate.\u003c/p\u003e","manuscriptTitle":"Computational and Machine Learning Approaches for Optimizing Anti-CD3ε Nanobody: Humanization and Characterization for Enhanced Therapeutic Efficacy","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-13 04:05:48","doi":"10.21203/rs.3.rs-5769566/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":"615a2f4b-a362-491e-bb34-5d6b231bbf5c","owner":[],"postedDate":"January 13th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":42612489,"name":"Biological sciences/Biotechnology"},{"id":42612490,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":42612491,"name":"Biological sciences/Drug discovery"},{"id":42612492,"name":"Biological sciences/Immunology"},{"id":42612493,"name":"Biological sciences/Molecular biology"},{"id":42612494,"name":"Biological sciences/Structural biology"}],"tags":[],"updatedAt":"2025-05-23T12:53:09+00:00","versionOfRecord":[],"versionCreatedAt":"2025-01-13 04:05:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5769566","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5769566","identity":"rs-5769566","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}