Abstract Triple-negative breast cancer (TNBC) remains a major therapeutic challenge because of its aggressive behavior, molecular heterogeneity, and limited subtype-specific targets. This study used an integrated immunoinformatics workflow to design a recombinant tumor-associated antigen-derived multi-epitope vaccine based on MMP1, CXorf61/CT83, and COL11A1, prioritized according to their tumor-to-normal transcript-expression ratios. Candidate cytotoxic T-lymphocyte, helper T-lymphocyte, and linear B-cell epitopes were screened for predicted HLA binding, antigenicity, allergenicity, toxicity, and IFN-γ-induction potential. The selected epitopes were assembled with PADRE, a MyD88-derived exploratory immunomodulatory domain, class-specific linkers, and a C-terminal histidine tag to generate a 621-amino-acid construct. Combined HLA class I and II analysis predicted 99.36% worldwide population coverage, with 3.14 epitope–HLA hits per individual. The construct was predicted to be antigenic, non-allergenic, soluble, and physicochemically compatible with recombinant production. Structural modeling yielded a ProSA Z-score of −7.43 and 97.45% of residues in favored Ramachandran regions, while disulfide engineering identified ten candidate intramolecular bridges. Docking produced an HDOCK score of −298.65 for the modeled vaccine–TLR4 complex, and flexibility analysis showed an average RMSF of 1.93 Å. Immune simulation predicted Th1-associated cytokine production, T-cell expansion, antibody responses, and memory-cell formation. Codon optimization generated a CAI of 0.95 and a GC content of 53.2%, followed by virtual cloning into pET-28a(+). These findings support the computational feasibility of the proposed vaccine candidate, which requires experimental validation of expression, antigen processing, HLA presentation, immunogenicity, safety, and antitumor activity.

Keywords: Immunoinformatics, Recombinant multi-epitope vaccine, MyD88-derived immunomodulatory domain, Tumor-associated antigens, Triple-negative breast cancer

Citation:  Afnani, M.R., Setiawan, V.K., and Rovik, A. 2026. Expression-guided immunoinformatics design of a recombinant multi-epitope vaccine candidate for triple-negative breast cancer incorporating a MyD88-derived immunomodulatory domain. Natural and Life Sciences Communications. 25(4): e2026091.

Graphical Abstract:

INTRODUCTION

Breast cancer remains a significant global health challenge, with its incidence projected to reach approximately 3.2 million new cases annually by 2050. In this landscape, triple-negative breast cancer (TNBC) stands out as a particularly aggressive subtype, disproportionately contributing to high mortality rates among younger patients and specific ethnic groups (Xiong et al., 2024; Bouzid et al., 2025). Defined by the absence of expression of the estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), TNBC lacks the molecular targets that have made endocrine and HER2-directed therapies successful in other breast cancer subtypes (Rovik et al., 2024; Xiong et al., 2024; Yulian and Fachriza, 2024).

In the Indonesian clinical context, TNBC represents a substantial disease burden. Data from national tertiary referral centers, including Hasan Sadikin Hospital and Cipto Mangunkusumo Hospital, indicate that TNBC accounts for approximately 15%–25% of breast cancer cases (Ng et al., 2023; Yulian and Fachriza, 2024). Compared with many Western cohorts, Indonesian patients with TNBC frequently present at a younger age, with the highest incidence reported between 41 and 50 years. A considerable proportion are also diagnosed with locally advanced disease, particularly stage III tumors, contributing to 5-year survival estimates of approximately 60%–81%, depending on disease extent, access to surgery, and the presence of visceral metastases (Ng et al., 2023; Xiong et al., 2024; Yulian and Fachriza, 2024). These clinical characteristics underscore the need for therapeutic approaches that complement the current standard of care.

Historically, systemic chemotherapy has constituted the principal treatment option for TNBC. More recently, poly(ADP-ribose) polymerase inhibitors have been introduced for selected patients with BRCA-associated disease, whereas immune checkpoint inhibitors, including pembrolizumab, have been incorporated into specific clinical settings. Nevertheless, treatment responses remain heterogeneous, particularly among patients with residual invasive, recurrent, or metastatic disease (Prawiningrum et al., 2022; Bouzid et al., 2025; Zahraei et al., 2025). The molecular diversity of TNBC and the immunosuppressive characteristics of its tumor microenvironment contribute to both primary and acquired resistance to immune-based therapies (T et al., 2023; Zahraei et al., 2025). These limitations support continued investigation of multi-antigen immunotherapeutic strategies capable of directing immune recognition toward several tumor-associated antigens simultaneously.

Immunoinformatics provides a systematic framework for prioritizing candidate antigens, predicting MHC class I- and class II-restricted epitopes, excluding potentially allergenic or toxic sequences, and estimating HLA population coverage before experimental evaluation. This approach has already been applied to several TNBC-associated antigens, including MZF-1, MUC1, SOX9, TROP-2, α-lactalbumin, and cancer-testis antigens (Prawiningrum et al., 2022; Krishnamoorthy and Karuppasamy, 2023; Zahraei et al., 2025; Zhou et al., 2025).

Previous computational TNBC vaccine studies have established the feasibility of this strategy but have differed substantially in antigen selection and construct architecture. Krishnamoorthy and Karuppasamy (2023) developed a single-antigen vaccine derived from MZF-1 and incorporated the 50S ribosomal L7/L12 protein as an adjuvant. Zhou et al. (2025) subsequently evaluated seven recognized TNBC-associated proteins—MZF-1, MUC1, SOX9, keratins 5 and 14, TWIST1, and progranulin—and compared five constructs containing GM-CSF, β-defensin, IL-2, cholera enterotoxin, or 50S ribosomal L7/L12. Zahraei et al. (2025) employed a literature-informed scoring framework to prioritize nine extracellular and intracellular antigens and developed both recombinant protein and mRNA vaccine formats. Although these studies broadened the TNBC vaccine landscape, their antigen panels were primarily selected from previously established or therapeutically explored targets rather than ranked directly according to tumor-to-normal transcript-expression differences.

The present study addresses this gap by using TNBC transcriptomic data to prioritize candidate antigens according to their tumor-to-normal expression ratios. This strategy identified MMP1, CXorf61/CT83, and COL11A1 as the leading antigen sources for epitope prediction. The resulting panel represents biologically distinct aspects of TNBC, including extracellular-matrix remodeling, cancer-testis-associated antigen expression, and stromal or mesenchymal tumor biology. To our knowledge, this specific three-antigen combination has not been incorporated into the previously reported TNBC multi-epitope constructs described above.

A further challenge in recombinant multi-epitope vaccines is their potentially limited intrinsic immunogenicity, which commonly motivates the incorporation of immunomodulatory components (Prawiningrum et al., 2022). In the present construct, a MyD88-derived domain was included as an exploratory immunomodulatory component. MyD88 is a central intracellular adaptor for most Toll-like receptor and interleukin-1 receptor pathways and contributes to signaling processes associated with dendritic-cell activation and pro-inflammatory cytokine production (Bechelli et al., 2016; Chen et al., 2020). However, because MyD88 physiologically functions within the intracellular signaling compartment, the proposed activity of a MyD88-derived sequence within a recombinant extracellular protein construct remains hypothetical and requires direct experimental assessment. 

Accordingly, this study developed an immunoinformatics-guided, tumor-associated antigen-derived multi-epitope vaccine containing CTL, HTL, and linear B-cell epitopes from MMP1, CXorf61/CT83, and COL11A1. The selected epitopes were assembled with PADRE and a MyD88-derived immunomodulatory domain and subsequently evaluated for predicted physicochemical properties, structural organization, disulfide-engineering potential, TLR4 interaction, molecular flexibility, population coverage, immune-response profiles, codon adaptation, and virtual cloning feasibility. The principal contribution of this work therefore lies in the integration of expression-guided antigen prioritization, a distinct MMP1–CXorf61/CT83–COL11A1 panel, and an exploratory MyD88-derived immunomodulatory architecture. The resulting construct represents a computationally supported candidate that warrants recombinant expression and experimental immunogenicity evaluation.

MATERIALS AND METHODS

Screening of protein candidates

Gene expression data for TNBC and normal breast tissue were retrieved from the UALCAN database (https://ualcan.path.uab.edu/). UALCAN processes level-3 RNA-sequencing data from The Cancer Genome Atlas using transcripts-per-million normalization and enables subgroup-based comparisons between primary tumors and normal tissues. Target prioritization was based on the ratio of mean tumor TPM to mean normal-tissue TPM (Chandrashekar et al., 2017, 2022). Genes were ranked in descending order according to this tumor-to-normal expression ratio (Table 1).
A higher ratio indicates preferential expression in tumor tissue but does not independently establish tumor specificity or oncogenic function.

Table 1. Six genes with the highest tumor-to-normal TPM expression ratios.

Sample retrieval

Protein sequences were retrieved from the National Center for Biotechnology Information (NCBI) database using keywords relevant to the research topic to ensure appropriate species, target genes, and annotation status (O’Leary et al., 2016; Schoch et al., 2020). Each sequence was validated using its accession number and curation information to ensure data accuracy. All eligible sequences were downloaded in FASTA format for subsequent bioinformatics analyses, including epitope mapping, immunogenicity prediction, and vaccine design. Detailed information for all retrieved samples, including accession number, organism, sequence length, and curation status, is presented in Table 2.

Table 2. List of protein sequences retrieved from the NCBI database used in this study.

MHC I prediction

MHC class I epitope prediction was performed to identify cytotoxic T lymphocyte (CTL) peptides derived from highly expressed TNBC-associated proteins with strong affinity for HLA molecules. The analysis used the IEDB Next-Generation Tools platform (https://nextgen-tools.iedb.org/pipeline?tool=tc1) with the NetMHCpan 4.1 EL algorithm.  NetMHCpan 4.1 EL applies pan-allelic artificial neural networks trained on peptide–HLA binding-affinity and mass-spectrometry-derived eluted-ligand data, with individual HLA molecules represented by binding-groove pseudo-sequences to support predictions across diverse alleles (Reynisson et al., 2021; Yan et al., 2024). Twenty-seven globally representative HLA-I alleles were selected to maximize population coverage. Epitope length was fixed at nine amino acids, the optimal length for MHC I presentation. Predicted peptides were ranked separately for MMP1, CXorf61/CT83, and COL11A1, and approximately the top 1% of each protein-specific output was retained for downstream immunological screening. (Alnuqaydan and Eisa, 2024; Afnani and Purnama, 2025).

MHC II prediction

HLA class II epitopes were predicted to identify peptide regions capable of inducing CD4⁺ T helper responses essential for long-term immunity. Screening was conducted using the IEDB Next-Generation Tools server (https://nextgen-tools.iedb.org/pipeline?tool=tc2), implementing the NetMHCIIpan 4.1 EL algorithm. NetMHCIIpan 4.1 EL uses pan-allelic neural networks trained on HLA class II eluted-ligand and peptide-binding data and identifies the most probable nine-residue binding core within each 15-mer peptide (Jensen et al., 2018; Reynisson et al., 2021). Fifteen high-frequency HLA-II alleles were selected to ensure broad population relevance. A peptide core length of 15 amino acids was used, corresponding to the standard binding motif recognized by class II molecules (Tahir ul Qamar et al., 2020; Mamun et al., 2025). The resulting epitopes were ranked, and only the top 1% of peptides were shortlisted, thereby identifying highly immunogenic helper T-cell epitopes that can support durable and coordinated vaccine responses.

Linear B-cell epitope prediction

To complement T-cell immunogenicity, linear B-cell epitopes capable of stimulating antibody responses were identified using the BepiPred 2.0 tool on the IEDB server (https://tools.iedb.org/bcell/) (Jespersen et al., 2017). BepiPred 2.0 applies a random-forest classifier trained on experimentally characterized epitope and non-epitope residues to assign each amino acid a probability of belonging to a linear B-cell epitope. A threshold score of 0.500 was used to select regions with high predicted surface accessibility and antigenicity. Only peptide segments between 10 and 50 amino acids were included to align with typical experimentally validated linear epitopes and maintain structural specificity (Kar et al., 2020; Adam, 2021).

Population coverage

Population coverage analysis was conducted using the IEDB Population Coverage tool (https://tools.iedb.org/population/) to estimate the proportion of individuals whose HLA repertoire could present at least one selected T-cell epitope. The final epitope–HLA restriction pairs obtained from the CTL and HTL prediction pipelines were entered into the analysis, whereas linear B-cell epitopes were excluded because they are not restricted by HLA molecules. HLA class I- and class II-restricted epitopes were analyzed using the class-combined model across 16 geographic regions and the worldwide population dataset (Zhao et al., 2021; Martinelli, 2022).

The coverage value represents the percentage of individuals predicted to recognize at least one epitope–HLA combination, whereas the average-hit value represents the mean number of epitope–HLA combinations recognized per individual. The combined analysis was selected to capture the complementary contribution of CD8⁺ and CD4⁺ T-cell presentation pathways and to evaluate whether promiscuous epitopes restricted by multiple prevalent HLA alleles could provide broad theoretical population accessibility.

Profiling of antigenicity, allergenicity, toxicity, and IFN-γ induction

Comprehensive immunological screening was conducted to prioritize epitopes with favorable predicted immunological characteristics before vaccine assembly. Antigenicity was assessed using VaxiJen v2.0 (https://www.ddg-pharmfac.net/vaxijen/VaxiJen/VaxiJen.html), whereas allergenicity was evaluated using AllerTOP v2.1 (https://www.ddg-pharmfac.net/allertop_test/). VaxiJen applies an alignment-independent auto-cross-covariance transformation of amino-acid physicochemical properties to classify sequences according to their predicted antigenicity, whereas AllerTOP uses a k-nearest-neighbor model based on transformed sequence descriptors to distinguish allergens from non-allergens (Dimitrov et al., 2013, 2014; Doytchinova and Flower, 2007).

Toxicity was evaluated using ToxinPred (http://crdd.osdd.net/raghava/toxinpred/), and IFN-γ-induction potential was assessed using IFNepitope (https://webs.iiitd.edu.in/raghava/ifnepitope/index.php). ToxinPred applies machine-learning models derived from peptide composition and physicochemical descriptors, whereas IFNepitope uses motif-based and support-vector-machine classification to predict IFN-γ-inducing peptides (Dhanda et al., 2013; Gupta et al., 2013; Rathore et al., 2024).

A VaxiJen score greater than 0.400 was used as the predefined threshold for antigenicity. AllerTOP and ToxinPred provide categorical predictions; therefore, only peptides classified as non-allergenic and non-toxic, respectively, were retained. For IFN-γ prediction, only peptides classified as positive, corresponding to a prediction score greater than 0 under the selected model, were considered eligible for inclusion. Because the selection decision was based on the categorical IFN-γ outcome, the numerical IFN-γ scores were omitted from the final epitope tables, and only the “Positive” classification was reported. An epitope was retained only when it simultaneously fulfilled all applicable immunological screening criteria.

Vaccine construction and physicochemical properties

A multi-epitope vaccine was constructed by integrating immunomodulatory and epitope components into a single recombinant protein to support coordinated cellular and humoral immune responses against TNBC. The construct began with a MyD88-derived sequence (GenBank accession no. AAC50954.1), followed by the universal PADRE helper peptide, CTL epitopes, HTL epitopes, linear B-cell epitopes, and a C-terminal 6× histidine tag. The MyD88-derived sequence was incorporated as an exploratory immunomodulatory component because MyD88 functions as a central intracellular adaptor for most Toll-like receptor and interleukin-1 receptor signaling pathways and is associated with dendritic-cell activation and Th1-oriented cytokine responses (Bechelli et al., 2016; Chen et al., 2020). Its inclusion was intended to support the potential coupling of innate and adaptive immune signaling; however, the immunomodulatory activity and intracellular accessibility of this component within the recombinant construct remain hypothetical and require experimental validation.

PADRE was included as a broadly reactive helper T-cell epitope to facilitate CD4⁺ T-cell support across diverse HLA class II backgrounds and thereby complement the selected antigen-derived HTL epitopes. A rigid EAAK linker was positioned between the MyD88-derived domain and PADRE to promote functional separation and minimize unfavorable structural interference between these modules. CTL epitopes were joined using AAY linkers, HTL epitopes using GPGPG linkers, and B-cell epitopes using KK linkers. AAY linkers were selected because they are commonly used to facilitate proteasomal processing and separation of adjacent MHC class I–restricted epitopes. GPGPG linkers were used to maintain the relative independence of HTL epitopes and reduce the formation of unintended junctional epitopes, whereas KK linkers served as short, protease-sensitive spacers intended to preserve the accessibility of linear B-cell epitopes and limit unfavorable inter-epitope interactions (Adeleke et al., 2021; Gustiananda et al., 2021; Shamakhi and Kordbacheh, 2021; Priyamvada and Ramaiah, 2023). Physicochemical profiling of the assembled sequence was performed using the ProtParam web server (https://web.expasy.org/protparam/) to determine molecular weight, theoretical pI, stability, aliphatic index, and predicted half-life, ensuring that the final construct met biochemical criteria for protein expression and subsequent structural studies (Elalouf, 2023; Nshimiyimana et al., 2024).

Secondary and tertiary vaccine structure prediction

Secondary and tertiary vaccine structure prediction was conducted to assess the structural organization, folding behavior, and conformational stability of the designed multi-epitope vaccine. Secondary structure elements were predicted using SOPMA (https://npsa-prabi.ibcp.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_sopma_f.html) and PSIPRED (https://bioinf.cs.ucl.ac.uk/psipred/), and the results were integrated by consensus to enhance prediction confidence. PSIPRED uses position-specific sequence profiles as input to a two-stage neural-network classifier, whereas SOPMA combines local sequence information with homologous-sequence alignments to assign individual residues to predicted secondary-structure states (McGuffin et al., 2000; Buchan et al., 2013; Elalouf, 2023; Afnani et al., 2026). The three-dimensional structure of the vaccine construct was then modeled using the AlphaFold Server (https://alphafoldserver.com/), which applies deep–learning algorithms to generate high-accuracy protein folding models approaching near-experimental resolution (Laskowski and Thornton, 2022; Desai et al., 2024; Rubach et al., 2024). To evaluate the reliability and stereochemical quality of the predicted 3D model, structural validation was performed using ProSA (https://prosa.services.came.sbg.ac.at/prosa.php) to assess Z-score deviation from experimentally solved protein structures and the Structure Analysis and Verification Server (SAVES) (https://saves.mbi.ucla.edu/), incorporating ERRAT for non-bonded interaction quality assessment and Ramachandran plot analysis to verify backbone dihedral angle distributions (Wiederstein and Sippl, 2007; Chinthakunta et al., 2018; Zhu et al., 2023).

Disulfide engineering of the vaccine construct

Disulfide engineering was performed to identify residue pairs that could potentially be converted into cysteine pairs and form additional intramolecular disulfide bridges within the vaccine structure. The selected three-dimensional vaccine model was submitted to the Disulfide by Design 2 server. DbD2 evaluates the geometric compatibility of residue pairs by modeling cysteine substitutions and estimating the resulting disulfide-bond geometry and energy (Dombkowski, 2003; Craig and Dombkowski, 2013). Both intra-chain and inter-chain searches were enabled, and virtual Cβ atoms were generated for glycine residues where required. Candidate pairs were screened using χ3-angle targets of −87° or +97° with a tolerance of ±30° and a Cα–Cβ–Sγ angle of 114.6° ± 10°. Following inspection of the predicted geometry, bond energy, and spatial distribution, ten residue pairs were selected for cysteine substitution and mutant-model generation. The wild-type and engineered structures were subsequently visualized to map the positions of the proposed disulfide bridges.

Molecular docking and visualization

Molecular docking was conducted to assess the HDOCK score and interaction landscape between the designed vaccine and its immune receptor targets. Docking was performed using the HDOCK server (http://hdock.phys.hust.edu.cn/), which integrates template-based and ab initio algorithms for reliable protein–protein interaction prediction. HDOCK performs a global docking search through a hybrid template-based and ab initio strategy, followed by knowledge-based scoring and ranking of candidate vaccine–receptor complexes (Yan et al., 2020). To provide a detailed view of atomic-level stabilizing interactions, post-docking analyses of hydrogen bonds and salt bridges were performed using PDBsum (https://www.ebi.ac.uk/thornton-srv/databases/pdbsum/) (Laskowski and Thornton, 2022). Bond visualization and ligand–receptor interaction mapping were generated using LigPlus version 2.3, providing an intuitive depiction of contact networks within the binding interface (Arumugam and Varamballi, 2021). This pipeline provided a comprehensive overview of both binding strength and interaction specificity, thereby supporting the immunological relevance of the assembled vaccine.

Molecular flexibility and dynamics analysis

Molecular flexibility and dynamics analyses were performed to evaluate the structural flexibility and dynamic stability of the docked vaccine–receptor complex under near-physiological conditions. The initial simulation was carried out using the iMODS platform (https://imods.iqf.csic.es/), which assesses deformability, molecular mobility, and eigenvalues to estimate resistance to deformation and the overall robustness of the interaction (López-Blanco et al., 2014; Zaib et al., 2023). To further estimate residue-level fluctuation patterns, CABS-flex 2.0 (https://biocomp.chem.uw.edu.pl/CABSflex2/) was used to generate quantitative root-mean-square fluctuation (RMSF) profiles—a threshold of <2 Å was considered indicative of favorable structural stability (Kuriata et al., 2018; Kurcinski et al., 2019). The combined outputs were used to characterize the predicted flexibility and collective motions of the modeled vaccine–receptor complex. These analyses do not establish stability or immune activation under biological conditions.

Immune simulation

An immune response simulation was conducted to evaluate the capacity of the designed vaccine to generate coordinated humoral and cellular immunity following repeated antigen exposure. The simulation used the C-ImmSim server version 10.1 (http://kraken.iac.rm.cnr.it/C-IMMSIM/), an agent-based modeling platform that reproduces the dynamics of mammalian immune networks by incorporating interactions among B cells, T cells, antigen-presenting cells, and cytokine signaling pathways (Rapin et al., 2011; Cheng et al., 2022). The immunization protocol consisted of three injections at time points 1, 84, and 170, corresponding to approximately 4-week intervals between doses, mirroring conventional vaccination schedules. The simulation was extended to 1,050 time steps to monitor both primary and secondary immune responses, allowing assessment of memory cell development, antibody class switching, interferon-mediated signaling, and long-term immune persistence (Sarvmeili et al., 2024; Ullah et al., 2024; Chansap et al., 2025). This in silico model provides a predictive assessment of the vaccine’s immunogenic potential before experimental validation.

Codon optimization and in silico cloning

To enhance translational efficiency and recombinant protein expression in a bacterial production system, the final vaccine construct was codon-optimized for the Escherichia coli strain K12 using the Java Codon Adaptation Tool (JCat) server (https://www.jcat.de/). JCat replaces synonymous codons according to host-specific codon-usage frequencies while maximizing the codon adaptation index and excluding user-selected sequence motifs that may interfere with transcription or translation (Grote et al., 2005; Arumugam and Varamballi, 2021). Optimization parameters included avoiding rho-independent termination sites, restriction enzyme cleavage sites, and prokaryotic instability motifs, while maximizing the codon adaptation index and GC balance to promote efficient transcription and translation. The optimized nucleotide sequence was then subjected to in silico cloning into the pET-28a(+) expression vector using SnapGene version 8.2 (Naderian et al., 2025). SnapGene simulates restriction digestion, fragment insertion, ligation, construct orientation, and reading-frame continuity by mapping enzyme-recognition sites and annotated sequence features within the recombinant plasmid (Fatoba et al., 2021). Restriction sites for EcoRI and BamHI were incorporated to facilitate directional cloning while maintaining compatibility with the vector’s multiple cloning sites (Hossan et al., 2021; Parvin et al., 2024). This computational approach ensured the feasibility of heterologous expression prior to future wet-lab synthesis.

The complete analytical workflow used in this study is summarized in Figure 1. The analysis proceeded sequentially from expression-based target prioritization and protein-sequence retrieval to MHC class I, MHC class II, and linear B-cell epitope prediction, followed by immunological screening and population-coverage assessment. The retained epitopes were subsequently assembled into a multi-epitope construct and evaluated through physicochemical characterization, structural modeling, disulfide engineering, molecular docking, flexibility analysis, immune simulation, codon optimization, and in silico cloning. The specific algorithms, thresholds, and selection criteria applied at each stage are described in the corresponding subsections above.

Figure 1. Overview of the immunoinformatics workflow used to design and computationally evaluate the TNBC multi-epitope vaccine candidate.

RESULTS

Screening of protein candidates

The comparative gene expression analysis identified a set of candidate proteins with markedly elevated expression in TNBC tumors compared with normal breast tissue (Table 1). Matrix metalloproteinase-1 (MMP1) exhibited the highest tumor-to-normal expression ratio (~280.6), indicating extreme overexpression in TNBC (Wang et al., 2018). This was followed by CXorf61 (also known as CT83) with a ratio of ~129.0 and collagen type XI alpha-1 (COL11A1) with a ratio of ~63.6 (Abdou et al., 2022). Other genes, such as COL10A1, PRAME, and MMP11, also showed high differential expression (ratios of ~48–57), although less pronounced than the top three. Based on these findings, three proteins (MMP1, CXorf61, and COL11A1) were selected for downstream epitope prediction and vaccine design owing to their exceptionally high tumor-specific expression and potential oncogenic relevance. The pronounced overexpression of these gene products in TNBC suggests that they represent tumor-associated antigens suitable for vaccine targeting, as their limited normal expression may reduce the risk of off-target effects.

Figure 2. Differential expression of candidate tumor-associated antigens in TNBC versus normal breast tissue. Genes are ranked by the mean tumor/normal TPM expression ratio. The figure shows expression levels of the top 25 genes, highlighting their preferential expression in tumor samples.

MHC I prediction

Computational screening of cytotoxic T-lymphocyte epitopes generated protein-specific sets of MHC class I peptide predictions. For MMP1, 12,403 initial predictions were generated, of which the 124 highest-ranked candidates, representing approximately the top 1%, were advanced to subsequent immunological screening. The analysis of CXorf61 generated 2,949 initial predictions, from which 29 top-ranked candidates were retained, whereas 451 of the 45,132 predictions generated from COL11A1 were selected for further evaluation. The approximately 1% selection threshold was applied independently to each protein to prevent differences in protein length and the number of generated predictions from disproportionately influencing candidate prioritization.

The shortlisted candidates were subsequently assessed using the predefined screening criteria. Peptides were retained only when they had a VaxiJen antigenicity score greater than 0.400, were classified as non-allergenic by AllerTOP v2.1 and non-toxic by ToxinPred, and received a positive IFN-γ prediction. The numerical IFN-γ scores were not included in Table 3 because the categorical positive or negative outcome was used for candidate selection.

Following sequential screening, only the MMP1- and COL11A1-derived peptides presented in Table 3 simultaneously fulfilled all applied criteria. These epitopes were also associated with multiple HLA-A or HLA-B alleles, supporting their computational prioritization as promiscuous CTL components of the final construct. Although 29 CXorf61-derived peptides proceeded to secondary screening, none simultaneously met the antigenicity, allergenicity, toxicity, IFN-γ, and HLA-binding criteria. Consequently, no CXorf61-derived 9-mer was retained in the final MHC class I epitope set. These results provide the computational basis for selecting the CTL components incorporated into the vaccine construct but do not constitute experimental validation of antigen presentation or T-cell activation.

Table 3. Predicted HLA class I–restricted CTL epitopes from the top TNBC antigens.

MHC II prediction

Helper T-cell epitope prediction generated protein-specific sets of MHC class II candidates. The analysis produced 1,419 initial predictions from MMP1, of which 14 top-ranked candidates were retained for downstream screening. For CXorf61, 8 of the 832 initial predictions were selected, whereas 53 of the 5,386 COL11A1-derived predictions were advanced. As in the MHC class I analysis, the approximately 1% ranking threshold was applied separately to each protein before secondary immunological filtering.

The shortlisted peptides were subsequently evaluated using the same predefined immunological criteria, including a VaxiJen antigenicity score >0.400, non-allergenic classification, non-toxic classification, and positive predicted IFN-γ induction. Sequential screening yielded three final MMP1-derived HTL epitopes and one epitope each from CXorf61 and COL11A1, as presented in Table 4. Their predicted binding to multiple HLA-DR molecules and fulfillment of all applicable screening criteria supported their computational prioritization for inclusion in the HTL component of the construct.

Table 4. Predicted HLA class II–restricted T-cell epitopes from selected TNBC-associated proteins.

Linnear B-cell epitope prediction

Linear B-cell epitope mapping identified multiple antigenic regions across MMP1, CXorf61, and COL11A1 that met predefined thresholds for surface accessibility and antigenicity (Table 5). All selected peptides were predicted to be non-allergenic and non-toxic, indicating theoretical suitability for eliciting humoral immune responses. These B-cell epitopes complement the T-cell components of the vaccine design by providing antibody-targetable regions within the selected tumor-associated antigens. Because the analysis was limited to sequence-based linear epitopes, these findings do not establish the presence or accessibility of conformational antibody-binding sites in the folded vaccine protein.

Table 5. Predicted linear B-cell epitopes from the high-expression TNBC target proteins.

Population coverage

The class-combined analysis of the selected HLA class I- and class II-restricted epitopes predicted 99.36% worldwide population coverage, with an average of 3.14 epitope–HLA hits per individual (Table 6). Across the 16 geographic regions evaluated, the mean predicted coverage was 93.34% ± 10.55%, whereas the corresponding mean average-hit value was 2.39 ± 0.71. Coverage exceeded 90% in 14 of the 16 regions, with the highest estimates observed in Europe (99.86%), North America (99.64%), East Asia (99.28%), and the West Indies (98.93%). Southeast Asia showed predicted coverage of 94.57%, with an average-hit value of 2.04.

Regional variability nevertheless remained evident. Central America displayed the lowest estimated coverage at 53.14%, followed by South Africa at 88.25%, indicating that the broad worldwide estimate was not uniformly distributed across all populations. The high worldwide coverage was associated with the inclusion of promiscuous T-cell epitopes restricted by multiple prevalent HLA alleles and with the complementary contribution of class I- and class II-mediated antigen presentation. These values represent theoretical HLA-based accessibility and do not directly predict the magnitude or clinical effectiveness of an immune response.

Table 6. Predicted combined HLA class I and class II population coverage of the selected T-cell epitopes across 16 geographic regions and the worldwide population.

Vaccine construction and physicochemical properties

Following epitope selection, the multi-epitope vaccine construct was assembled in silico by concatenating the adjuvant and epitope components with appropriate linkers. The N-terminal MyD88-derived sequence, included as an exploratory immunomodulatory component, was fused to the universal PADRE helper peptide (AKFVAAWTLKAAA) to promote CD4⁺ T-cell activation (Zhao et al., 2025). This was followed by an array of selected CTL, HTL, and B-cell epitopes in tandem, separated by linkers designed to preserve epitope immunogenicity and facilitate proper processing. Specifically, AAY linkers were used between CTL epitopes to favor efficient proteasomal cleavage, GPGPG linkers were placed between HTL epitopes to maintain their conformational independence, and KK linkers were used as flexible spacers between B-cell epitopes (Adeleke et al., 2021; Gustiananda et al., 2021). A C-terminal hexahistidine tag (6×His) was added to aid in
the purification of the recombinant protein (Booth et al., 2018). The modular arrangement was intended to maintain separation among the functional components and support their predicted processing. The final multi-epitope construct comprised 621 amino acids (Figure 3).

Figure 3. A schematic representation of the final chimeric vaccine protein, with the MyD88 adjuvant and PADRE at the amino terminus, followed by each epitope (white boxes) connected by the specified linker sequences (colored segments), and the His-tag at the carboxyl terminus.

In silico physicochemical analysis indicated that the chimeric construct met key biochemical thresholds for further development (Table 7). The protein exhibited a suitable molecular weight (~68.9 kDa), a basic theoretical pI (9.07), and a stable instability index (<40), suggesting favorable structural stability (Khatoon et al., 2017; Nshimiyimana et al., 2024). The aliphatic index and negative GRAVY value, which supported predicted thermostability and hydrophilicity, respectively, were consistent with a high solubility score (Oluwagbemi et al., 2022). In addition, the construct surpassed established antigenicity cutoffs in both VaxiJen and ANTIGENpro assessments and was classified as non-allergenic. Collectively, these results suggest that the designed multi-epitope vaccine possesses physicochemical and immunological properties suitable for recombinant expression and downstream evaluation.

Table 7. Predicted physicochemical and immunogenic properties of the designed multi-epitope vaccine protein.

Secondary and tertiary vaccine structure prediction

Secondary structure analysis of the designed multi-epitope vaccine was performed to assess the distribution of fundamental structural elements along the protein sequence (Elalouf, 2023). The PSIPRED prediction indicated that the vaccine construct is predominantly α-helical and coil-like, with a smaller proportion of β-strands (Figure 4a). This pattern aligns with the construct's chimeric nature, in which the MyD88-derived adjuvant domain contributes substantial α-helical content. At the same time, epitope-rich regions and flexible linkers introduce coil-dominated segments. A complementary analysis using SOPMA yielded a similar secondary-structure distribution, reinforcing the reliability of the predicted folding pattern through methodological consensus (Figure 4b). The agreement between PSIPRED and SOPMA indicated a broadly comparable predicted distribution of helices, strands, and coil regions.

Figure 4. Secondary structure distribution predicted using PSIPRED (a). Secondary structure prediction obtained from SOPMA, demonstrating a comparable structural composition and supporting consensus-based prediction reliability (b).

Tertiary-structure modeling was performed using AlphaFold to generate a three-dimensional representation of the full-length vaccine construct. The predicted 3D structure adopts a compact, globular conformation, with well-organized α-helical bundles and β-sheet elements connected by flexible loops (Figure 5a). Structural validation using ProSA yielded a Z-score of –7.43, within the range typically observed for experimentally solved proteins of comparable size, indicating the absence of major structural anomalies (Figure 5b–c) (Chinthakunta et al., 2018; Zaib et al., 2023). Further stereochemical assessment via Ramachandran plot analysis showed that 97.45% of residues were in favored regions, exceeding the commonly accepted threshold of >90% for high-quality models (Figure 5d) (Rao et al., 2020; Mamun et al., 2025). Collectively, these validation metrics indicate that the predicted tertiary structure is structurally plausible and suitable for downstream analyses of interactions and dynamics.

Figure 5. 3D structure of the vaccine protein predicted using AlphaFold (a). Structural validation using ProSA, showing the Z-score profile with an overall Z-score of –7.43
(b-c). Ramachandran plot with 97.45% in favored regions (d).

Disulfide engineering of the vaccine construct

Disulfide by Design 2 identified 15 geometrically compatible residue pairs within chain A of the modeled vaccine structure. Of these, ten pairs were selected for construction of the cysteine-engineered model: Ala24–Tyr116, Ser34–Thr71, Leu37–Trp47, Ala45–Ala49, Pro139–Glu143, Leu150–Glu159, Tyr167–Val222, Val193–Arg196, Cys203–Ser206, and Ser224–Tyr227. The corresponding predicted χ3 angles ranged from −100.43° to +119.12°, whereas the estimated disulfide-bond energies ranged from 0.84 to 6.27 kcal/mol (Table 8). All selected pairs were located within chain A and were modeled as ten potential intramolecular disulfide bridges. Because residue Cys203 was already present in the original sequence, formation of the Cys203–Cys206 pair required mutation only at position 206. Comparison of the wild-type and engineered structures showed that the proposed bonds were distributed across several regions of the vaccine model rather than being concentrated within a single structural segment (Figure 6).

Table 8. Predicted residue pairs selected for disulfide engineering.

Figure 6. Disulfide engineering of the three-dimensional multi-epitope vaccine model. (a) Wild-type vaccine structure before cysteine substitution. (b) Cysteine-engineered model containing ten proposed intramolecular disulfide bridges. The positions of the modeled disulfide bonds are indicated by yellow circles.

Molecular docking and visualization

Molecular docking was performed to examine the predicted structural compatibility between the vaccine model and the TLR4 ectodomain (Kanse et al., 2023; Li et al., 2014). The top-ranked vaccine–TLR4 complex yielded an HDOCK score of −298.65 (Table 9). Interface analysis identified seven predicted salt bridges and 12 hydrogen bonds, indicating a plausible modeled contact interface. However, the HDOCK value represents a relative docking-ranking score rather than an experimentally measured binding free energy. Accordingly, these findings do not establish stable receptor binding, TLR4 agonism, receptor activation, or MyD88-dependent signaling.

Table 9. Molecular docking result and protein–protein docking interactions between the vaccine construct and TLR4.

The predicted interfacial contacts further support the structural compatibility of the modeled vaccine–TLR4 complex (Figure 7). Nevertheless, docking alone cannot determine whether the construct binds TLR4 under physiological conditions or activates downstream immune signaling. These possibilities require confirmation using receptor-binding and TLR4-dependent functional assays.

Figure 7. Predicted docked complex between the multi-epitope vaccine model and the TLR4 ectodomain, showing the modeled interface and selected interfacial contacts.

Molecular flexibility and dynamics analysis

Molecular flexibility analysis was conducted using normal-mode approaches to assess the intrinsic flexibility and collective motion of the vaccine–receptor complex. Deformability analysis showed that structural flexibility was limited and localized primarily to loop and linker regions, whereas the majority of the protein backbone exhibited low deformability (Figure 8a). Consistent with this, the eigenvalue profile indicated high resistance to large-scale deformation (Figure 8b), and variance analysis revealed low-amplitude collective motions across most residues (Figure 8c), collectively suggesting that the complex possesses a mechanically stable framework with restricted global mobility (Parvin et al., 2024).

Localized atomic fluctuations were further illustrated by the B-factor distribution, which showed increased flexibility primarily in terminal and loop regions. In contrast, epitope-containing domains and the adjuvant core remained comparatively rigid (Figure 8d). A complementary residue-level flexibility analysis using CABS-flex yielded an average RMSF of 1.93 Å (Figure 8f), below the commonly accepted 2.0 Å threshold for stable protein complexes (Elfadil et al., 2025; Nirob et al., 2025). These results suggest that the designed multi-epitope vaccine exhibits a favorable balance between structural rigidity and localized flexibility, considered advantageous for maintaining conformational integrity while permitting adaptive interactions during immune recognition.

Figure 8. Molecular flexibility analysis of the vaccine–receptor complex, including the deformability profile along the protein backbone (a), eigenvalue distribution from normal mode analysis indicating resistance to global deformation (b), variance associated with collective residue motions (c), B-factor representation highlighting localized atomic fluctuations (d), elastic network model illustrating residue connectivity and motion constraints (e), and RMSF profile from CABS-flex analysis with an average fluctuation value of 1.93 Å (f).

Immune simulation

In silico immune simulation suggested that the multi-epitope vaccine has the theoretical capacity to induce coordinated humoral and cellular immune responses following repeated antigen exposure (Figure 9). The predicted immune profile indicated effective antibody class switching and memory formation, sustained activation of CD4⁺ helper and CD8⁺ cytotoxic T cells, and a cytokine environment indicative of a Th1-skewed response. Concurrent activation of antigen-presenting cells and maintenance of a broad immune repertoire further support the hypothesis that the construct could engage both innate and adaptive immune pathways in a balanced manner (Omoniyi et al., 2022). These findings align with the vaccine's intended immunological design.

Figure 9. Predicted immune-response profiles generated by C-ImmSim following simulated administration of the multi-epitope vaccine. Antigen and antibody profiles (IgM, IgG1, IgG2) (a). Cytokine and interleukin concentrations (b). B cell and memory cell populations by isotype (c). B lymphocyte population by entity state (d). CD4⁺ T-helper lymphocyte counts (e). CD8⁺ T-cytotoxic lymphocyte counts (f).

Codon optimization and in silico cloning

To facilitate experimental validation, the final vaccine protein sequence was reverse-translated and optimized for expression in a suitable host; E. coli was chosen as the model expression system. The codon optimization process adjusted the DNA sequence of the vaccine construct to match E. coli codon usage preferences without altering the amino acid sequence. This optimization yielded a coding sequence of 1839 base pairs (excluding the stop codon), compatible with standard cloning and expression vectors (Figure 10a).

Key metrics of the optimized gene suggest efficient expression. The codon adaptation index (CAI) of the optimized sequence is 0.95, very close to the ideal value of 1.0, indicating excellent agreement with E. coli codon bias (values above ~0.8 are generally considered optimal for high expression) (Dong et al., 2020; Zubair et al., 2025). The final GC content of the DNA sequence is 53.2%, within the desirable range (30–70%) for stability and transcriptional efficiency in E. coli (Mia et al., 2022; Sarvmeili et al., 2024). Additionally, the sequence was screened to remove problematic elements: no prokaryotic ribosome-binding sites or terminator motifs were inadvertently introduced, and rare codons were minimized. These adjustments reduce the risk of translational pausing or secondary-structure issues in the bacterial host.

Figure 10. The optimized sequence was restricted with EcoRI and BamHI (a). In silico insertion of the optimized gene into the pET-28a(+) expression vector (b). In silico agarose gel electrophoresis preview of the recombinant vaccine construct (c).

Following sequence optimization, in silico cloning was performed to verify that the vaccine gene could be inserted into an expression plasmid. The optimized gene was virtually cloned into the pET-28a(+) expression vector, a commonly used plasmid for protein expression in E. coli. In this virtual cloning, appropriate restriction sites (EcoRI at the 5′ end and BamHI at the 3′ end) were added to the gene, and the sequence was inserted into the vector’s multiple cloning site under the control of the T7 promoter (Figure 10b). A detailed view of the integration confirmed that the gene’s reading frame aligns correctly with the vector’s translational start and that no premature stop codons were introduced (Figure 10c). The virtual cloning analysis supports the technical feasibility of inserting the optimized sequence into the selected expression vector. By achieving a high CAI and an appropriate GC balance, and by successfully simulating its insertion into an expression vector, the study addresses the downstream feasibility of producing the vaccine.

DISCUSSION

Triple-negative breast cancer (TNBC) remains difficult to manage because of its aggressive behavior, molecular heterogeneity, metastatic propensity, and absence of endocrine- or HER2-directed therapeutic targets. In this study, an integrated immunoinformatics pipeline was used to design a tumor-associated antigen-derived multi-epitope vaccine incorporating CTL, HTL, and linear B-cell epitopes from MMP1, CXorf61/CT83, and COL11A1. These antigens were prioritized according to their tumor-to-normal transcript-expression ratios and assembled with PADRE and a MyD88-derived immunomodulatory domain. The resulting analyses provide computational evidence regarding HLA accessibility, physicochemical properties, structural plausibility, receptor-docking compatibility, immune-response profiles, and recombinant-expression feasibility, but they do not establish biological immunogenicity or antitumor efficacy.

The antigen-selection strategy distinguishes the present construct from previous TNBC multi-epitope designs. Krishnamoorthy and Karuppasamy (2023) focused on MZF-1, Zhou et al. (2025) combined seven established TNBC-associated proteins and compared five adjuvant-containing constructs, and Zahraei et al. (2025) prioritized nine extracellular and intracellular antigens using a literature-informed framework and evaluated both protein and mRNA platforms. By contrast, the present study used differential transcript abundance between TNBC and normal breast tissue as the initial screening criterion and identified MMP1, CXorf61/CT83, and COL11A1 as the leading antigen sources. This expression-guided three-antigen combination may broaden immune recognition across biologically distinct features of TNBC, including extracellular-matrix remodeling, stromal or mesenchymal biology, and cancer-testis antigen expression. Nevertheless, transcript overexpression does not establish tumor exclusivity, protein abundance, or safety; these targets require verification using independent transcriptomic and proteomic datasets, immunohistochemistry, and systematic normal-tissue expression analysis.

The final T-cell epitope set showed complementary restriction across multiple HLA class I and class II alleles, producing predicted worldwide population coverage of 99.36% and an average of 3.14 epitope–HLA hits per individual. Coverage in Southeast Asia reached 94.57%, supporting the theoretical regional relevance of the construct. The high estimate likely reflects the inclusion of promiscuous epitopes associated with several prevalent HLA alleles and the combined contribution of class I- and class II-restricted presentation. However, population coverage does not account for antigen processing, HLA expression, immunodominance, T-cell repertoire diversity, antigen abundance, or tumor-mediated immune suppression. It should therefore be interpreted as theoretical HLA accessibility rather than the proportion of individuals expected to develop an effective antitumor response.

A recombinant protein platform was selected because it enables the CTL, HTL, B-cell, PADRE, linker, and immunomodulatory components to be produced as a single, compositionally defined polypeptide that can be directly characterized for identity, purity, aggregation, and antigen content. This format is compatible with established microbial-expression and affinity-purification systems and may support standardized off-the-shelf production. Clinical studies of recombinant HER2 protein immunotherapeutics have demonstrated manufacturing feasibility and antigen-specific immune responses in breast cancer, although these findings were obtained in HER2-positive disease and cannot be extrapolated directly to TNBC (Curigliano et al., 2016; Limentani et al., 2016). The principal limitation is that exogenous proteins are primarily processed through endosomal pathways, whereas induction of CTL responses depends on efficient cross-presentation by dendritic cells. Recombinant proteins may also exhibit modest intrinsic immunogenicity, extracellular degradation, aggregation, and dependence on adjuvants or delivery systems.

Alternative platforms offer different advantages but also introduce distinct translational constraints. Peptide vaccines are chemically defined and comparatively straightforward to manufacture, yet may be limited by HLA specificity, rapid degradation, and weak immunogenicity. Trials of HER2-derived peptide vaccines such as GP2 and AE37 showed acceptable safety and immune activity but did not establish uniform clinical benefit across all study populations (Brown et al., 2020). Nanoliposomal delivery of the E75–AE36 multi-epitope peptide with PADRE improved immune responses in a murine breast-cancer model, although nanoparticle systems introduce additional formulation, stability, biodistribution, and scale-up challenges (Zamani et al., 2020). DNA and mRNA vaccines support intracellular antigen expression and may facilitate both MHC class I and II presentation; a mammaglobin-A DNA vaccine induced antigen-specific CD8⁺ T-cell responses in a small phase I breast-cancer study, whereas personalized RNA vaccines have generated poly-epitope responses in early cancer trials (Tiriveedhi et al., 2014; Sahin et al., 2017). Viral vectors can provide strong antigen expression and innate stimulation, as illustrated by the clinical evaluation of PANVAC, but pre-existing or induced vector immunity and more complex manufacturing may limit repeated administration (Heery et al., 2015). Thus, the recombinant protein format was selected for product definition and manufacturing feasibility rather than presumed superiority. Comparative testing of protein, nucleic-acid, and nanoparticle formulations of the same epitope architecture would be informative.

The modular organization of the 621-amino-acid construct was designed to accommodate distinct antigen-processing requirements. EAAK was used to separate the MyD88-derived component from PADRE, whereas AAY, GPGPG, and KK linkers were incorporated between CTL, HTL, and linear B-cell epitopes, respectively (Adeleke et al., 2021; Gustiananda et al., 2021; Shamakhi and Kordbacheh, 2021; Priyamvada and Ramaiah, 2023). These functions remain predictive because sequence arrangement and secondary-structure modeling cannot establish proteasomal cleavage, endosomal processing, junctional-epitope formation, or HLA presentation. Similarly, the MyD88-derived sequence was selected because MyD88 is central to TLR/IL-1R signaling and dendritic-cell activation (Bechelli et al., 2016; Chen et al., 2020). However, MyD88 is physiologically a cytosolic adaptor rather than a TLR4 ectodomain ligand; its incorporation into an extracellular recombinant protein does not demonstrate cytosolic delivery or signaling activity. Cellular uptake, intracellular localization, dendritic-cell activation, and cytokine induction therefore require direct evaluation.

Structural analyses indicated that the construct could adopt a stereochemically plausible tertiary model, while disulfide engineering identified ten candidate intramolecular bridges that may restrict local flexibility. Emadi et al. (2026) similarly applied cysteine substitution to optimize a computationally designed multi-epitope cancer vaccine. Nevertheless, the predicted bond energies varied, and cysteine substitutions could alter folding, solubility, antigen processing, or epitope accessibility. Comparative simulations and experimental characterization of wild-type and engineered proteins are needed before any stabilizing effect can be established. Docking also generated a favorable HDOCK score and several predicted hydrogen bonds and salt bridges with the TLR4 ectodomain. This finding indicates possible structural compatibility but does not represent experimentally measured binding free energy or demonstrate receptor agonism, dimerization, or MyD88-dependent signaling. These functions should be tested using receptor-binding assays, TLR4 reporter systems, and appropriate positive, negative, and receptor-deficient controls.

C-ImmSim predicted antibody production, CTL and helper T-cell expansion, IFN-γ-associated responses, and memory-cell formation following repeated virtual immunization. Comparable Th1-oriented patterns have been reported in previous computational TNBC vaccine studies (Krishnamoorthy and Karuppasamy, 2023; Zahraei et al., 2025; Zhou et al., 2025). However, the simulation does not reproduce the full TNBC microenvironment, including antigen heterogeneity, regulatory T cells, myeloid-derived suppressor cells, immune-checkpoint signaling, stromal exclusion, T-cell exhaustion, and HLA loss. The simulated responses should therefore be regarded as hypothesis-generating rather than evidence of tumor-cell killing or therapeutic efficacy.

Codon optimization generated a CAI of 0.95 and a GC content of 53.2%, while virtual cloning into pET-28a(+) supported the technical feasibility of constructing the proposed expression plasmid. These parameters do not confirm transcription, translation, solubility, correct folding, or biological activity. Expression in Escherichia coli may also result in inclusion-body formation and requires rigorous endotoxin removal, particularly because residual lipopolysaccharide could confound TLR4-related assays. The next experimental priorities should include small-scale recombinant expression, purification, endotoxin quantification, biophysical characterization, antigen uptake, and direct assessment of epitope presentation. Subsequent studies should examine dendritic-cell maturation, antigen-specific CD4⁺ and CD8⁺ T-cell activation, cytokine production, tumor-cell recognition, and cross-reactivity with normal tissues. Comparative constructs lacking the MyD88-derived domain or containing alternative adjuvant and delivery systems would further clarify the contribution of each component. Overall, this study provides a rational computational foundation for a recombinant TNBC vaccine candidate, but its biological and translational relevance remains contingent on systematic in vitro and in vivo validation.

CONCLUSION

This study presents an expression-guided immunoinformatics strategy for designing a recombinant multi-epitope vaccine candidate derived from MMP1, CXorf61/CT83, and COL11A1. The selected CTL, HTL, and linear B-cell epitopes fulfilled the predefined computational screening criteria and provided complementary HLA class I and class II restriction profiles, resulting in predicted worldwide population coverage of 99.36%. Physicochemical assessment, structural modeling, disulfide engineering, docking, flexibility analysis, immune simulation, codon optimization, and virtual cloning collectively support the technical feasibility of advancing the construct to recombinant-expression studies. However, the predicted vaccine–TLR4 association does not establish receptor agonism or MyD88-dependent signaling, and the simulated immune responses do not constitute evidence of biological immunogenicity or antitumor efficacy. The functional contribution of the MyD88-derived domain, natural processing and presentation of the incorporated epitopes, protein folding, cross-presentation, and potential cross-reactivity with normal tissues remain unresolved. Further studies should therefore include recombinant expression and purification, endotoxin-controlled biophysical characterization, antigen-presentation assays, dendritic-cell and T-cell functional testing, normal-tissue reactivity assessment, and subsequent in vivo evaluation. The proposed construct should consequently be regarded as a computationally prioritized TNBC vaccine candidate that warrants systematic experimental validation.

ACKNOWLEDGEMENT

We acknowledge the developers and maintainers of the databases and web servers used in this study.

AUTHOR CONTRIBUTIONS

Moh Royhan Afnani: Conceptualization (Equal), Investigation (Lead), Data Curation (Lead), Validation (Equal), Visualization (Equal), Writing – Original Draft (Lead); Volta Kellik Setiawan: Investigation (Equal), Data Curation (Equal), Validation (Equal), Visualization (Equal), Writing – Original Draft (Equal); Anwar Rovik: Conceptualization (Equal), Investigation (Equal), Supervision (Lead), Validation (Equal), Visualization (Equal), Writing – Original Draft (Equal), Writing – Review & Editing (Lead).

CONFLICT OF INTEREST

The authors declare that they have no conflicts of interest.

REFERENCES

Abdou, Y., Goudarzi, A., Yu, J.X., Upadhaya, S., Vincent, B., and Carey, L.A. 2022. Immunotherapy in triple negative breast cancer: Beyond checkpoint inhibitors. npj Breast Cancer. 8(1): 121. https://doi.org/10.1038/s41523-022-00486-y

Adam, K.M. 2021. Immunoinformatics approach for multi-epitope vaccine design against structural proteins and ORF1a polyprotein of severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2). Tropical Diseases, Travel Medicine and Vaccines. 7(1): 22. https://doi.org/10.1186/s40794-021-00147-1

Adeleke, V.T., Adeniyi, A.A., Adeleke, M.A., Okpeku, M., and Lokhat, D. 2021. The design of multiepitope vaccines from plasmids of diarrheagenic Escherichia coli against diarrhoea infection: Immunoinformatics approach. Infection, Genetics and Evolution. 91(March): 104803. https://doi.org/10.1016/j.meegid.2021.104803

Afnani, Moh. R. and Purnama, E.R. 2025. Design of a recombinant multi-epitope subunit vaccine targeting glycoproteins of herpes simplex virus type 2 using an immunoinformatics approach. Bioactivities. 3(1): 55–71. https://doi.org/10.47352/bioactivities.2963-654X.280

Afnani, Moh. R., Emilia, N.F., Nurlaili, E., Rovik, A., Ansori, A.N.M., Parikesit, A.A., Chatterjee, S., Mahata, J., Firdous, A., Sahadewa, S., et al. 2026. In silico design of a multi-epitope rabies vaccine candidate incorporating African HLA diversity: A reverse vaccinology approach. Narra X. 4(1): e263. https://doi.org/10.52225/narrax.v4i1.263

Alnuqaydan, A.M. and Eisa, A.A. 2024. Targeting polyprotein to design potential multiepitope vaccine against omsk hemorrhagic fever virus (OHFV) by evaluating allergenicity, antigenicity, and toxicity using immunoinformatic approaches. Biology. 13(9): 738. https://doi.org/10.3390/biology13090738

Arumugam, S. and Varamballi, P. 2021. In-silico design of envelope based multi-epitope vaccine candidate against Kyasanur forest disease virus. Scientific Reports. 11(1): 17118. https://doi.org/10.1038/s41598-021-94488-8

Bechelli, J., Smalley, C., Zhao, X., Judy, B., Valdes, P., Walker, D.H., and Fang, R. 2016. MyD88 mediates instructive signaling in dendritic cells and protective inflammatory response during Rickettsial infection. Infection and Immunity. 84(4): 883–893. https://doi.org/10.1128/IAI.01361-15

Booth, W.T., Schlachter, C.R., Pote, S., Ussin, N., Mank, N.J., Klapper, V., Offermann, L.R., Tang, C., Hurlburt, B.K., and Chruszcz, M. 2018. Impact of an N-terminal polyhistidine tag on protein thermal stability. ACS Omega. 3(1): 760–768. https://doi.org/10.1021/acsomega.7b01598

Bouzid, R.S., Bouzid, R., Labed, H., Serhani, I., Hellal, D., Oumeddour, L., Boudhiaf, I., Ibrir, M., Khadraoui, H., and Belaaloui, G. 2025. Molecular subtyping and target identification in triple negative breast cancer through immunohistochemistry biomarkers. BMC Cancer. 25(1): 454. https://doi.org/10.1186/s12885-025-13832-7

Brown, T.A., Mittendorf, E.A., Hale, D.F., Myers, J.W., Peace, K.M., Jackson, D.O., Greene, J.M., Vreeland, T.J., Clifton, G.T., Ardavanis, A., et al. 2020. Prospective, randomized, single-blinded, multi-center phase II trial of two HER2 peptide vaccines, GP2 and AE37, in breast cancer patients to prevent recurrence. Breast Cancer Research and Treatment. 181(2): 391–401. https://doi.org/10.1007/s10549-020-05638-x

Buchan, D.W.A., Minneci, F., Nugent, T.C.O., Bryson, K., and Jones, D.T. 2013. Scalable web services for the PSIPRED protein analysis workbench. Nucleic Acids Research. 41(W1): W349–W357. https://doi.org/10.1093/nar/gkt381

Chandrashekar, D.S., Bashel, B., Balasubramanya, S.A.H., Creighton, C.J., Ponce-Rodriguez, I., Chakravarthi, B.V.S.K., and Varambally, S. 2017. UALCAN: A portal for facilitating tumor subgroup gene expression and survival analyses. Neoplasia. 19(8): 649–658. https://doi.org/10.1016/j.neo.2017.05.002

Chandrashekar, D.S., Karthikeyan, S.K., Korla, P.K., Patel, H., Shovon, A.R., Athar, M., Netto, G J., Qin, Z.S., Kumar, S., Manne, U., et al. 2022. UALCAN: An update to the integrated cancer data analysis platform. Neoplasia. 25: 18–27. https://doi.org/10.1016/j.neo.2022.01.001

Chansap, S., Cheukamud, W., Suthisintong, T., Kueakhai, P., and Changklungmoa, N. 2025. Development of multi-epitope cathepsin L driven short peptide vaccine against Fasciola gigantica. Frontiers in Veterinary Science. 12: 1547937. https://doi.org/10.3389/fvets.2025.1547937

Chen, L., Zheng, L., Chen, P., and Liang, G. 2020. Myeloid differentiation primary response protein 88 (MyD88): The central hub of TLR/IL-1R signaling. Journal of Medicinal Chemistry. 63(22): 13316–13329. https://doi.org/10.1021/acs.jmedchem.0c00884

Cheng, P., Xue, Y., Wang, J., Jia, Z., Wang, L., and Gong, W. 2022. Evaluation of the consistence between the results of immunoinformatics predictions and real-world animal experiments of a new tuberculosis vaccine MP3RT. Frontiers in Cellular and Infection Microbiology. 12: 1047306. https://doi.org/10.3389/fcimb.2022.1047306

Chinthakunta, N., Cheemanapalli, S., Chinthakunta, S., Anuradha, C.M., and Chitta, S.K. 2018. A new insight into identification of in silico analysis of natural compounds targeting GPR120. Network Modeling Analysis in Health Informatics and Bioinformatics. 7(1): 8. https://doi.org/10.1007/s13721-018-0166-0

Craig, D.B. and Dombkowski, A.A. 2013. Disulfide by design 2.0: A web-based tool for disulfide engineering in proteins. BMC Bioinformatics. 14: 346. https://doi.org/10.1186/1471-2105-14-346

Curigliano, G., Romieu, G., Campone, M., Dorval, T., Duck, L., Canon, J.L., Roemer-Becuwe, C., Roselli, M., Neciosup, S., et al. 2016. A phase I/II trial of the safety and clinical activity of a HER2-protein based immunotherapeutic for treating women with HER2-positive metastatic breast cancer. Breast Cancer Research and Treatment. 156(2): 301–310. https://doi.org/10.1007/s10549-016-3750-y

Desai, D., Kantliwala, S.V, Vybhavi, J., Ravi, R., Patel, H., and Patel, J. 2024. Review of AlphaFold 3: Transformative advances in drug design and therapeutics. Cureus. 16(7): e63646. https://doi.org/10.7759/cureus.63646

Dhanda, S.K., Vir, P., and Raghava, G.P. 2013. Designing of interferon-gamma inducing MHC class-II binders. Biology Direct. 8(1): 30. https://doi.org/10.1186/1745-6150-8-30

Dimitrov, I., Bangov, I., Flower, D.R., and Doytchinova, I. 2014. AllerTOP v.2—A server for in silico prediction of allergens. Journal of Molecular Modeling. 20(6): 2278. https://doi.org/10.1007/s00894-014-2278-5

Dimitrov, I., Flower, D.R., and Doytchinova, I. 2013. AllerTOP - a server for in silico prediction of allergens. BMC Bioinformatics. 14(SUPPL6): S4. https://doi.org/10.1186/1471-2105-14-S6-S4

Dombkowski, A.A. 2003. Disulfide by design: A computational method for the rational design of disulfide bonds in proteins. Bioinformatics. 19(14): 1852–1853. https://doi.org/10.1093/bioinformatics/btg231

Dong, R., Chu, Z., Yu, F., and Zha, Y. 2020. Contriving multi-epitope subunit of vaccine for COVID-19: Immunoinformatics approaches. Frontiers in Immunology. 11: 1784. https://doi.org/10.3389/fimmu.2020.01784

Doytchinova, I.A. and Flower, D.R. 2007. VaxiJen: A server for prediction of protective antigens, tumour antigens and subunit vaccines. BMC Bioinformatics. 8(1): 4. https://doi.org/10.1186/1471-2105-8-4

Elalouf, A. 2023. In-silico structural modeling of human immunodeficiency virus proteins. Biomedical Engineering and Computational Biology. 14: 11795972231154402. https://doi.org/10.1177/11795972231154402

Elfadil, M.M., Samhoon, S.O.A., Saadaldin, M.M., Ibrahim, S.A.E., Mohamed, A.A.M., Suliman, O.H., Mohamed, O., Damiri, N., Firdaus-Raih, M., Mohamed, S.B., et al. 2025. Reverse vaccinology and immunoinformatics approaches for multi-epitope vaccine design against Klebsiella pneumoniae reveal a novel vaccine target protein. Journal of Genetic Engineering and Biotechnology. 23(3): 100510. https://doi.org/10.1016/j.jgeb.2025.100510

Emadi, S.H., Ahmad, S., Rahmanian, M., Baharlou, R., Ahmad, F., Aghaamoo, S., and Sanami, S. 2026. Design and computational evaluation of a prophylactic and therapeutic multi-epitope vaccine candidate against cervical cancer. Virology Journal. 23: 45. https://doi.org/10.1186/s12985-026-03072-x

Fatoba, A.J., Adeleke, V.T., Maharaj, L., Okpeku, M., Adeniyi, A.A., and Adeleke, M.A. 2021. Immunoinformatics design of multiepitope vaccine against Enterococcus faecium infection. International Journal of Peptide Research and Therapeutics. 27(4): 2183–2198. https://doi.org/10.1007/s10989-021-10245-5

Grote, A., Hiller, K., Scheer, M., Munch, R., Nortemann, B., Hempel, D.C., and Jahn, D. 2005. JCat: A novel tool to adapt codon usage of a target gene to its potential expression host. Nucleic Acids Research. 33(Web Server): W526–W531. https://doi.org/10.1093/nar/gki376

Gupta, S., Kapoor, P., Chaudhary, K., Gautam, A., Kumar, R., and Raghava, G.P.S. 2013. In silico approach for predicting toxicity of peptides and proteins. PLoS One. 8(9): e73957. https://doi.org/10.1371/journal.pone.0073957

Gustiananda, M., Sulistyo, B.P., Agustriawan, D., and Andarini, S. 2021. Immunoinformatics analysis of SARS-CoV-2 ORF1ab polyproteins to identify promiscuous and highly conserved T-cell epitopes to formulate vaccine for Indonesia and the world population. Vaccines. 9(12): 1459. https://doi.org/10.3390/vaccines9121459

Heery, C.R., Ibrahim, N.K., Arlen, P.M., Mohebtash, M., Murray, J L., Koenig, K., Madan, R.A., McMahon, S., Marte, J.L., Steinberg, S.M., et al. 2015. Docetaxel alone or in combination with a therapeutic cancer vaccine (PANVAC) in patients with metastatic breast cancer: A randomized clinical trial. JAMA Oncology. 1(8): 1087–1095. https://doi.org/10.1001/jamaoncol.2015.2736

Hossan, M.I., Chowdhury, A.S., Hossain, M.U., Khan, M.A., Mahmood, T.B., and Mizan, S. 2021. Immunoinformatics aided-design of novel multi-epitope based peptide vaccine against Hendra henipavirus through proteome exploration. Informatics in Medicine Unlocked. 25: 100678. https://doi.org/10.1016/j.imu.2021.100678

Jensen, K.K., Andreatta, M., Marcatili, P., Buus, S., Greenbaum, J.A., Yan, Z., Sette, A., Peters, B., and Nielsen, M. 2018. Improved methods for predicting peptide binding affinity to MHC class II molecules. Immunology. 154(3): 394–406. https://doi.org/10.1111/imm.12889

Jespersen, M.C., Peters, B., Nielsen, M., and Marcatili, P. 2017. BepiPred-2.0: Improving sequence-based B-cell epitope prediction using conformational epitopes. Nucleic Acids Research. 45(W1): W24–W29. https://doi.org/10.1093/nar/gkx346

Kanse, S., Khandelwal, M., Pandey, R.K., Khokhar, M., Desai, N., and Kumbhar, B.V. 2023. Designing a multi-epitope subunit vaccine against VP1 major coat protein of JC polyomavirus. Vaccines. 11(7): 1182. https://doi.org/10.3390/vaccines11071182

Kar, T., Narsaria, U., Basak, S., Deb, D., Castiglione, F., Mueller, D.M., and Srivastava, A.P. 2020. A candidate multi-epitope vaccine against SARS-CoV-2. Scientific Reports. 10(1): 10895. https://doi.org/10.1038/s41598-020-67749-1

Khatoon, N., Pandey, R.K., and Prajapati, V.K. 2017. Exploring Leishmania secretory proteins to design B and T cell multi-epitope subunit vaccine using immunoinformatics approach. Scientific Reports. 7(1): 8285. https://doi.org/10.1038/s41598-017-08842-w

Krishnamoorthy, H.R. and Karuppasamy, R. 2023. Design and in silico validation of a novel MZF-1-based multi-epitope vaccine to combat metastatic triple negative breast cancer. Vaccines. 11(3): 577. https://doi.org/10.3390/vaccines11030577

Kurcinski, M., Oleniecki, T., Ciemny, M.P., Kuriata, A., Kolinski, A., and Kmiecik, S. 2019. CABS-flex standalone: A simulation environment for fast modeling of protein flexibility. Bioinformatics. 35(4): 694–695. https://doi.org/10.1093/bioinformatics/bty685

Kuriata, A., Gierut, A.M., Oleniecki, T., Ciemny, M.P., Kolinski, A., Kurcinski, M., and Kmiecik, S. 2018. CABS-flex 2.0: A web server for fast simulations of flexibility of protein structures. Nucleic Acids Research. 46(W1): W338–W343. https://doi.org/10.1093/nar/gky356

Laskowski, R.A. and Thornton, J.M. 2022. PDBsum extras: SARS‐CoV‐2 and AlphaFold models. Protein Science. 31(1): 283–289. https://doi.org/10.1002/pro.4238

Li, H., Liu, H., Chen, K., Feng, W., and Guo, J. 2014. HSV-2 increases TLR4-dependent phosphorylated IRFs and IFN-beta induction in cervical epithelial cells (INM9P.460). The Journal of Immunology, 192(Supplement_1): 189.13. https://doi.org/10.4049/jimmunol.192.Supp.189.13

Limentani, S.A., Campone, M., Dorval, T., Curigliano, G., deBoer, R., Vogel, C., White, S., Bachelot, T., Canon, J.L., Disis, M., et al. 2016. A non-randomized dose-escalation phase I trial of a protein-based immunotherapeutic for the treatment of breast cancer patients with HER2-overexpressing tumors. Breast Cancer Research and Treatment. 156(2): 319–330. https://doi.org/10.1007/s10549-016-3751-x

López-Blanco, J.R., Aliaga, J.I., Quintana-Ortí, E.S., and Chacón, P. 2014. iMODS: Internal coordinates normal mode analysis server. Nucleic Acids Research. 42(W1): W271–W276. https://doi.org/10.1093/nar/gku339

Mamun, T.I., Ali, Md. A., Hosen, Md. N., Rahman, J., Islam, Md. A., Akib, Md. G., Zaman, K., Rahman, Md. M., Hossain, F.M.A., Ibenmoussa, S., et al. 2025. Designing a multi-epitope vaccine candidate against human rhinovirus C utilizing immunoinformatics approach. Frontiers in Immunology. 15: 1364129. https://doi.org/10.3389/fimmu.2024.1364129

Martinelli, D.D. 2022. In silico vaccine design: A tutorial in immunoinformatics. Healthcare Analytics. 2: 100044. https://doi.org/10.1016/j.health.2022.100044

McGuffin, L.J., Bryson, K., and Jones, D.T. 2000. The PSIPRED protein structure prediction server. Bioinformatics. 16(4): 404–405. https://doi.org/10.1093/bioinformatics/16.4.404

Mia, M.M., Hasan, M., Ahmed, S., and Rahman, M.N. 2022. Insight into the first multi-epitope-based peptide subunit vaccine against avian influenza A virus (H5N6): An immunoinformatics approach. Infection, Genetics and Evolution. 104: 105355. https://doi.org/10.1016/j.meegid.2022.105355

Naderian, R., Ahmad, S., Rahmanian, M., Aghaamoo, S., Rahbar, A., Pajand, O., Alizadeh, A., Nazarian, S., Sanami, S., and Eslami, M. 2025. Rational design of multi-epitope vaccine for Chandipura virus using an immunoinformatics approach. PLoS One. 20(10): e0335147. https://doi.org/10.1371/journal.pone.0335147

Ng, B., Puspitaningtyas, H., Wiranata, J.A., Hutajulu, S.H., Widodo, I., Anggorowati, N., Sanjaya, G.Y., Lazuardi, L., and Sripan, P. 2023. Breast cancer incidence in Yogyakarta, Indonesia from 2008–2019: A cross-sectional study using trend analysis and geographical information system. PLoS One. 18(7): e0288073. https://doi.org/10.1371/journal.pone.0288073

Nirob, Md. N.H., Sultana, I., al Arian, T., Islam, MD.S., Moniruzzzaman, M., Nosi, S.J., Halder, S.K., Shil, A., and Himel, M.K. 2025. A multi-epitope vaccine targeting a key virulence factor ClfA: An in-silico approach to combat Staphylococcus aureus infections. PLoS One. 20(10): e0334885. https://doi.org/10.1371/journal.pone.0334885

Nshimiyimana, S., Kimotho, J., and Ngugi, C.W. 2024. Immunoinformatics approach to develop a novel chimeric L1/L2 mRNA-based vaccine against a broad spectrum of human papillomavirus (HPV) species. World's Veterinary Journal. 15(2): 341-374. https://doi.org/10.20944/preprints202410.1416.v1

O’Leary, N.A., Wright, M.W., Brister, J.R., Ciufo, S., Haddad, D., McVeigh, R., Rajput, B., Robbertse, B., Smith-White, B., Ako-Adjei, D., et al. 2016. Reference sequence (RefSeq) database at NCBI: Current status, taxonomic expansion, and functional annotation. Nucleic Acids Research. 44(D1): D733–D745. https://doi.org/10.1093/nar/gkv1189

Oluwagbemi, O.O., Oladipo, E.K., Dairo, E.O., Ayeni, A.E., Irewolede, B.A., Jimah, E.M., Oyewole, M.P., Olawale, B.M., Adegoke, H.M., and Ogunleye, A.J. 2022. Computational construction of a glycoprotein multi-epitope subunit vaccine candidate for old and new South-African SARS-CoV-2 virus strains. Informatics in Medicine Unlocked. 28: 100845. https://doi.org/10.1016/j.imu.2022.100845

Omoniyi, A.A., Adebisi, S.S., Musa, S.A., Nzalak, J.O., Bauchi, Z.M., Bako, K.W., Olatomide, O.D., Zachariah, R., and Nyengaard, J.R. 2022. In silico design and analyses of a multi-epitope vaccine against Crimean-Congo hemorrhagic fever virus through reverse vaccinology and immunoinformatics approaches. Scientific Reports. 12(1): 8736. https://doi.org/10.1038/s41598-022-12651-1

Parvin, R., Masum, M.H.U., Ferdous, J., Mahdeen, A.A., and Khan, M.S.I. 2024. Designing of a chimeric multiepitope vaccine against bancroftian lymphatic filariasis through immunoinformatics approaches. PLoS One. 19(9): e0310398. https://doi.org/10.1371/journal.pone.0310398 

Prawiningrum, A.F., Paramita, R.I., and Panigoro, S.S. 2022. Immunoinformatics approach for epitope-based vaccine design: Key steps for breast cancer vaccine. Diagnostics. 12(12): 2981. https://doi.org/10.3390/diagnostics12122981

Priyamvada, P. and Ramaiah, S. 2023. Pan-genome and reverse vaccinology approaches to design multi-epitope vaccine against Epstein-Barr virus associated with colorectal cancer. Immunologic Research. 71(6): 887–908. https://doi.org/10.1007/s12026-023-09403-2

Rao, S., Sayeeda, M., Prakash, T., Ap, P., Imran, S., and Ravi, L. 2020. Construction of computational 3D structures of protein drug targets of Mycobacterium tuberculosis. Asian Journal of Pharmaceutical and Clinical Research. 13(11): 82–85. https://doi.org/10.22159/ajpcr.2020.v13i11.39242

Rapin, N., Lund, O., and Castiglione, F. 2011. Immune system simulation online. Bioinformatics. 27(14): 2013–2014. https://doi.org/10.1093/bioinformatics/btr335

Rathore, A.S., Choudhury, S., Arora, A., Tijare, P., and Raghava, G.P.S. 2024. ToxinPred 3.0: An improved method for predicting the toxicity of peptides. Computers in Biology and Medicine. 179: 108926. https://doi.org/10.1016/j.compbiomed.2024.108926

Reynisson, B., Alvarez, B., Paul, S., Peters, B., and Nielsen, M. 2021. NetMHCpan-4.1 and NetMHCIIpan-4.0: Improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data. Nucleic Acids Research. 48(W1): W449–W454. https://doi.org/10.1093/NAR/GKAA379

Rovik, A., Andyra, V.U., and Afifah, L. 2024. In silico exploration of suruhan leaves (Peperomia pellucida) for triple-negative breast cancer therapy targeting AURKA. Journal of Experimental and Clinical Pharmacy (JECP). 4(2): 115. https://doi.org/10.52365/jecp.v4i2.1119

Rubach, P., Sikora, M., Jarmolinska, A.I., Perlinska, A.P., and Sulkowska, J.I. 2024. AlphaKnot 2.0: A web server for the visualization of proteins’ knotting and a database of knotted AlphaFold-predicted models. Nucleic Acids Research. 52(W1): W187–W193. https://doi.org/10.1093/nar/gkae443

Sahin, U., Derhovanessian, E., Miller, M., Kloke, B.P., Simon, P., Lower, M., Bukur, V., Tadmor, A.D., Luxemburger, U., et al. 2017. Personalized RNA mutanome vaccines mobilize poly-specific therapeutic immunity against cancer. Nature. 547(7662): 222–226. https://doi.org/10.1038/nature23003

Sarvmeili, J., Baghban Kohnehrouz, B., Gholizadeh, A., Shanehbandi, D., and Ofoghi, H. 2024. Immunoinformatics design of a structural proteins driven multi-epitope candidate vaccine against different SARS-CoV-2 variants based on fynomer. Scientific Reports. 14(1): 10297. https://doi.org/10.1038/s41598-024-61025-2

Schoch, C.L., Ciufo, S., Domrachev, M., Hotton, C.L., Kannan, S., Khovanskaya, R., Leipe, D., Mcveigh, R., O’Neill, K., Robbertse, B., et al. 2020. NCBI taxonomy: A comprehensive update on curation, resources and tools. Database. 2020: baaa062. https://doi.org/10.1093/database/baaa062

Shamakhi, A. and Kordbacheh, E. 2021. Immunoinformatic design of an epitope-based immunogen candidate against Bacillus anthracis. Informatics in Medicine Unlocked. 24: 100574. https://doi.org/10.1016/j.imu.2021.100574

T, D., Kamaraj, B., Vasudevan, K., Gopikrishnan, M., Dassegowda, K.R., Rambabu, M., and Doss, C.G.P. 2023. Structural immunoinformatics approach for rational design of a multi-epitope vaccine against triple negative breast cancer. International Journal of Biological Macromolecules. 243: 125209. https://doi.org/10.1016/j.ijbiomac.2023.125209

Tahir ul Qamar, M., Rehman, A., Tusleem, K., Ashfaq, U.A., Qasim, M., Zhu, X., Fatima, I., Shahid, F., and Chen, L.L. 2020. Designing of a next generation multiepitope based vaccine (MEV) against SARS-COV-2: Immunoinformatics and in silico approaches. PLoS One. 15(12): 1–25. https://doi.org/10.1371/journal.pone.0244176

Tiriveedhi, V., Tucker, N., Herndon, J., Li, L., Sturmoski, M., Ellis, M., Ma, C., Naughton, M., Lockhart, A.C., Gao, F., et al. 2014. Safety and preliminary evidence of biologic efficacy of a mammaglobin-A DNA vaccine in patients with stable metastatic breast cancer. Clinical Cancer Research. 20(23): 5964–5975. https://doi.org/10.1158/1078-0432.CCR-14-0059 

Ullah, H., Ullah, S., Li, J., Yang, F., and Tan, L. 2024. An in silico design of a vaccine against all serotypes of the dengue virus based on virtual screening of B-cell and T-cell epitopes. Biology. 13(9): 681. https://doi.org/10.3390/biology13090681

Wang, Q., Lv, L., Tang, Y., Zhang, L., and Wang, L. 2018. MMP‑1 is overexpressed in triple‑negative breast cancer tissues and the knockdown of MMP‑1 expression inhibits tumor cell malignant behaviors in vitro. Oncology Letters. 17(2): 1732-1740. https://doi.org/10.3892/ol.2018.9779

Wiederstein, M. and Sippl, M.J. 2007. ProSA-web: Interactive web service for the recognition of errors in three-dimensional structures of proteins. Nucleic Acids Research. 35: W407–W410. https://doi.org/10.1093/nar/gkm290

Xiong, N., Wu, H., and Yu, Z. 2024. Advancements and challenges in triple-negative breast cancer: A comprehensive review of therapeutic and diagnostic strategies. Frontiers in Oncology. 14: 1405491. https://doi.org/10.3389/fonc.2024.1405491

Yan, Y., Tao, H., He, J., and Huang, S.Y. 2020. The HDOCK server for integrated protein–protein docking. Nature Protocols. 15(5): 1829–1852. https://doi.org/10.1038/s41596-020-0312-x

Yan, Z., Kim, K., Kim, H., Ha, B., Gambiez, A., Bennett, J., de Almeida Mendes, M.F., Trevizani, R., Mahita, J., Richardson, E., et al. 2024. Next-generation IEDB tools: A platform for epitope prediction and analysis. Nucleic Acids Research. 52(W1): W526–W532. https://doi.org/10.1093/nar/gkae407

Yulian, E.D. and Fachriza, I. 2024. Five-year survival of triple-negative breast cancer and the associated clinicopathological factors: A study in an Indonesian tertiary hospital. Medical Journal of Indonesia. 32(4): 230–237. https://doi.org/10.13181/mji.oa.247161

Zahraei, M., Roohparvar Basmenj, E., Behrouzi, G., Heidari Keshel, S., and Alem, M. 2025. In silico design of a multi-epitope vaccine against the triple negative breast cancer. Scientific Reports. 15(1): 42425. https://doi.org/10.1038/s41598-025-26511-1

Zaib, S., Rana, N., Areeba, Hussain, N., Alrbyawi, H., Dera, A.A., Khan, I., Khalid, M., Khan, A., and Al-Harrasi, A. 2023. Designing multi-epitope monkeypox virus-specific vaccine using immunoinformatics approach. Journal of Infection and Public Health. 16(1): 107–116. https://doi.org/10.1016/j.jiph.2022.11.033 

Zamani, P., Teymouri, M., Nikpoor, A.R., Navashenaq, J.G., Gholizadeh, A., Darban, S.A., and Jaafari, M.R. 2020. Nanoliposomal vaccine containing long multi-epitope peptide E75-AE36 pulsed PADRE-induced effective immune response in mice TUBO model of breast cancer. European Journal of Cancer. 129: 80–96. https://doi.org/10.1016/j.ejca.2020.01.010

Zhao, X., Tatapudi, H., Corey, G., and Gopalappa, C. 2021. Threshold analyses on combinations of testing, population size, and vaccine coverage for COVID-19 control in a university setting. PLoS One. 16(8): e0255864. https://doi.org/10.1371/journal.pone.0255864

Zhao, Y., Guo, S., Liu, J., Wang, Y., Wang, B., Peng, C., and Du, E. 2025. Adjuvant-free, self−assembling ferritin nanoparticle vaccine coupled with influenza virus hemagglutinin protein carrying M1 and PADRE epitopes elicits cross-protective immune responses. Frontiers in Immunology. 16: 1519866. https://doi.org/10.3389/fimmu.2025.1519866

Zhou, P., Shi, X., Xia, J., and Hu, H. 2025. In silico design and in vitro validation of a multi-epitope peptide vaccine targeting triple-negative breast cancer. Frontiers in Oncology. 15: 1611991. https://doi.org/10.3389/fonc.2025.1611991

Zhu, F., Tan, C., Li, C., Ma, S., Wen, H., Yang, H., Rao, M., Zhang, P., Peng, W., Cui, Y., et al. 2023. Design of a multi-epitope vaccine against six Nocardia species based on reverse vaccinology combined with immunoinformatics. Frontiers in Immunology. 14: 1100188. https://doi.org/10.3389/fimmu.2023.1100188

Zubair, S., Parvaiz, F., Abualait, T., Al-Regaiey, K., Anwar, T., Zafar, M., Kaleem, I., and Bashir, S. 2025. Computational design of multi-epitope vaccine against Hepatitis C virus infection using immunoinformatics techniques. PLoS One. 20(1): e0317520. https://doi.org/10.1371/journal.pone.0317520        

OPEN access freely available online

Natural and Life Sciences Communications

Chiang Mai University, Thailand. https://cmuj.cmu.ac.th