silico In analysis of cancer antigens of non-small cell lung cancer (NSCLC) for dendritic cell-based immune-gene therapy application

Hend Allaw ⁽¹⁾     Lama Youssef   ^(1,2)  Majd Aljamali  ^(1,3)

⁽¹⁾       Syrian Virtual University, ⁽²⁾ Dept. Pharmaceutics, Faculty of Pharmacy, Damascus University, ⁽³⁾ Dept. Biochemistry and Microbiology, Faculty of Pharmacy, Damascus University.

Hendallaou10@gmail.com

Abstract

One of the advanced cancer immunotherapy approaches is the development of efficient therapeutic cancer vaccines. Cancer vaccines are based on tumor antigens expressed in the context of major histocompatibility complex (MHC) molecules able to elicit a strong tumor-specific CTL response, which may result in the killing of tumor cells and cancer regression. We here propose a strategy to design a polytope cancer vaccine that has many unique characteristics. Combining different HLA-restricted epitopes from CTAs into one polytope vaccine construct allows the fusion antigen to efficiently enter the ER, then be processed and presented to MHC class I to induce the related CTL responses against all epitopes simultaneously. The goal of using multiple antigenic epitopes instead of a single antigen is to avoid the specific antigen being lost or mutated and would cover a wide range of histocompatibility complex polymorphisms.

Keywords: Cancer immunotherapy, epitopes, Major Histocompatibility Complex, Antigens, NSCLC.

Introduction

The incidence of lung cancer is high, with 2.1 million new cases and 1.8 million casualties estimated worldwide, accounting for 18.4% of all cancer cases [1]. Two major categories are discerned—small-cell lung cancer (SCLC) and non-small-cell lung cancer (NSCLC). The two most common entities within the NSCLC category are pulmonary adenocarcinoma (ADC) and pulmonary squamous cell carcinoma (SqCC), which represent about 90% of all NSCLC [2]. Presentation of specific tumor cell antigens inducing an effective immune response is one of the most important immunotherapeutic mechanisms against tumor cell immune evasion [5]. In vivo immune responses generally begin through unique functions of dendritic cells (DCs), followed by priming of naïve T cells [6]; therefore, to design a DC-based cancer immunotherapy, we can transfect DCs with nucleic acids encoding tumor specific antigens (Ags) or incubate them with tumor-specific molecules, such as proteins, peptides, or lysates [7].

Exclusively expressed tumor antigens, which are recognized by T cells, have been studied in various antitumor immunotherapies [8]. These antigens as potential immunogens are processed to short peptides that bind to MHC class I molecules and present to T-cell receptors of tumor-reactive cytotoxic T lymphocytes (CTLs). Epitope-based vaccines powerfully stimulate immune responses against immunogenic epitopes of different antigens while avoiding unknown properties of using whole gene products [8]. The goal of immunization with these peptide epitopes is to achieve therapeutic benefits; however, to prevent the escape of tumor cells, it may be advantageous to use T cell epitopes of different tumor Antigens simultaneously [9].

Some of the most interesting cancer antigens for development of cancer vaccines are cancer testis antigens or CTA. CTAs are aberrantly expressed by different tumor cell types while their normal expression is restricted to a few somatic tissues, including testis. Therefore, CTAs are primary candidates for vaccination in cancer patients [10].

NSCLC, along with melanoma and ovarian cancer, are the most frequently expressed CTAs among the various cancers analyzed [11].

Dataset:

232 CTAs from the CT database were evaluated in a data set of 199 NSCLC cases and 32 normal tissues obtained from 141 individuals. The analysis revealed 96 CTAs that were expressed in NSCLC and showed exclusive expression in testis and placenta among normal tissues. These CTAs were designated as "confirmed CTAs". The information of previous studies regarding mRNA and protein expression for each gene in NSCLC was obtained from CT database [12,13].

mRNA and protein expression profile analysis:

The protein expression of reported CTAs in NSCLC. From 232 genes, 68 were expressed in NSCLC based on the CT database. The CT database (http://www.cta.lncc.br) is a systematic data repository for CTAs, currently including 276 genes designated as CTAs, with curated information about gene and protein expression in normal and cancer tissues. Of these, 24 were described based on mRNA and protein levels, while the remaining 44 CTAs were only defined based on mRNA levels [12,14].

Using the Human Protein Atlas (HPA) image database [15], we confirmed the protein expression of 8 CTAs in NSCLC (MAGEC2, MAGEB6, PAGE2, PAGE5, PAGE2B, CT45A2, SAGE1 and MAGEA8) [12]. According to the data collected from the HPA and RNA sequence analysis, three CTAs will be chosen for the polytope chimeric syntheses. MAGEA8 (melanoma-associated antigen 8), CT45A2 (Cancer/testis antigen family 45 member A2), and SAGE1 (sarcoma antigen 1) [10].

The mRNA expression levels for the three CTAs were compared in accordance with the CT database.

Methods:

Figure 1 shows a flowchart that summarizes the project methodology steps

.

Results:

I. Design and structure of the chimeric construct

After searching for each of the CTA genes in Genebank, we obtained the sequences, and downloaded each sequence from genebank in Fasta file format [16]. Using Emboss to translate the genes for each of the proteins into amino acids sequences for epitopes prediction [17]. Using SYFPEITHI webserver tool which predicts epitopes according to HLA specific selection [18]. The scoring system evaluates every amino acid within a given peptide.  high score indicates a strong binder [19]. 

Table 1. mRNA expression levels of the three CTAs chosen for further analysis [13]

Figure 1. Flowchart of methodology used

·         Epitopes predictions:

Nonamer had been chosen in the input of SYFPEITHI server. Getting multiple epitopes prediction and using IEDB for immunogenicity score, Epitopes containing amino acids of 115-123 of MAGEA8[18,20], Epitopes containing amino acids of 841-849 of

SAGE1[18,19], and Epitopes containing amino acids of 7-15 of CTA45A2 [18,20]. (Table 2,3)

All the peptides are immunogenic and reported to be recognized by CTLs. These three peptides were used to design the chimeric construct. The epitopes were linked by Gly-Pro-Gly-Pro-Gly (GPGPG) repeats. These repeats are expected to prevent formation of junctional epitopes when the protein is cleaved during the presentation process in antigen presenting cells (APCs) [8]. It has been shown that GPGPG spacers eliminate responses against the junctional epitope, allowing the development of a balanced response [21]. To increase the accuracy and efficiency of translation in a human host, the Kozak sequence was added 5’ to the start codon. Efficient entrance and accumulation of the recombinant protein in the endoplasmic reticulum (ER) can facilitate processing of epitopes [22]; therefore, an ER signal sequence was

added at the 5’ end of the construct, and the KDEL sequence was added at the 3’ end to make it resident in the ER [22]. The HA epitope tag (YPYDVPDYA) from the human CTL influenza hemagglutinin protein was used to track the gene product in downstream assays. The HA-tag was placed 3’ to the ER signal sequence to minimize any potential disruption in tertiary structure, and thus function, of the protein. ER signal sequence was obtained from NCBI query search [8]. The structure of the chimeric gene and arrangements of fragments and linker sites are shown in Figure 2.

·         Signal Prediction:

Signal peptide was analysed for the prediction significance after constructing the polytope chimeric sequence. Phobius web server a combined transmembrane topology and signal peptide predictor was used for the signal peptide significance prediction.

·         Chimeric polytope peptide sequence:

MGMQVQIQSLFLLLLWVPGSRGYPYDVPDYAGPGPGKVAELVRFLGPGPGNYERIFILLGPGPGKVAVDPETVGPGPGKDEL

Table 2: Epitopes prediction scores via SYFPEITHI.

Table 3: Epitopes immunogenicity scores via IEDB.

Figure 2: Schematic model of the construct. The selected epitopes of MAGEA8, SAGE1 and CTA45A2 are bound together by the linkers for expression in human. These fragments were selected on the basis of HLA restriction of MHC class I, to be recognized by CTLs.

II. In silico analysis of original chimeric gene

Human codon bias was considered to design the chimeric gene. Codon bias and the GC content of the chimeric gene were analyzed. Codon Adaptation Index (CAI) of the gene is 0.90. A CAI of 1.0 is considered ideal while a CAI of >0.8 is rated as good for expression in the desired expression organism. The lower the number, the higher the chance that the gene will be expressed poorly as shown in figure 8 [8,23]. The GC content of the gene is 62.58%. The ideal percentage range of GC content is between 30% and 70%. No optimization has been conducted. Finally, HindIII and EcoRI restriction sites were introduced at the 5’ and 3’ ends of the sequence, respectively.

III. mRNA structure prediction

To verify potential folding of the chimeric mRNA, Mfold webserver for the prediction of peptide secondary structure of single strand nucleic acids was used. Using Jpred 4 for secondary structure prediction the results showed two beta sheets, and one alpha helix comprised of amino acids 18 (figure 3) [24].

IV. Prediction of secondary and tertiary structures of chimeric protein

Chimeric protein 3D models, produced by ab initio modelling using Robetta web server (figure 12) [25] and were uploaded to the Swiss-PdbViewer server to render the tertiary structural illustrations [26]. The final structure of the protein was predicted by Pymol software [27].

V. Evaluation of model stability

By minimizing the energy of a molecule, the stability of the model is confirmed.

Energy minimization was determined by analysis of 3D structural stability of the chimeric protein using Swiss-PdbViewer [8]. The energy minimization profile performed by spdbv (Swiss-PdbViewer) and calculated to be -555.381 KJ/mol [26].

This result refers to that the recombinant protein was relatively stable. Also, the structural stability of the chimeric protein was confirmed based on data generated by a Ramachandran plot (figure 15) [24].

Figure 3: secondary structure prediction Using Jpred 4 for secondary structure prediction the results

showed two beta sheets, and one alpha helix comprised of amino acids 18 [24]

Figure 4: Protein 3D structure Ab initio modelling using Robetta web server De novo models are

built using the Rosetta de novo protocol. The procedure is fully automated [25].

VI. Solvent accessibility prediction

The analysis of the fractional accessible surface area (ASA) and fractional residue volume showed that all residues have fractional volumes below 1.0.

According to VARDAR accessible surface area can be reported in square angstroms or as a fractional ASA (ranging from 0.00 to 1.00). 

Therefore, the protein is efficiently packed with no major packing defects. Stereochemical/packing quality analysis revealed that most residues have good quality scores near 8 and this showed that the protein has a high-resolution structure. 3D profile quality analysis examined local environment, packing, and hydrophobic energy for the protein structure, and the results showed an acceptable quality index [8,31].

VII. Prediction of cleavage sites

NetChop is a tool to predict cleavage sites of the human proteasome. Cleavage sites on the construct protein were analyzed with NetChop. As expected, no cleavage sites were predicted inside the linkers so the production of junctional epitopes was prevented. Also, the cleavage sites with high prediction scores were located at both ends of each selected epitope. The results are summarized in table 4 [32].

Figure 5: Evaluation of model stability, the structure stability was confirmed based on the Ramachandran plot, the dihedral angles of amino acid residues appear as crosses in the plot. The red and yellow regions represent the favored and allowed regions. The red regions correspond to conformations where there are no steric clashes in the model. These favored regions include the dihedral angles typical of the alpha-helical and beta-sheet conformations. The green areas correspond to conformations where atoms in the protein come closer than the sum of their van der Waals radii. These regions are sterically forbidden for all amino acids with side chains [31].

Table 4. Prediction of cleavage sites on the constructed protein using NetChop server.  Each amino acid in the table is the location of cleavage while no sites are located in the linkers. The threshold is 0.5 [32].

VIII. Validation of T-cell epitopes

NetCTL 1.2 server predicted CTL epitopes in the chimeric protein sequence. The server predicted CTL epitopes restricted to 12 MHC class I supertypes using ANNs [33]. The scores from the individual prediction methods were integrated, and thresholds for the integrated scores of each peptide were translated into sensitivity and specificity values (Table 5) [34].

The SYFPEITHI epitope prediction algorithm was also used. To find out the ligation strength to a defined HLA type for a sequence of amino acids. The maximum score for HLAA*0201 peptides is 36. The scores for epitopes of the chimeric protein are shown in table 5 [18].

CTLPred, is a direct method for prediction of CTL epitopes crucial in subunit vaccine design. The methods are based on elegant machine learning techniques as Artificial Neural Network ANN and Support Vector Machine SVM [35]. The scores of CTLPred predicted epitopes for the chimeric protein are shown in table 5 [36]. The default cut-off score at which the sensitivity and specificity of prediction methods are highly similar was 0.51.

IX. MHC binding peptides affinity

NetMHC 4 server Predicted peptide binding to a number of different HLA alleles using artificial neural networks (ANNs) [37]. Rank Threshold for Strong binding peptides is 0.500 and rank threshold for weak binding peptides is 2.000. The results are summarized in table 6 [38].

X. Prediction of post-translational Modifications

To predict post-translational modifications, three web-based servers were used.

NetOglyc server Find the presence of N-Glycosylation sites in human proteins.  Produced neural network predictions of mucin-type GalNAc O-glycosylation sites [39]. The NetNglyc server predicts N-Glycosylation sites in human proteins using artificial neural networks that examine the sequence context of Asn-Xaa-Ser/Thr sequences [40]. The NetPhos 3.1 server produces neural network predictions for serine, threonine and tyrosine phosphorylation sites [41]. No O-linked glycosylation were predicted. Furthermore, based on the result of the ANN method, which predicts phosphorylation sites with sensitivity in the 69% to 96% range, the construct is potentially phosphorylated at residues Ser-10, Ser-20, and Tyr-30.

Discussion:

the goal of developing efficient therapeutic cancer vaccines based on tumor antigens expressed in MHC molecules [42]. Vital questions during vaccine design include the appropriate choice of peptides, formulation, delivery mode, and monitoring of induced immune responses. The article also notes that mutated or lost immunogenic epitopes by tumors and insufficient antigen presentation by APCs are major factors in the failure of the immune system to establish effective immune responses against tumor antigens [43]. therefore, an approach that generates both Ag-specific CD4+ (Th) and CD8+ (CTL) responses may provide optimal immunization against tumors [8,44]. Tumor escape from CTL surveillance, through down regulation of individual tumor antigens and MHC alleles, might be overcome by polytope vaccines, which simultaneously target multiple cancer Antigens [45]. This strategy has advantages over using individual epitopes or intact target antigens, where individual epitopes may be lost or mutated, or where the target antigens may be oncogenic [8].  

KVAELVRFL

0.8147

NYERIFILL

1.6923

KVAVDPETV

0.7260

*Score>1.25: (sensitivity=0.54, specificity= 0.993), score>1.00: (sensitivity=0.70, specificity= 0.985), score>0.90: (sensitivity=0.74, specificity= 0.980), score>0.75: (sensitivity=0.80, specificity= 0.970), score>0.50: (sensitivity=0.89, specificity= 0.940)

Table 5: Prediction of T-cell epitopes of the construct using different web-based servers.

Table 6: Predictions of MHC-binding peptide affinity for the construct by NetMHC version 4.0. server using ANNs approximation [28].

KVAELVRFL

NYERIFILL

KVAVDPETV

Tumor escape from CTL surveillance, through down regulation of individual tumor antigens and MHC alleles, might be overcome by polytope vaccines, which simultaneously target multiple cancer Antigens [45]. This strategy has advantages over using individual epitopes or intact target antigens, where individual epitopes may be lost or mutated, or where the target antigens may be oncogenic [8]. Antigenic epitopes from diverse antigens can be linked together in a single polytope construct; such insertion of different MHC class I-restricted epitopes allows wide coverage of an MHC polymorphic population [46]. CTAs have been considered promising targets for immunotherapy approaches thanks to their tumor-specificity and strong immunogenicity for the absence of immune tolerance [45]. We describe here the strategy in the design of a polytope cancer vaccine that has many unique characteristics. Combining different HLA-restricted epitopes from CTAs into one polytope vaccine construct allows the fusion Ag to efficiently enter the ER, then be processed and presented to MHC class I to induce the related CTL responses against all epitopes simultaneously [8]. In this study, we designed a new chimeric construct of CTAs including HLA-restricted epitopes of MAGEA8, SAGE1, and CTA45A2, which contained essential determinants to be recognized by CTLs. These potential antigenic epitopes' DNA fragment was created as a chimeric construct that is best suited for expression in humans. [47].

Based on the codon usage of highly expressed nuclear-encoded human genes, the chimeric gene was developed. Complex cellular mechanisms control every stage of gene expression, from the transcription of DNA into mRNA to the folding and post-translational modification of proteins. A relationship between mRNA expression and protein solubility can now be predicted [48].

In eukaryotic cell mRNAs, the consensus sequence surrounding the start codon (Kozak seq. 5'GCCACCATGGC) can affect the precision and efficiency of translation. In the chimeric gene, the 5'GCCACC sequence was inserted 5’ to the ATG codon. The next codon following the initial methionine ATG codon, GGA, encoding Gly, and the necessary G was provided [8]. Efficient entrance and accumulation of the recombinant protein in the endoplasmic reticulum (ER) can facilitate processing of epitopes. For successful CTL induction, the antigen peptide of interest should be efficiently delivered to the MHC class I-restricted presentation pathway via direct or cross-priming. Various DNA vaccination studies have suggested that cross-priming is more efficient than direct priming while other studies indicate that direct priming is a very important process for CTL responses. Probably both processes occur following DNA vaccination and the predominant process would be determined by the experimental conditions used, including the type of construct or antigen, and the route of administration. We believe that optimization of the intracellular trafficking of expressed antigen peptide in DCs following direct transfection would be a useful approach for improving the efficacy of MHC class I-restricted presentation and subsequent CTL induction. It has been reported that the direct delivery of antigen peptide to ER improved the efficiency of CTL induction [55]. Codon Adaptation Index (CAI) of the gene is 0.90. A CAI of 1.0 is considered ideal. Moreover, The GC content of the gene is 62.58%. The ideal percentage range of GC content is between 30% and 70%. In addition, the required restriction enzyme sites were added to the ends of the designate gene for future assays [8]. The model stability was evaluated using the Ramachandran plot, which showed most residues in a stable zone. Proteasome cleavage sites were predicted to identify potential immunogenic regions in the chimeric protein. The use of GPGPG as a hydrophobic linker restricts the production of junctional epitopes, allowing efficient downstream processing of the chimeric protein [55]. The NetCTL 1.2 server predicts CTL epitopes in protein sequences and shows that the selected epitopes in the chimeric construct have high-affinity binding to MHC molecules and acceptable sensitivity and specificity to be recognized by CTLs.

Therefore, based on the prediction results, the selected epitopes of our chimeric construct also showed high-affinity binding to MHC molecules and acceptable sensitivity and specificity to be recognized by CTLs (Tables 5 and 6, respectively).

Conclusion:

We used in silico approaches to design our chimeric polytope construct of immune-gene therapy applications. We used several web servers and applications to predict different features of the construct, including GC content, secondary and tertiary structure of the protein, solvent accessibility of the chimeric protein, proteasomal cleavage site, validation of the epitope’s prediction, MHC binding affinity, and post-translational modifications. Three epitopes with high immunogenicity scores were included in the study; MAGEA8, SAGE1, and CTA45A2. Both the MAGEA8 epitope and SAGE1 epitope gave a good binding prediction. However, only the SAGE1 epitope showed a strong binding affinity with MCH molecules. For future studies, the CTA45A2 epitope could be substituted with an epitope with a better binding prediction and affinity in order to develop a more effective structural model for cancer immune-gene therapy. Considering all of these results together, this study showed potential for the rational design of multiepitope chimeric cancer vaccines using immunoinformatics and various computational methods. With the ultimate objective of developing therapeutic vaccinations for cancer patients, this study provides the foundation for further refinement and optimization of the fusion gene expression approach.

References:

1.      Cronin, K.A. et al. (2018). Annual report to the nation on the status of cancer, part I: National cancer statistics, Cancer, 124(13), pp. 2785–2800.

2.      Chen, Z. et al. (2014) “Non-small-cell lung cancers: A heterogeneous set of diseases,” Nature Reviews Cancer, 14(8), pp. 535–546.

3.      Hirsch, F.R. et al. (2008) “The prognostic and predictive role of histology in advanced non-small cell lung cancer: A literature review,” Journal of Thoracic Oncology, 3(12), pp. 1468–1481.

4.      Paech, D.C. et al. (2011) “A systematic review of the interobserver variability for histology in the differentiation between squamous and nonsquamous non-small cell lung cancer,” Journal of Thoracic Oncology, 6(1), pp. 55–63.

5.      Gholamin, M. et al. (2010) “Induction of cytotoxic T lymphocytes primed with tumor RNA-loaded dendritic cells in esophageal squamous cell carcinoma: Preliminary step for DC Vaccine Design,” BMC Cancer, 10(1).

6.      Landi, A., Babiuk, L.A. and van Drunen Littel-van den Hurk, S. (2007) “High transfection efficiency, gene expression, and viability of monocyte-derived human dendritic cells after nonviral gene transfer,” Journal of Leukocyte Biology, 82(4), pp. 849–860.

7.      Banchereau, J. and Palucka, A.K. (2005) “Dendritic cells as therapeutic vaccines against cancer,” Nature Reviews Immunology, 5(4), pp. 296–306.

8.      Forghanifard, M.M. “In silico analysis of chimeric polytope of cancer/testis antigens for dendritic cell-based immune-gene therapy applications.,” Gene Therapy and Molecular Biology , 14(2012), 87-96.

9.      Buteau, C., Markovic, S.N. and Celis, E. (2002) “Challenges in the development of effective peptide vaccines for cancer,” Mayo Clinic Proceedings, 77(4), pp. 339–349.

10.  Fratta, E. et al. (2011) “The biology of cancer testis antigens: Putative function, regulation and therapeutic potential,” Molecular Oncology, 5(2), pp. 164–182.

11.  Grunwald, C. et al. (2006) “Expression of multiple epigenetically regulated cancer/germline genes in nonsmall cell lung cancer,” International Journal of Cancer, 118(10), pp. 2522–2528.

12.  Djureinovic, D. et al. (2016) “Profiling cancer testis antigens in non–small-cell lung cancer,” JCI Insight, 1(10).

13.  CTDatabase. CTpedia. Available at: http://www.cta.lncc.br/.

14.  Almeida, L.G. et al. (2009) “CTdatabase: A knowledge-base of high-throughput and curated data on cancer-testis antigens,” Nucleic Acids Research, 37(Database).

15.  The human protein atlas.The Human Protein Atlas. Available at: https://www.proteinatlas.org/.

16.  GenBank Overview. National Center for Biotechnology Information. U.S. National Library of Medicine. Available at: https://www.ncbi.nlm.nih.gov/genbank/.

17.  Embl-Ebi. Emboss Transeq, EBI. Available at: https://www.ebi.ac.uk/Tools/st/emboss_transeq/.

18.  Syfpeithi. SYFPEITHI. Available at: http://www.syfpeithi.de/.

19.  Ip, P., Nijman, H. and Daemen, T. (2015) “Epitope prediction assays combined with validation assays strongly narrows down putative cytotoxic T lymphocyte epitopes,” Vaccines, 3(2), pp. 203–220.

20.  IEDB.org: Free epitope database and prediction resource:Available at: IEDB.org: Free epitope database and prediction resource

21.  Signori, E. et al. (2010) “DNA vaccination strategies for anti-tumour effective gene therapy protocols,” Cancer Immunology, Immunotherapy, 59(10), pp. 1583–1591.

22.  Lu, J. et al. (2004) “Multiepitope trojan antigen peptide vaccines for the induction of antitumor CTL and th immune responses,” The Journal of Immunology, 172(7), pp. 4575–4582.

23.  Gene synthesis & DNA synthesis – guaranteed delivery time | GenScript. Available at: GenScript - Make Research Easy - The leader in molecular cloning and gene synthesis, peptide synthesis, protein and antibody engineering.

24.  A protein secondary structure prediction server (no date) JPred: A Protein Secondary Structure Prediction Server. Available at: https://www.compbio.dundee.ac.uk/jpred/.

25.  Robetta web Server https://robetta.bakerlab.org/login.php?next_url=%2Fsubmit.php

26.  Swiss-PdbViewer. Swiss PDB Viewer - Home. Available at: https://spdbv.unil.ch/.

27.  Pymol is a user-sponsored molecular visualization system on an open-source foundation, maintained and distributed by schrödinger. we are happy to introducepymol 2.5! PyMOL. Available at: https://pymol.org/2/.

28.  Zuker, M. (2003) “Mfold web server for nucleic acid folding and hybridization prediction,” Nucleic Acids Research, 31(13), pp. 3406–3415.

29.  Robetta. Available at:  https://robetta.bakerlab.org/

30.  Edwards, Y.J. and Cottage, A. (2003) “Bioinformatics methods to predict protein structure and function: A practical approach,” Molecular Biotechnology, 23(2), pp. 139–166.

31.  VADAR, Available at: http://vadar.wishartlab.com/

32.  NetChop-3.1. Available at: https://services.healthtech.dtu.dk/service.php?NetChop-3.1

33.  Larsen, M.V. et al. (2007) “Large-scale validation of methods for cytotoxic T-lymphocyte epitope prediction,” BMC Bioinformatics, 8(1).

34.  .NetCTL. Available at:   https://services.healthtech.dtu.dk/service.php?NetChop-3.1

35.  Bhasin, M. and Raghava, G.P.S. (2004) “Prediction of CTL epitopes using QM, SVM and Ann Techniques,” Vaccine, 22(23-24), pp. 3195–3204.

36.  CTLPred. Available at: http://www.imtech.res.in/raghava/ctlpred/index.html

37.  Buus, S. et al. (2003) “Sensitive quantitative predictions of peptide-MHC binding by a ‘query by committee’ Artificial Neural Network Approach,” Tissue Antigens, 62(5), pp. 378–384.

38.  NetMHC. Available at: https://services.healthtech.dtu.dk/service.php?NetMHC-4.0

39.  NetOglyc 4.0. Available at:  https://services.healthtech.dtu.dk/service.php?NetOGlyc-4.0

40.  NetNglyc. Available at: https://services.healthtech.dtu.dk/service.php?NetNGlyc-1.0

41.  NetPhos 3.1. Available at: https://services.healthtech.dtu.dk/service.php?NetPhos-3.1

42.  Buonaguro, L. and Tagliamonte, M. (2020) “Selecting target antigens for cancer vaccine development,” Vaccines, 8(4), p. 615.

43.  Dutoit, V. et al. (2002) “Multiepitope CD8+ T cell response to a NY-ESO-1 peptide vaccine results in imprecise tumor targeting,” Journal of Clinical Investigation, 110(12), pp. 1813–1822.

44.  QIN, H. et al. (2005) “Specific antitumor immune response induced by a novel DNA vaccine composed of multiple CTL and T helper cell epitopes of prostate cancer associated antigens,” Immunology Letters, 99(1), pp. 85–93.

45.  Mateo, L. et al. (1999) “An HLA-A2 polyepitope vaccine for melanoma immunotherapy,” The Journal of Immunology, 163(7), pp. 4058–4063.

46.  Doan, T. et al. (2004) “A polytope DNA vaccine elicits multiple effector and memory CTL responses and protects against human papillomavirus 16 E7-expressing tumour,” Cancer Immunology, Immunotherapy, 54(2), pp. 157–171.

47.  Bar-Haim, E. et al. (2004) “Mage-A8 overexpression in transitional cell carcinoma of the bladder: Identification of two tumour-associated antigen peptides,” British Journal of Cancer, 91(2), pp. 398–407.

48.  Zhang, Y. et al. (2021) “Sage1: A potential target antigen for lung cancer T-cell immunotherapy,” Molecular Cancer Therapeutics, 20(11), pp. 2302–2313.

49.  Chen, Y.-T. et al. (2009) “Cancer/testis antigen CT45: Analysis of mrna and protein expression in human cancer,” International Journal of Cancer, 124(12), pp. 2893–2898.

50.  Amani, J. et al. (2010) “Immunogenic properties of chimeric protein from ESPA, EAE and TIR genes of escherichia coli O157:H7,” Vaccine, 28(42), pp. 6923–6929.

51.  Tartaglia, G.G. et al. (2009) “A relationship between mrna expression levels and protein solubility in E. coli,” Journal of Molecular Biology, 388(2), pp. 381–389.

52.  Wilson, J.H. and Hunt, T. (2002) Molecular biology of the cell, 4th edition: A problems approach. New York: Garland Science.

53.  Eggers, M. et al. (1995) “The cleavage preference of the proteasome governs the yield of antigenic peptides.,” Journal of Experimental Medicine, 182(6), pp. 1865–1870.

54.  RNA folding form v2.3. Available at: http://www.mfold.org/mfold/applications/rna-folding-form-v2.php

55.  Julenius, K. et al. (2004) “Prediction, conservation analysis, and structural characterization of mammalian mucin-type O-glycosylation sites,” Glycobiology, 15(2), pp. 153–164.