Bioinformatics Advance Access originally published online on November 7, 2006
Bioinformatics 2007 23(2):177-183; doi:10.1093/bioinformatics/btl563
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
In silico grouping of peptide/HLA class I complexes using structural interaction characteristics
1 Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore 8 Medical Drive, Singapore 117597
2 Institute for Infocomm Research 21 Heng Mui Keng Terrace, Singapore 119613
3 Department of Chemistry and Biomolecular Sciences & Biotechnology Research Institute, Macquarie University NSW 2109, Australia
*To whom correspondence should be addressed.
 |
ABSTRACT |
|---|
 TOP ABSTRACT 1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
Motivation: Classification of human leukocyte antigen (HLA) proteins into supertypes underpins the development of epitope-based vaccines with wide population coverage. Current methods for HLA supertype definition, based on common structural features of HLA proteins and/or their functional binding specificities, leave structural interaction characteristics among different HLA supertypes with antigenic peptides unexplored.
Methods: We describe the use of structural interaction descriptors for the analysis of 68 peptide/HLA class I crystallographic structures. Interaction parameters computed include the number of intermolecular hydrogen bonds between each HLA protein and its corresponding bound peptide, solvent accessibility, gap volume and gap index.
Results: The structural interactions patterns of peptide/HLA class I complexes investigated herein vary among individual alleles and may be grouped in a supertype dependent manner. Using the proposed methodology, eight HLA class I supertypes were defined based on existing experimental crystallographic structures which largely overlaps (77% consensus) with the definitions by binding motifs. This mode of classification, which considers conformational information of both peptide and HLA proteins, provides an alternative to the characterization of supertypes using either peptide or HLA protein information alone.
Contact: shoba{at}els.mq.edu
|
1 INTRODUCTION |
|---|
|
TOPABSTRACT1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
Major histocompatibility complex (MHC) are cell surface glycoproteins that play a vital role in adaptive immune response (Rammensee et al., 1993). In order to generate maximal immunological protection against a large repertoire of possible pathogens, MHC bind peptides of diverse sequences and present them on the surface of antigen-presenting cells for recognition by T cell receptors (Lefranc and Lefranc, 2001). T cell recognition of the peptide–MHC complex initiates a cascade of immunological events necessary for initiation and regulation of immune responses. Two classes of MHC are responsible for antigen presentation: (1) MHC class I presents endogenous peptides; and MHC class II presents exogenous peptides. The MHC binding clefts contain polymorphic cavities (or ‘pockets’) that fit the side-chains of complementary residues on the binding peptide (Falk et al., 1991a; Hunt et al., 1992). These corresponding peptide residues are termed anchor residues because they ‘anchor’ the peptides firmly at various positions in the MHC binding cleft and contribute to most of the binding interactions. Specific MHC alleles can bind peptides with similar anchor residues, leading to the definition of ‘peptide motif’ for an array of class I and class II alleles (Falk et al., 1991b; Rotzschke et al., 1991). The subsequent discovery that certain MHC alleles can recognize very similar motifs led to the definition of MHC ‘supermotifs’ or ‘supertypes’ (Del Guercio et al., 1995).
The characterization and classification of MHC alleles into supertypes is important for the development of epitope-based vaccine (Sette et al., 2001, 2002; Zhu et al., 2006). By clustering MHC alleles based on their structural features and/or peptide binding specificities, promiscuous T cell epitopes that bind multiple MHC alleles can be identified. Such peptides are key targets for the design of vaccines and immunotherapies because they are applicable to higher proportions of human population. However, experimental determination of binding specificities for even a single MHC allele is an expensive, laborious and time consuming process; and not practical for the study of MHC supertypes which involve large numbers of alleles (Kobayashi et al., 2001; Panigada et al., 2002; Doytchinova and Flower, 2005). In silico, bioinformatics is emerging as an alternative and viable approach for MHC supertype classification. A number of clustering methods for MHC supertype definitions are available, including those based on local sequence similarities in binding pockets (Chelvanayagam, 1996; Zhang et al., 1998; Zhao et al., 2003), global sequence similarities (Cano et al., 1998; McKenzie et al., 1999) and peptide binding motifs (Lund et al., 2004). For the latter, the success of the approach depends on the availability of sufficient binding data. Where data is limited or there is bias in the experimental binding motifs, mixed results have been reported (Tong et al., 2006a). Recently, Doytchinova et al. (2004) and Doytchinova and Flower (2005) employed the use of hierarchical clustering and principal component analysis to classify human MHC or human leukocyte antigen (HLA) alleles according to MHC sequences and structures. The approach successfully identified MHC class I and class II supertype fingerprints and illustrated that only 1–3 amino acid are sufficient for an allele to be classified within a particular supertype. Kangueane et al. (2005) defined critical polymorphic functional residue positions within the binding grooves of HLA-A, -B and -C alleles and grouped 47% of 295 HLA-A alleles, 44% of 540 HLA-B alleles and 35% of 156 HLA-C alleles to 36, 71 and 18 groups, respectively.
In the present study, we applied the use of structural interaction parameters previously reported as significant for peptide/MHC interactions (Kangueane et al., 2001) for in-depth analysis of 68 peptide/HLA crystallographic structures from the MPID-T database (Tong et al., 2006b). Our analysis revealed the striking observation that peptide/HLA structural interaction patterns vary among different alleles and may be grouped in a supertype dependent manner. The results obtained in this study shed new light into HLA supertype definition, suggesting that HLA supertype definitions may not be limited to peptide binding motifs or receptor information, and can be characterized at the intermolecular level, based on the interactions between HLA proteins and their associated peptides, consistent with solutions from X-ray crystallography. Through the use of structural interaction parameters described herein, a novel HLA class I supertype classification schema has been developed for alleles with available crystallographic structures.
|
2 METHODS |
|---|
|
TOPABSTRACT1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
2.1 Data
A total of 68 peptide–HLA complexes spanning 13 classes I alleles from the MPID-T database (Tong et al., 2006b, http://surya.bic.nus.edu.sg/mpidt) were used in the current analysis (Table 1).
|
2.2 Hierarchical clustering
A hierarchical clustering technique using the agglomerative algorithm (Barnard and Downs, 1992; Doytchinova et al., 2004; Doytchinova and Flower, 2005) was applied. The distance between the structures was computed by the single-linkage method as implemented in MATLAB version 7.0 based on the separation between the each pair of data points (Barnard and Downs, 1992). The nearest neighbours were merged into clusters. Smaller clusters were then merged into larger clusters based on inter-cluster distances, until all structures are combined. We have considered the last three levels for defining HLA class I supertypes.
2.3 Interaction parameters
Some interaction parameters have been identified as being significant for the characterization of peptide/MHC interface (Kangueane et al., 2001; Govindarajan et al., 2003) and can be computed from the 3D coordinates of a peptide–MHC complex (Berman et al., 2000; Kaas et al., 2004). These parameters were applied in this study for analyzing the binding characteristics of HLA supertypes:
2.4 Intermolecular hydrogen bonds
The number of intermolecular hydrogen bonds between the bound peptide and MHC protein was calculated using HBPLUS (McDonald and Thornton, 1994) in which hydrogen bonds are defined in accordance to standard geometric parameters. Hydrogen bonding patterns of all complexes presented in this study are available in MPID-T (Govindarajan et al., 2003; Tong et al., 2006b; http://surya.bic.nus.edu.sg/mpidt).
2.5 Interface area between peptide and MHC
The accessible surface area (ASA; Å2) between the bound peptide and MHC is measured by tracing out the maximum permitted van der Waals' contact that is covered by the center of a water molecule as it rolls over the surface of the protein. Interface area for MHC class I complexes is defined as the mean 
ASA on complexation when going from a monomeric MHC protein to a dimeric peptide/MHC complex state and calculated as half the sum of the total ASA for both molecules for each type of complex. Interface area for MHC class II complexes is similarly defined (Kangueane et al., 2001).
2.6 Gap volume
The gap volume or volume (Å3) enclosed by the bound peptide and MHC protein is computed using the SURFNET program (Laskowski, 1995). The algorithm places a series of spheres (maximum radius 5.00 Å) midway between the surfaces of each pair of subunit atoms, such that its surface is in contact with the surfaces of the atoms of the pair. The size of each sphere is reduced accordingly whenever it is intercepted by other atoms and subsequently discarded if it falls below a minimum allowed radius (1.00 Å). The sizes of all the remaining allowable gap-spheres are subsequently used to compute the gap volume between the two subunits.
2.7 Gap index
One essential feature in receptor-ligand binding is the electrostatic and geometric complementarity observed between associating molecules. In this study, we adopted the use of gap index (reviewed in Jones and Thornton, 1996) as means to evaluate complementarity of interacting interfaces between the bound peptide and HLA protein:
 |
|
3 RESULTS |
|---|
|
TOPABSTRACT1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
3.1 HLA-A supertypes
Three main clusters are observed: A-I (A*0101), A-II (A*0201) and A-III (A*6801), which are consistent with the assignments of from the groups of Sette (Sette and Sidney, 1999) and Flower (Doytchinova et al., 2004). Since A-I and A-III are confined to single structures; the analysis of A-II structures is described. The mean interface area for A-II is 846.3 ± 48.9 Å2. On average, the number of intermolecular hydrogen bonds and gap index is 11.1 ± 1.9 and 0.9 ± 0.2, respectively. Extensive hydrogen bonding networks are found in binding pockets A, B and F, which corresponds to IMGT peptide/MHC contact sites (Kaas and Lefranc, 2005). No clear difference is observed in the number of intermolecular hydrogen bonds for 9mer (11.0 ± 1.8) and 10mer (11.8 ± 2.2) complexes. The gap indices, however, for the 9mer and 10mer complexes are 1.0 ± 0.2 and 0.8 ± 0.3, respectively, indicating that the interacting surfaces of 10mer complexes are generally more complementary than 9mer complexes. The interface area inversely correlates with gap index, indicating that A-II complexes with larger interface area have better geometric and electrostatic complementarities, stabilized by the formation of several intermolecular hydrogen bonds.
3.2 HLA-B supertypes
The hierarchical clustering for the structural interaction characteristics of 10 HLA-B alleles is given in Figure 1. Five main clusters can be identified in this study: B-I (B*2705, B*2709, B*4403), B-II (B*0801, B*3501, B*5301), B-III (B*4402, B*4405), B-IV (B*5101) and B-V (B*1501). Our HLA-B supertype definition largely overlaps (70% consensus) with the definition by binding motifs (Sette and Sidney, 1999; Rammensee et al., 1999). However, our data indicates that B*5101 does not share similar interaction patterns with either B*3501 and B*5301 from the SS-B7 supertype (Sette and Sidney, 1999; Rammensee et al., 1999; Lund et al., 2004); or B*4402, B*4403 and B*4405 from the DF-B44 supertype (Doytchinova et al., 2004) and may form a separate supertype instead. On average, B*5101 complexes have a smaller gap volume (766.5 Å3) and fewer intermolecular hydrogen bonds (7.5) in comparison with B*3501 (gap volume = 870.2 Å3, intermolecular hydrogen bonds = 12.0) and B*5301 (gap volume = 875.0 Å3, intermolecular hydrogen bonds = 11.0). The B*0801 allele clustered within B-II is in agreement with the B7 supertype definition based on hierarchical clustering and consensus principle component analysis on the functional residues of HLA proteins (Doytchinova et al., 2004). The deciding factors for this cluster are interface area (876.2 ± 3.0 Å2) and gap volume (850.8 ± 19.5 Å3). This allele has, however, been classified as an outlier by Sette (Sette and Sidney, 1999).
|
B-I has the largest average interface area (947.8 ± 148.0
2) among all class I supertypes investigated in this study. On average, the number of intermolecular hydrogen bonds, interface area, gap volume, and gap index for B-I is 15.5 ± 7.1, 919.5 ± 25.1, 909.4 ± 8.8 and 1.0 ± 0.0, respectively. The number of intermolecular hydrogen bonds inversely correlates with gap volume (Figure 2G; r =–0.42) and gap index (Figure 2H; r = –0.46). Hydrogen bonds are localized in pockets A, B and F. More hydrogen bonds are formed at smaller gap index compared to other HLA class I complexes B-I has the largest average interface area (947.8 ± 148.0 2) among all class I supertypes investigated in this study. On average, the number of intermolecular hydrogen bonds, interface area, gap volume, and gap index for B-I is 15.5 ± 7.1, 919.5 ± 25.1, 909.4 ± 8.8 and 1.0 ± 0.0, respectively. The number of intermolecular hydrogen bonds inversely correlates with gap volume (Figure 2G; r = –0.42) and gap index (Figure 2H; r = –0.46). Hydrogen bonds are localized in pockets A, B and F. More hydrogen bonds are formed at smaller gap index compared to other HLA class I complexes investigated herein. The average number of intermolecular hydrogen bonds, gap index, and interface area for B-II is 13.8 ± 2.5, 1.0 ± 0.2 and 879.2 ± 72.4 2, respectively. In general, intermolecular hydrogen bonds for this supertype are well distributed and do not correlate with the interface area (Figure 2I; r = 0.01), gap volume (Figure 2K, r = –0.01) and gap index (Figure 2L, r = –0.01) of complex. B*4403 clusters with B*2705 and B*2709 although the three alleles have different specificities based on existing binding motifs (Sette and Sidney, 1999). The gap volume of B*4403 (1013.0 Å3) is clearly similar to that of B27 alleles (985.6 ± 4.2 Å3) and not the other B44 alleles. Based on existing binding data (Rammensee et al., 1999; de Castro et al., 2004), B*2705 exhibits strong predominance for Arg at position 2, while B*4403 prefers Glu at this position. Nonetheless, the existence of Glu at position 2 has also been reported in B*2705-binding peptides derived from fresh blood lymphocytes (Stod
lková et al., 2004), and this dual Arg/Glu substitution mechanism also exist in the auxiliary anchor residues of other alleles, including A*3101, and B*07 (Rammensee et al., 1999). Given the caveat of existing binding data, peptide motifs for many of these alleles are primarily derived from limited sources (Rammensee et al., 1999) and only sequences that have been studied are reported. It is possible that the current motifs for many less-studied alleles are under-represented and biased towards selected anchor residues investigated in the relevant studies. This could be the norm rather than the exception as more binding assays are carried out in the laboratory.
|
|
4 DISCUSSION |
|---|
|
TOPABSTRACT1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
The extremely high polymorphism of HLA alleles (Williams, 2001) has been a confounding factor in the study of HLA peptide binding specificities. For a HLA protein to recognize specific peptide, geometric and electrostatic complementarity between the receptor and its corresponding ligand is essential for the formation of chemical bonds between their functional groups, which in turn determines the net stability of the complex. In this context, we introduced the use of structural interaction information to analyze high-level relationships hidden within peptide/HLA crystallographic structures at supertype level. Such descriptors should better reflect peptide/HLA interactions than just sequence data alone.
In general, three types of hydrogen bonding patterns for peptide–HLA class I complexes investigated herein can be identified: (1) the gap index directly correlates with the number of intermolecular hydrogen bonds (Figure 2D); (2) the gap index does not correlate with the number of intermolecular hydrogen bonds (Figure 2L) and (3) the gap index inversely correlates with the number of intermolecular hydrogen bonds (Figure 2H). For the first group (A2 supertype), the majority of intermolecular hydrogen bonds are concentrated at both ends of the binding groove (in pockets A, B and F). More hydrogen bonds are observed with decreasing geometric and electrostatic complementarity (i.e. increasing gap index; Figure 2D) as well as increasing gap volume (Figure 2C). However, the correlations are weak. For the second group (B-II supertype), the gap index is independent of the number of intermolecular hydrogen bonds formed. A loss of four hydrogen bonds is noticed for B*0801, when the peptide GGKKKYQL (PDB ID: 1AGC) is modified to GGKKRYKL (PDB ID: 1AGF), while the gap index is unchanged. A possible explanation for these observations is that the interaction mechanism employed by this group may be degenerate and a combination of non-covalent interactions (hydrogen bonds, hydrophobic and ionic interactions) may be involved in peptide selection. However, it is not clear to what extent the different interactive forces contribute to the net stability of complex. For the third group (B-I supertype), the number of intermolecular hydrogen bonds increases with higher geometric and electrostatic complementarity (smaller gap index). For instance, the number of intermolecular hydrogen bonds between GRFAAAIAK and B*2705 (PDB ID: 1JGE [PDB] ) increases from 15 to 19 when the same peptide binds to B*2709 (PDB ID: 1K5N [PDB] ). Thus although the structures of B*2705 and B*2709 bind identical or similar peptides, the four parameters used in this study result in a more comprehensive characterization of the peptide–MHC interaction. This group also consists of the highest mean number of intermolecular hydrogen bonds. The results strongly indicate that the complexes formed by this group may be more stable, with higher overall geometric and electrostatic complementarity.
Computational techniques with different degrees of accuracy have been developed to model the structure of peptide–MHC complexes, and predict the binding free energy of peptides to MHC proteins (reviewed in Tong et al., 2006c). While some studies showed excellent results when applied to specific sets of alleles, the results presented here suggest that the use of a standardized set of structural interaction rules or free energy scoring functions to discriminate binding peptides may not be applicable for all MHC alleles as interaction characteristics vary across MHC supertypes. The modeling of MHC bound peptide ligands is non-trivial. Although the conformational space of MHC binding peptides is restricted by the highly conserved binding groove, peptide side-chain prediction still requires the screening of a large conformational space (Schueler-Furman et al., 1998). The use of ab initio techniques (Tong et al., 2004, 2006d; Bordner and Abagyan, 2006) and rotamer libraries (Schueler-Furman et al., 1998) are excellent approaches. For the latter, the results presented herein suggest that existing rotamer libraries within this context may be further refined by selecting rotamers from representative sets of known structures within the same MHC superfamily. Likewise, where binding free energy prediction is of concern, the design of generalized free energy scoring functions applicable for all MHC alleles may not be appropriate. Accordingly, it is important to identify key parameters for optimal predictive results for the alleles of interest (Tong et al., 2006c; Davies et al., 2006). The classification scheme presented herein is applicable to facilitate the identification of binding peptides for alleles with scarce experimental peptide information, as well as the development of predictive algorithms based on machine learning and artificial intelligence approaches.
The present analysis is difficult due to the limited number of peptide/HLA crystallographic structures in the current PDB (Berman et al., 2000). Nonetheless, we have demonstrated that different HLA proteins employ the use of different binding mechanism for selectivity of antigenic peptides in a supertype dependent manner. By focusing solely on the use of experimental 3D structures, our analysis is supported and verified by existing data. The proposed classification scheme provides an alternative to HLA supertype analysis using either sequence (Chelvanayagam, 1996; Zhang et al., 1998; Zhao et al., 2003; Cano et al., 1998; McKenzie et al., 1999; Lund et al., 2004) or receptor structure information (Doytchinova et al., 2004; Doytchinova and Flower, 2005; Kangueane et al., 2005) alone. In silico analysis of peptide–HLA interaction characteristics opens the way for more in-depth understanding of the binding mechanism involved in peptide selection and better characterization of HLA supertypes. Future work will focus on the use of molecular modeling techniques for large-scale classification of HLA class I and class II supertypes for which experimental crystallographic structures are unavailable as well as in-depth analysis of their structural interaction patterns.
|
Acknowledgments |
|---|
Conflicts of Interest: none declared.
|
FOOTNOTES |
|---|
Associate Editor: Dmitrij Frishman
Received on July 11, 2006; revised on November 3, 2006; accepted on November 3, 2006
|
REFERENCES |
|---|
|
TOPABSTRACT1 INTRODUCTION2 METHODS3 RESULTS4 DISCUSSIONREFERENCES |
|---|
Barnard, J.M. and Downs, G.M. (1992) Clustering of chemical structures on the basis of two-dimensional similarity measures. J. Chem. Inf. Comput. Sci, . 32, 644–649[CrossRef].
Berman, H.M., et al. (2000) The Protein Data Bank. Nucleic Acids Res, . 28, 235–242
Bordner, A.J. and Abagyan, R. (2006) Ab initio prediction of peptide-MHC binding geometry for diverse class I MHC allotypes. Proteins, 63, 512–526[CrossRef][Web of Science][Medline].
Chelvanayagam, G. (1996) A roadmap for HLA-A, HLA-B, and HLA-C peptide binding specificities. Immunogenetics, 45, 15–26[CrossRef][Web of Science][Medline].
Cano, P., et al. (1998) A geometric study of the amino acid sequence of class I HLA molecules. Immunogenetics, 48, 324–334[CrossRef][Web of Science][Medline].
Davies, M.N., et al. (2006) Statistical convolution of enthalpic energetic contributions to MHC-peptide binding affinity. BMC Struct. Biol, . 6, 5[CrossRef][Medline].
de Castro, J.A.L., et al. (2004) HLA-27: a registry of constitutive peptide ligands. Tissue Antigens, 63, 424–445[CrossRef][Web of Science][Medline].
Del Guercio, M.F., et al. (1995) Binding of a peptide antigen to multiple HLA alleles allows definition of an A2-like supertype. J. Immunol, . 154, 685–693[Abstract].
Doytchinova, I.A., et al. (2004) Identifying human MHC supertypes using bioinformatic methods. J. Immunol, . 172, 4314–4323
Doytchinova, I.A. and Flower, D.R. (2005) In silico identification of supertypes for class II MHCs. J. Immunol, . 174, 7085–7095
Falk, K., et al. (1991a) Identification of naturally processed viral nonapeptides allows their quantification in infected cells and suggests an allele-specific T cell epitope forecast. J. Exp. Med, . 174, 425–434
Falk, K., et al. (1991b) Allele-specific motifs revealed by sequencing of self-peptides eluted from MHC molecules. Nature, 351, 290–296[CrossRef][Medline].
Govindarajan, K.R., et al. (2003) MPID: MHC-Peptide Interaction Database for sequence-structure-function information on peptides binding to MHC molecules. Bioinformatics, 19, 309–310
Hunt, D.F., et al. (1992) Characterization of peptides bound to the class I MHC molecule HLA-A2.1 by mass spectrometry. Science, 255, 1261–1263
Jones, S. and Thornton, J.M. (1996) Principles of protein–protein interactions. Proc. Natl Acad. Sci. USA, 93, 13–20
Kaas, Q., et al. (2004) IMGT/3Dstructure-DB and IMGT/StructuralQuery, a database and a tool for immunoglobulin, T cell receptor and MHC structural data. Nucleic Acids Res, . 32, D208–D210
Kaas, Q. and Lefranc, M.P. (2005) T cell receptor/peptide/MHC molecular characterization and standardized pMHC contact sites in IMGT/3Dstructure-DB. In Silico Biol, . 5, 505–528[Medline].
Kangueane, P., et al. (2001) Towards the MHC-peptide combinatorics. Hum Immunol, . 62, 539–556[CrossRef][Web of Science][Medline].
Kangueane, P., et al. (2005) A framework to sub-type HLA supertypes. Front. Biosci, . 10, 879–886[Web of Science][Medline].
Kobayashi, H., et al. (2001) Identification of helper T cell epitopes that encompass or lie proximal to cytotoxic T cell epitopes in the gp100 melanoma tumor antigen. Cancer Res, . 61, 7577–7584
Laskowski, R.A. (1995) SURFNET: a program for visualizing molecular surfaces, cavities and intermolecular interactions. J. Mol. Graph, . 13, 323–330[CrossRef][Web of Science][Medline].
Lefranc, M.P. and Lefranc, G. The T Cell Receptor FactsBook, (2001) , London Academic Press.
Lund, O., et al. (2004) Definition of supertypes for HLA molecules using clustering of specificity matrices. Immunogenetics, 12, 797–810.
McDonald, I.K. and Thornton, J.M. (1994) Satisfying hydrogen bonding potential in proteins. J. Mol. Biol, . 238, 777–793[CrossRef][Web of Science][Medline].
McKenzie, L.M., et al. (1999) Taxonomic hierarchy of HLA class I allele sequences. Genes Immun, . 1, 120–129[CrossRef][Web of Science][Medline].
Panigada, M., et al. (2002) Identification of promiscuous T cell epitope in Mycobacterium tuberculosis Mce proteins. Infect. Immun, . 70, 79–85
Rammensee, H.G., et al. (1993) Peptides naturally presented by MHC class I molecules. Annu. Rev. Immunol, . 11, 213–244[CrossRef][Web of Science][Medline].
Rammensee, H.G., et al. (1999) SYFPEITHI: database for MHC ligands and peptide motifs. Immunogenetics, 50, 213–219[CrossRef][Web of Science][Medline].
Rotzschke, O., et al. (1991) Exact prediction of a natural T cell epitope. Eur. J. Immunol, . 21, 2891–2894[Web of Science][Medline].
Schueler-Furman, O., et al. (1998) Knowledge-based structure prediction of MHC class I bound peptides: a study of 23 complexes. Fold Des, . 3, 549–564[CrossRef][Web of Science][Medline].
Sette, A. and Sidney, J. (1999) Nine major HLA class I supertypes account for vast preponderance of HLA-A and -B polymorphism. Immunogenetics, 50, 201–212[CrossRef][Web of Science][Medline].
Sette, A., et al. (2001) The development of multi-epitope vaccines: epitope identification, vaccine design and clinical evaluation. Biologicals, 29, 271–276[CrossRef][Web of Science][Medline].
Sette, A., et al. (2002) Optimizing vaccine design for cellular processing, MHC binding and TCR recognition. Tissue Antigens, 59, 443–451[CrossRef][Web of Science][Medline].
Stodlková, E., et al. (2004) Peptides eluted from HLA-B27 of human splenocytes and blood cells reveal a similar but partially different profile compared to in vitro grown cell lines. Immunology Lett, . 94, 261–265[CrossRef][Web of Science][Medline].
Tong, J.C., et al. (2004) Modeling the structure of bound peptide ligands to major histocompatibility complex. Protein Sci, . 13, 2523–2532[CrossRef][Web of Science][Medline].
Tong, J.C., et al. (2006a) Modeling the bound conformation of Pemphigus vulgaris-associated peptides to MHC class II DR and DQ alleles. Immunome Res, . 2, 1[CrossRef][Medline].
Tong, J.C., et al. (2006b) MPID-T: database for sequence-structure-function information on TCR/peptide/MHC interactions. Appl. Bioinformatics, 5, 111–114[CrossRef][Medline].
Tong, J.C., et al. (2006c) Methods and protocols for predicting immunogenic epitopes. Brief. Bioinformatics, in press, Advance Access published on October 31, 2006. doi:10.1093/bib/bbl038.
Tong, J.C., et al. (2006d) Prediction of HLA-DQ3.2ß ligands: evidence of multiple registers in class II binding peptides. Bioinformatics, 22, 1232–1238
Williams, T.M. (2001) ‘Human leukocyte antigen gene polymorphism and the histocompatibility laboratory’. J. Mol. Diagn, . 3, 98–104
Zhang, C., et al. (1998) Structural principles that govern the peptide-binding motifs of class I MHC molecules. J. Mol. Biol, . 281, 929–947[CrossRef][Web of Science][Medline].
Zhao, B., et al. (2003) Compression of functional space in HLA-A sequence diversity. Hum. Immunol, . 64, 718–728[CrossRef][Web of Science][Medline].
Zhu, S., et al. (2006) Improving MHC binding peptide prediction by incorporating binding data of auxiliary MHC molecules. Bioinformatics, 22, 1648–1655
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||