Genome-wide structural modelling of TCR-pMHC interactions
© Liu et al.; licensee BioMed Central Ltd. 2013
Published: 16 October 2013
The adaptive immune response is antigen-specific and triggered by pathogen recognition through T cells. Although the interactions and mechanisms of TCR-peptide-MHC (TCR-pMHC) have been studied over three decades, the biological basis for these processes remains controversial. As an increasing number of high-throughput binding epitopes and available TCR-pMHC complex structures, a fast genome-wide structural modelling of TCR-pMHC interactions is an emergent task for understanding immune interactions and developing peptide vaccines.
We first constructed the PPI matrices and iMatrix, using 621 non-redundant PPI interfaces and 398 non-redundant antigen-antibody interfaces, respectively, for modelling the MHC-peptide and TCR-peptide interfaces, respectively. The iMatrix consists of four knowledge-based scoring matrices to evaluate the hydrogen bonds and van der Waals forces between sidechains or backbones, respectively. The predicted energies of iMatrix are high correlated (Pearson's correlation coefficient is 0.6) to 70 experimental free energies on antigen-antibody interfaces. To further investigate iMatrix and PPI matrices, we inferred the 701,897 potential peptide antigens with significant statistic from 389 pathogen genomes and modelled the TCR-pMHC interactions using available TCR-pMHC complex structures. These identified peptide antigens keep hydrogen-bond energies and consensus interactions and our TCR-pMHC models can provide detailed interacting models and crucial binding regions.
Experimental results demonstrate that our method can achieve high precision for predicting binding affinity and potential peptide antigens. We believe that iMatrix and our template-based method can be useful for the binding mechanisms of TCR-pMHC complexes and peptide vaccine designs.
An adaptive immune response protects an organism from the infection by identifying and killing pathogens [1, 2]. It is antigen-specific and allows for a stronger immune response after the recognition of specific "non-self" antigens by the T-cell receptor (TCR) . As an increasing number of high-throughput experiments providing available and reliable binding epitopes related to various TCRs [4–6], a systematic and fast method to search similar complexes (i.e. TCR-pMHC molecules) is an important task for understanding potential immune interactions and developing pathogen vaccines.
Since rapidly increasing three-dimensional structure complexes in Protein Data Bank (PDB), many structure-based works have been proposed to utilize physical interacting interfaces of these complexes to study protein-protein interactions [7–10], MHC-peptide interactions [11, 12], and structural systems biology [13–15]. Most of these works [7–9, 11, 12] used a scoring-based matrix to evaluate the protein-protein and MHC-peptide interface preferences. In addition, sequence-based matrix methods (e.g. SYFPEITHI , MAPPP , IEDB ) have been proposed for predicting peptide-MHC interactions.
Recently, we have proposed a template-based strategy, called PAComplex , which is the first method investigating both peptide-MHC and peptide-TCR interfaces to infer peptide antigens and homologous peptide antigens of a query. This study utilized four scoring matrices and one scoring matrix to calculate the binding scores of peptide-MHC (which is similar to protein-protein interface (PPI)) and TCR-peptide (which is similar to antigen-antibody (Ag-Ab) [20, 21]) interfaces, respectively. Our previous works showed that four scoring matrices yielded significantly higher accuracies than one scoring matrix for inferring structure-based PPIs [22, 23]. The four scoring matrices include sidechain-sidechain and sidechain-backbone van der Waals energies; and sidechain-sidechain and sidechain-backbone hydrogen-bond energies. In addition, two main factors that deteriorate the performance of PAComplex using one-matrix scores are (i) the hydrogen-bond energies and van der Waals interactions were considered as the same and (ii) the sidechain-sidechain and sidechain-backbone interactions were not discriminated. For example, we observed that the average experimental energies of the residues forming hydrogen bonds and van der Waals interactions 2.54 and 1.08, respectively, based on 70 mutated residues on Ag-Ab interfaces.
To address these issues, we proposed four-matrices scoring function to enhance one-matrix scoring function to infer the peptide antigens using TCR-pMHC complex structures. The major enhancements are as follows: 1) four scoring matrices (named iMatrix) can predict template-based binding energies of TCR to pMHC interfaces by separating the van der Waals (vdW) forces from special bonding forces; 2) iMatrix discriminates sidechain-sidechain and sidechain-backbone interactions into two matrices; 3) a fast and genomic-scale searching method for identifying peptide antigens of a template TCR-pMHC structure; 4) iMatrix highlights the critical hydrogen bonds for key interacting residues between TCR-pMHC compexes.
To validate the reliability and enlarge the number of potential antigens, we evaluate our methods on experimental free energy data and 389 complete pathogen genomes. Experimental results indicated that iMatrix can achieve a high correlation of the binding interface energies. In addition, the homologous peptide antigens derived from iMatrix have a high precision value and keep the hydrogen bonds based on template then they should be the reliable peptide antigens. The iMatrix also reveals detailed interacting models for TCR-pMHC complexes distinctively and display the mechanisms of crucial binding regions. Furthermore, the iMatrix scoring function can provide important insights into heightened immunogenicity derived from the potential peptide antigens or epitopes and can infer valuable vaccine design for clinical trials.
Overview for genome-wide structural modelling of TCR-pMHC interactions
Here, J z ≥ 4.0 is considered a significant similarity according to the statistical analysis of 17 TCR-pMHC structure complexes (i.e. TCR-peptide-HLA-A0201 complexes); 80,057 experimental peptide antigens; and ≥ 108 peptide candidates derived from 864,628 protein sequences in 389 pathogens.
Scoring function and iMatrix
Following calculation of the interaction scores (E tot ), these scores are transformed into Z-values (i.e., Z MHC and Z TCR ) of peptide-MHC and peptide-TCR interfaces using the mean and standard deviation derived from 10,000 random interfaces by mutating each peptide position. For a TCR-pMHC template collected from the Protein Data Bank (PDB) , these 10,000 random interfaces are generated by substituting with another amino acid according to the amino acid composition derived from UniProt . Finally, we computed J Z (Equation 1) of the TCR-pMHC complex.
Data set of constructing iMatrix
Therefore, we built a dataset, consists of 398 Ag-Ab interactions, to generate the iMatrix for modelling TCR-pMHC interfaces (Figure 1A and 2). We first manually collected 679 crystal structures of Ag-Ab complexes from the PDB (April 2012) at a resolution less than or equal to 3Å. The binding interfaces consist of one protein antigen and one antibody whose fragments outside of variable regions are excluded from the analysis. All protein chains were pairwise aligned to make non-redundant sequence set using BLASTClust . Finally, the 229 Ag-Ab complexes (Table S1 in Additional file 3) with 398 Ag-Ab interfaces (Table S2 in Additional file 4) were collected in this set.
Experimental free energy dataset
To further investigate the relationship between the predicted energy and experimental free energy, we collected 70 mutated residues, which are contact residues in Ag-Ab interfaces in 4 structural complexes from the ASEdb (Table S3 in Additional file 5). The Alanine Scanning Energetics database is a repository for energetics of sidechain interactions determined by alanine-scanning mutagenesis . ASEdb gives the corresponding ΔΔG value representing the change in free energy of binding upon mutation to alanine for each experimentally mutated residue.
The experimental peptide antigens derived from IEDB
To further evaluate the reliability of homologous peptide antigen derived from the iMatrix, we collected the 80,057 experimental peptides from the IEDB (January 2013) for 389 pathogens; and 17 TCR-pMHC complexes (i.e. TCR-peptide-HLA-A0201, Table S4 in Additional file 6) from the PDB. Then, we filtered 4,987 positive nonamers and 4,322 negative nonamers of TCR-peptide-HLA-A0201. Here, the definition of positive records is at least one positive measurement in T cell response or MHC binding assays; negative records are data with only negative measurements. We also prepared the H-2-Kb (Mus musculus) and H-2-Ld (Mus musculus) alleles for validation of iMatrix.
In addition, in these 389 pathogens, the vaccinia virus has the largest amount (19.7%) of experimental records in the IEDB, including 1,131 positive nonamers and 706 negative nonamers. Here, the complete genomes of vaccinia virus are 320 proteins recorded in UniProt , and we processed them into 79,157 nonamers (56,030 non-redundant nonamers). This vaccinia virus subset was used in case studies.
Results and discussion
The high scores in four scoring matrices of iMatrix are often superior frequency of interacting residue pairs. The sidechain-sidechain scoring matrices are symmetric. In sidechain-backbone matrices (e.g., Figure 2B, 2D, S2B, and S2D in Additional file 2), y-axis denotes side chain and x-axis denotes backbone. The interacting score is set to zero if the frequency of an entry (a contacted pair residue) is 0.
For vdW scoring matrices of iMatrix (Figure 2A and 2B), the scores are high when aromatic residues (i.e., Phe, Trp, and Tyr) interact to aromatic and large-sidechain residues (e.g., Met, Ile, and Arg). The result is consistent to the previous results that residues Tyr and Trp play key roles in epitopes and paratopes . Conversely, the result is different from the vdW matrices of protein-protein interactions , which the aromatic residues only prefer interacting aromatic residues (yellow blocks; Figs. S2A and S2B in Additional file 2). Additionally, the scores are low while aliphatic residues (i.e. Ala, Val, Leu, Ile, Met, and Pro) interact to the other residues (orange blocks; Figure 2A) for immune complexes. The results are significantly different from the vdW matrices of protein-protein interfaces (yellow blocks; Figure S2A in Additional file 2).
For special-bond scoring matrices (Figure 2C and 2D), the scores (blue blocks in Figure 2C) are significantly high when the residues with polar groups (i.e. Tyr, Trp, Asn, and Gln; yellow blocks) or basic residues (i.e. His, Arg, and Lys) interact to acidic residues (i.e. Asp and Glu). These results are consistent to the results of protein-protein interfaces (orange block; Figure S2C in Additional file 2).
Based on our previous researches, the template-based scoring function achieves good agreement for the binding affinity in PPIs . The novel knowledge-based matrices were derived using a general mathematical structure  from a non-redundant set of 621 3D-dimer complexes proposed by Glaser et al. . This dataset is composed of 217 heterodimers and 404 homodimers and the sequence identity is less than 30% to each other. However, the matrices may not be applied to model TCR-peptide binding because previous studies have indicated that the TCR-pMHC interface resembles Ag-Ab interactions [20, 21]. We compared the TCR-pMHC, Ag-Ab, and protein-protein interfaces and presented our observations in global and local views. The TCR-pMHC and Ag-Ab co-crystal complexes were collected from the PDB (April 2012), including 105 and 398 non-redundant interfaces, respectively. PPIs set derived from 621 non-redundant interfaces [23, 35].
Amino acid preferences
where I i represnts the numbers of the amino acid type i in the interfaces. Next, we derived the interfaces similarity by pairwise comparison using the Pearson's correlation coefficient (PCC). The PCC of 20 amino acid types between any two sets of TCR-pMHC, Ag-Ab, and protein-protein interfaces are shown in Figure 3A. Since the strong positive PCC (0.76) between TCR-pMHC and Ag-Ab interfaces, their amino acid preferences are significantly similar. However, neither TCR-pMHC nor Ag-Ab interfaces are similar to protein-protein interfaces. This result indicates that the composition of TCR-pMHC and Ag-Ab interfaces seems to resemble each other closely.
Propensities of interface sizes and hydrogen bonds
We then gathered the sizes and proportions of hydrogen bonds (H-bonds) among TCR-pMHC, Ag-Ab, and protein-protein interfaces to analyse their properties. The average numbers of interacting residue pairs of TCR-pMHC (19.7 contact pairs/interface) and Ag-Ab (40.7 contact pairs/interface) interfaces are significantly less than the one of the protein-protein interfaces (94.4 contact pairs/interface) (Figure 3B). This informs that such immune-related binding regions are small than average. Interestingly, the H-bonds proportions of TCR-pMHC interfaces (20.1%) and Ag-Ab interfaces (19.1%) are slight higher than protein-protein interfaces (14.7%). H-bonds are extremely important in biological systems and play a key role in the structure of polymers, both synthetic and natural. These results suggest that although the TCR-pMHC and Ag-Ab interfaces are short and discontinuous, H-bonds might contribute a crucial part.
Local structural alignment of binding domains
TCR and antibody are composed of six variable loops (CDRs) and have the same domain annotation (i.e. V set domains (antibody variable domain-like)) based on SCOP  database. For local analysis the binding regions, we performed a structural alignment of the functional domains in TCR and antibody using MultiProt , an efficient and accurate method for local structural pairwise and multiple alignment. Figure 3C shows that the V set domains of TCRs and antibodies share highly structural similarity (in general, RMSD ≤ 2.0 Å). Currently, it is postulated that the CDR3 loops of TCR α and β chains specifically recognize the diversity of bound peptides of pMHC  thus play a key role of TCR-pMHC binding. We observed the details of structural alignment and found that CDR3 and contact regions of TCR (Figure 3C, red loops) and antibody (Figure 3C, blue loops) were well aligned together.
Evaluation of binding affinity
In addition, iMatrix were evaluated on these 70 mutated residues to observe the correlation between experimental ΔΔG values and predicted energies. The PCC between two scoring systems (i.e. iMatrix (red) and one matrix used in PAComplex (blue)) and free energies are shown in Figure 4C. The PCC values of iMatrix and one matrix are 0.59 and 0.47, respectively. Our results show that the iMatrix which separate vdW forces, hydrogen bonds, sidechain contact, and backbone contact could have higher correlation of the binding interface energies. This result is also consistence with the ΔΔG contribution of H-bond and sidechain contact (Figure 4A and 4B). These results imply that iMatrix considering H-bond energies and highlight sidechain contact can yield the benefits to model the binding energy to gather statistics of the Ag-Ab interfaces.
Large-scale peptide antigen identification on 389 pathogens
Comparisons between iMatrix and one-matrix on 389 complete pathogen database
No. of hits (A)
No. of hits (B)
where the H-bond ratio is equal to 1 while the number of H-bond within homologous peptide is equal to the template peptide (i.e. identical H-bond). Figure 5B illustrates the ratio of peptide which H-bond ratio equal to 1 within the peptide antigen family during different joint Z-value. The ratios of peptide with identical H-bond derived from the iMatrix have significant increasing while the threshold of joint Z-value is increasing. More importantly, the homologous peptides with joint Z-value > 6 derived from iMatrix have a significantly highest value of H-bond ratio (92%; Figure 5B). According our analysis described above, the H-bonds play an important role on the free energy of interface. Therefore, these peptide antigens with joint Z-value > 6 derived from iMatrix have a high precision value (Table 1) and keep the H-bond based on template (Figure 5) should be the more reliable peptide antigen than derived from one matrix.
Homologous peptide antigens of Tax-1
Furthermore, we would like to know whether the homologous peptide antigens of Tax peptide derived from iMatrix and one-matrix are different. The amino acid composition of the homologous peptide antigens was generated by by WebLogo, which is a graphical representation of an amino acid multiple sequence alignment . The homologous peptide antigens originated in iMatrix are more than a double of the number originated in one-matrix (102 vs 46). The amino acid composition of the homologous peptide antigens iMatrix (Figure 6B) and one-matrix scoring function (Figure 6C) generating by WebLogo, which is a graphical representation of an amino acid multiple sequence alignment . Two homologous peptide antigen sets maintained the important position 5 in peptide and conserved to Tyr (red frames in Figure 6B and 6C). This result conformed to the template-based atomic binding model (Figure 6A). Interestingly, position 5 in Figure 6B preferred all polar residues (Tyr, His, and Arg), whereas position 5 appeared Phe in Figure 6C (yellow background). However, Phe in position 5 of peptide is unreasonable and causes the loss of the critical H-bond. The iMatrix corrected such inaccuracy by considering special bond energies located in sidechain or backbone. Figure 2C provides the sidechain to sidechain special bond energies (SFss ij ). According to the scores, Tyr to Asp is 7.3 (green box) and Phe to Asp is 0.0 (red box), respectively. These related results show the iMatrix reveals the interacting environment by individually evaluating binding force and locations.
The 13 positive hits which are recorded in the IEDB derived from iMatrix scoring function shows a high consensus in position 5 (red background in Figure 6D); moreover, position 5 of 6 novel homologous peptides (not discovered by one-matrix) in the red frame are exact to Tyr.
Homologous peptide antigens of NY-ESO-1
Complementarity of interactions within a vdW network
We have developed the iMatrix, PPI-scoring matrices and a template-based approach for modelling of TCR-pMHC interactions in a genome-wide scale. Our scoring matrices, including four knowledge-based scoring matrixes, are able to identify the significant hydrogen bonds and stacking interactions in the both TCR-peptide and MHC-peptide interfaces. Experimental results demonstrate that these matrices can yield high precisions of binding affinity and infer homologous peptide antigens of a template TCR-pMHC structure on 389 pathogen genomes. In addition, our structural TCR-pMHC models can provide detailed interacting models and crucial binding regions. We believe that our scoring matrixes and template-based method are able to provide biological insights and binding mechanisms of TCR-pMHC and to reveal the immune reactions for peptide vaccine designs.
This paper was supported by National Science Council, partial supports of Ministry of Education and National Health Research Institutes (NHRI-EX100-10009PI). This paper is also particularly supported by "Center for Bioinformatics Research of Aiming for the Top University Program" of the National Chiao Tung University and Ministry of Education, Taiwan. We also thank Core Facility for Protein Structural Analysis supported by National Core Facility Program for Biotechnology.
Publication of this article was funded by the "Program for Interdisciplinary Research Project in Bioinformatics" of National Science Council.
This article has been published as part of BMC Genomics Volume 14 Supplement 5, 2013: Twelfth International Conference on Bioinformatics (InCoB2013): Computational biology. The full contents of the supplement are available online at http://www.biomedcentral.com/bmcgenomics/supplements/14/S5.
- Marrack P, Scott-Browne JP, Dai S, Gapin L, Kappler JW: Evolutionarily conserved amino acids that control TCR-MHC interaction. Annu Rev Immunol. 2008, 26: 171-203. 10.1146/annurev.immunol.26.021607.090421.PubMedPubMed CentralView ArticleGoogle Scholar
- Tikhonova AN, Van Laethem F, Hanada K, Lu JH, Pobezinsky LA, Hong CW, Guinter TI, Jeurling SK, Bernhardt G, Park JH, et al: alpha beta T Cell Receptors that Do Not Undergo Major Histocompatibility Complex-Specific Thymic Selection Possess Antibody-like Recognition Specificities. Immunity. 2012, 36 (1): 79-91. 10.1016/j.immuni.2011.11.013.PubMedPubMed CentralView ArticleGoogle Scholar
- Mlecnik B, Tosolini M, Kirilovsky A, Berger A, Bindea G, Meatchi T, Bruneval P, Trajanoski Z, Fridman WH, Pages F, et al: Histopathologic-Based Prognostic Factors of Colorectal Cancers Are Associated With the State of the Local Immune Reaction. J Clin Oncol. 2011, 29 (6): 610-618. 10.1200/JCO.2010.30.5425.PubMedView ArticleGoogle Scholar
- Bhasin M, Singh H, Raghava GP: MHCBN: a comprehensive database of MHC binding and non-binding peptides. Bioinformatics. 2003, 19 (5): 665-666. 10.1093/bioinformatics/btg055.PubMedView ArticleGoogle Scholar
- Brusic V, Rudy G, Harrison LC: MHCPEP, a database of MHC-binding peptides: update 1997. Nucleic Acids Research. 1998, 26 (1): 368-371. 10.1093/nar/26.1.368.PubMedPubMed CentralView ArticleGoogle Scholar
- Vita R, Zarebski L, Greenbaum JA, Emami H, Hoof I, Salimi N, Damle R, Sette A, Peters B: The immune epitope database 2.0. Nucleic Acids Research. 2010, 38 (Database): D854-862. 10.1093/nar/gkp1004.PubMedPubMed CentralView ArticleGoogle Scholar
- Aloy P, Russell RB: Interrogating protein interaction networks through structural biology. P Natl Acad Sci USA. 2002, 99 (9): 5896-5901. 10.1073/pnas.092147999.View ArticleGoogle Scholar
- Lu L, Lu H, Skolnick J: MULTIPROSPECTOR: An algorithm for the prediction of protein-protein interactions by multimeric threading. Proteins-Structure Function and Genetics. 2002, 49 (3): 350-364. 10.1002/prot.10222.View ArticleGoogle Scholar
- Aloy P, Pichaud M, Russell RB: Protein complexes: structure prediction challenges for the 21st century. Current Opinion in Structural Biology. 2005, 15: 15-22. 10.1016/j.sbi.2005.01.012.PubMedView ArticleGoogle Scholar
- Lefranc MP, Giudicelli V, Ginestoux C, Jabado-Michaloud J, Folch G, Bellahcene F, Wu Y, Gemrot E, Brochet X, Lane J, et al: IMGT, the international ImMunoGeneTics information system. Nucleic Acids Res. 2009, 37 (Database): D1006-1012. 10.1093/nar/gkn838.PubMedPubMed CentralView ArticleGoogle Scholar
- Altuvia Y, Sette A, Sidney J, Southwood S, Margalit H: A structure-based algorithm to predict potential binding peptides to MHC molecules with hydrophobic binding pockets. Hum Immunol. 1997, 58 (1): 1-11. 10.1016/S0198-8859(97)00210-3.PubMedView ArticleGoogle Scholar
- Kumar N, Mohanty D: MODPROPEP: a program for knowledge-based modeling of protein-peptide complexes. Nucleic Acids Research. 2007, 35 (Web Server): W549-555. 10.1093/nar/gkm266.PubMedPubMed CentralView ArticleGoogle Scholar
- Gavin AC, Aloy P, Grandi P, Krause R, Boesche M, Marzioch M, Rau C, Jensen LJ, Bastuck S, Dumpelfeld B, et al: Proteome survey reveals modularity of the yeast cell machinery. Nature. 2006, 440: 631-636. 10.1038/nature04532.PubMedView ArticleGoogle Scholar
- Aloy P, Russell RB: Ten thousand interactions for the molecular biologist. Nat Biotechnol. 2004, 22: 1317-1321. 10.1038/nbt1018.PubMedView ArticleGoogle Scholar
- Aloy P, Bottcher B, Ceulemans H, Leutwein C, Mellwig C, Fischer S, Gavin AC, Bork P, Superti-Furga G, Serrano L, et al: Structure-based assembly of protein complexes in yeast. Science. 2004, 303: 2026-2029. 10.1126/science.1092645.PubMedView ArticleGoogle Scholar
- Schuler MM, Nastke MD, Stevanovikc S: SYFPEITHI: database for searching and T-cell epitope prediction. Methods Mol Biol. 2007, 409: 75-93. 10.1007/978-1-60327-118-9_5.PubMedView ArticleGoogle Scholar
- Hakenberg J, Nussbaum AK, Schild H, Rammensee HG, Kuttler C, Holzhutter HG, Kloetzel PM, Kaufmann SH, Mollenkopf HJ: MAPPP: MHC class I antigenic peptide processing prediction. Appl Bioinformatics. 2003, 2 (3): 155-158.PubMedGoogle Scholar
- Peters B, Sette A: Generating quantitative models describing the sequence specificity of biological processes with the stabilized matrix method. Bmc Bioinformatics. 2005, 6: 132-10.1186/1471-2105-6-132.PubMedPubMed CentralView ArticleGoogle Scholar
- Liu IH, Lo YS, Yang JM: PAComplex: a web server to infer peptide antigen families and binding models from TCR-pMHC complexes. Nucleic Acids Res. 2011, 39 (Web Server): W254-260.PubMedPubMed CentralView ArticleGoogle Scholar
- Rudolph MG, Luz JG, Wilson IA: Structural and thermodynamic correlates of T cell signaling. Annu Rev Biophys Biomol Struct. 2002, 31: 121-149. 10.1146/annurev.biophys.31.082901.134423.PubMedView ArticleGoogle Scholar
- Rudolph MG, Wilson IA: The specificity of TCR/pMHC interaction. Curr Opin Immunol. 2002, 14 (1): 52-65. 10.1016/S0952-7915(01)00298-9.PubMedView ArticleGoogle Scholar
- Chen YC, Lo YS, Hsu WC, Yang JM: 3D-partner: a web server to infer interacting partners and binding models. Nucleic Acids Res. 2007, 35 (Web Server): W561-567. 10.1093/nar/gkm346.PubMedPubMed CentralView ArticleGoogle Scholar
- Lo YS, Chen YC, Yang JM: 3D-interologs: an evolution database of physical protein-protein interactions across multiple genomes. BMC genomics. 2010, 11 (Suppl 3): S7-10.1186/1471-2164-11-S3-S7.PubMedPubMed CentralView ArticleGoogle Scholar
- Henikoff S, Henikoff JG: Amino acid substitution matrices from protein blocks. Proc Natl Acad Sci USA. 1992, 89 (22): 10915-10919. 10.1073/pnas.89.22.10915.PubMedPubMed CentralView ArticleGoogle Scholar
- Yu YK, Wootton JC, Altschul SF: The compositional adjustment of amino acid substitution matrices. Proc Natl Acad Sci USA. 2003, 100 (26): 15688-15693. 10.1073/pnas.2533904100.PubMedPubMed CentralView ArticleGoogle Scholar
- Yu YK, Altschul SF: The construction of amino acid substitution matrices for the comparison of proteins with non-standard compositions. Bioinformatics. 2005, 21 (7): 902-911. 10.1093/bioinformatics/bti070.PubMedView ArticleGoogle Scholar
- Thorn KS, Bogan AA: ASEdb: a database of alanine mutations and their effects on the free energy of binding in protein interactions. Bioinformatics. 2001, 17 (3): 284-285. 10.1093/bioinformatics/17.3.284.PubMedView ArticleGoogle Scholar
- Chen JL, Stewart-Jones G, Bossi G, Lissin NM, Wooldridge L, Choi EM, Held G, Dunbar PR, Esnouf RM, Sami M, et al: Structural and kinetic basis for heightened immunogenicity of T cell vaccines. The Journal of experimental medicine. 2005, 201 (8): 1243-1255. 10.1084/jem.20042323.PubMedPubMed CentralView ArticleGoogle Scholar
- The Universal Protein Resource (UniProt) in 2010. Nucleic Acids Research. 2010, 38 (Database): D142-148.Google Scholar
- Crooks GE, Hon G, Chandonia JM, Brenner SE: WebLogo: A sequence logo generator. Genome Res. 2004, 14 (6): 1188-1190. 10.1101/gr.849004.PubMedPubMed CentralView ArticleGoogle Scholar
- Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, Shindyalov IN, Bourne PE: The Protein Data Bank. Nucleic Acids Res. 2000, 28 (1): 235-242. 10.1093/nar/28.1.235.PubMedPubMed CentralView ArticleGoogle Scholar
- Eisen HN, Hou XH, Shen C, Wang K, Tanguturi VK, Smith C, Kozyrytska K, Nambiar L, McKinley CA, Chen J, et al: Promiscuous binding of extracellular peptides to cell surface class I MHC protein. Proc Natl Acad Sci USA. 2012, 109 (12): 4580-4585. 10.1073/pnas.1201586109.PubMedPubMed CentralView ArticleGoogle Scholar
- Johnson M, Zaretskaya I, Raytselis Y, Merezhuk Y, McGinnis S, Madden TL: NCBI BLAST: a better web interface. Nucleic Acids Res. 2008, 36 (Web Server): W5-9. 10.1093/nar/gkn201.PubMedPubMed CentralView ArticleGoogle Scholar
- Zhao L, Li J: Mining for the antibody-antigen interacting associations that predict the B cell epitopes. BMC Struct Biol. 2010, 10 (Suppl 1): S6-10.1186/1472-6807-10-S1-S6.PubMedPubMed CentralView ArticleGoogle Scholar
- Glaser F, Steinberg DM, Vakser IA, Ben-Tal N: Residue frequencies and pairing preferences at protein-protein interfaces. Proteins-Structure Function and Genetics. 2001, 43 (2): 89-102. 10.1002/1097-0134(20010501)43:2<89::AID-PROT1021>3.0.CO;2-H.View ArticleGoogle Scholar
- Andreeva A, Howorth D, Chandonia JM, Brenner SE, Hubbard TJ, Chothia C, Murzin AG: Data growth and its impact on the SCOP database: new developments. Nucleic Acids Res. 2008, 36 (Database): D419-425.PubMedPubMed CentralView ArticleGoogle Scholar
- Shatsky M, Nussinov R, Wolfson HJ: A method for simultaneous alignment of multiple protein structures. Proteins. 2004, 56 (1): 143-156. 10.1002/prot.10628.PubMedView ArticleGoogle Scholar
- Garcia KC, Adams JJ, Feng D, Ely LK: The molecular basis of TCR germline bias for MHC is surprisingly simple. Nat Immunol. 2009, 10 (2): 143-147. 10.1038/ni.f.219.PubMedPubMed CentralView ArticleGoogle Scholar
- Bogan AA, Thorn KS: Anatomy of hot spots in protein interfaces. Journal of Molecular Biology. 1998, 280 (1): 1-9. 10.1006/jmbi.1998.1843.PubMedView ArticleGoogle Scholar
- Roomp K, Antes I, Lengauer T: Predicting MHC class I epitopes in large datasets. Bmc Bioinformatics. 2010, 11: 90-10.1186/1471-2105-11-90.PubMedPubMed CentralView ArticleGoogle Scholar
- Kashanchi F, Brady JN: Transcriptional and post-transcriptional gene regulation of HTLV-1. Oncogene. 2005, 24 (39): 5938-5951. 10.1038/sj.onc.1208973.PubMedView ArticleGoogle Scholar
- Grassmann R, Aboud M, Jeang KT: Molecular mechanisms of cellular transformation by HTLV-1 Tax. Oncogene. 2005, 24 (39): 5976-5985. 10.1038/sj.onc.1208978.PubMedView ArticleGoogle Scholar
- Waldele K, Schneider G, Ruckes T, Grassmann R: Interleukin-13 overexpression by tax transactivation: a potential autocrine stimulus in human T-cell leukemia virus-infected lymphocytes. Journal of virology. 2004, 78 (12): 6081-6090. 10.1128/JVI.78.12.6081-6090.2004.PubMedPubMed CentralView ArticleGoogle Scholar
- Aranda B, Achuthan P, Alam-Faruque Y, Armean I, Bridge A, Derow C, Feuermann M, Ghanbarian AT, Kerrien S, Khadake J, et al: The IntAct molecular interaction database in 2010. Nucleic Acids Research. 2010, 38: D525-D531. 10.1093/nar/gkp878.PubMedPubMed CentralView ArticleGoogle Scholar
- Ding YH, Smith KJ, Garboczi DN, Utz U, Biddison WE, Wiley DC: Two human T cell receptors bind in a similar diagonal mode to the HLA-A2/Tax peptide complex using different TCR amino acids. Immunity. 1998, 8 (4): 403-411. 10.1016/S1074-7613(00)80546-4.PubMedView ArticleGoogle Scholar
- Crooks GE, Hon G, Chandonia JM, Brenner SE: WebLogo: a sequence logo generator. Genome research. 2004, 14 (6): 1188-1190. 10.1101/gr.849004.PubMedPubMed CentralView ArticleGoogle Scholar
- Chen YT, Scanlan MJ, Sahin U, Tureci O, Gure AO, Tsang S, Williamson B, Stockert E, Pfreundschuh M, Old LJ: A testicular antigen aberrantly expressed in human cancers detected by autologous antibody screening. Proc Natl Acad Sci USA. 1997, 94 (5): 1914-1918. 10.1073/pnas.94.5.1914.PubMedPubMed CentralView ArticleGoogle Scholar
- Jager E, Chen YT, Drijfhout JW, Karbach J, Ringhoffer M, Jager D, Arand M, Wada H, Noguchi Y, Stockert E, et al: Simultaneous humoral and cellular immune response against cancer-testis antigen NY-ESO-1: definition of human histocompatibility leukocyte antigen (HLA)-A2-binding peptide epitopes. The Journal of experimental medicine. 1998, 187 (2): 265-270. 10.1084/jem.187.2.265.PubMedPubMed CentralView ArticleGoogle Scholar
- Lua RC: PyKnot: a PyMOL tool for the discovery and analysis of knots in proteins. Bioinformatics. 2012, 28 (15): 2069-2071. 10.1093/bioinformatics/bts299.PubMedView ArticleGoogle Scholar
- Buslepp J, Wang H, Biddison WE, Appella E, Collins EJ: A correlation between TCR Valpha docking on MHC and CD8 dependence: implications for T cell selection. Immunity. 2003, 19 (4): 595-606. 10.1016/S1074-7613(03)00269-3.PubMedView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.