Information Technology Reference
In-Depth Information
the H2-K
k
allele. In the remaining positions there seem to be no discernibly favored
interactions. Hydrophobic interaction shows a major disfavored interaction at posi-
tion 2 covering the whole side chain. The only favored interaction in the hydrogen
bond donor map in the H2-K
k
allele lies between positions 7 and 8. The main disfa-
vored interaction is found at position 2. Within the hydrogen bond acceptor map,
there is a strong disfavored interaction between the side chains at positions 2 and 3.
4.4.3 Iterative Self-Consistent (ISC) Algorithm - Class II Alleles
We have examined a recently developed bioinformatics method: the Iterative Self-
Consistent (ISC) Partial Least Squares (PLS)-based Additive Method, which was
applied to the prediction of class II Major Histocompatibility Complex (MHC)-
peptide binding affinity. We have shown previously that ISC is a reliable, quantita-
tive method for binding affinity prediction (Doytchinova and Flower 2003) develop-
ing a series of quantitative, systematic models, based on literature IC
50
values.
Experimental studies of T-cell epitope analogue binding and data from X-ray
crystallography, show that peptides bind to MHC molecules through the interaction
of side chains of certain peptide residues with pockets situated in the MHC class II
peptide-binding site: these side-chains extend into discrete pockets within the bind-
ing groove (Hennecke and Wiley 2002; Fremont, Monnaie, Nelson, Hendrickson,
and Unanue 1998; Corper, Stratmann, Apostolopoulos, Scott, Garcia, Kang, Wilson,
and Teyton 2000). Peptide side chains form favorable interactions with MHC side
chains within these pockets (Corper et al. 2000); the most critical determinant of
binding, other than the presence of appropriate types of side chain, is their relative
spacing. It has been suggested before that different MHC class II molecules can bind
the same peptide in several, alternative binding registers, whereby the peptide moves
sideways in the binding groove with side chains being bound by different pockets
(McFarland, Sant, Lybrand, and Beeson 1999; Li, Li, Martin, and Mariuzza 2000;
Vidal, Daniel, Vidavsky, Nelson, and Allen 2000). Reviewing this concept (Bank-
ovich, Girvin, Moesta, and Garcia 2004), identify two main alternative scenarios:
binding of the same peptide in different registers by the same or different alleles. The
more common second alternative is well demonstrated (Li et al. 2000; Vidal et al.
2000) and results from minor polymorphic differences in the amino acid residue
composition of the binding groove. In the DRB5 complex, the large P1 pocket
accommodates Phe from the peptide and Ile occupies the shallow pocket at P4. How-
ever, in the DRB1 allele, the small pocket at P1 is occupied by Val shifting the
peptide to the right, while Phe occupies a deeper pocket at P4. This also causes cer-
tain peptide side chains, which are orientated toward the TCR, to change (Li et al.
2000). Unequivocal evidence supporting the former alternative is somewhat scarce:
there are few, if any, proper examples of exactly the
same
peptide binding in differ-
ent registers to exactly the
same
MHC molecule.
Our results are consistent with the view that MHC binding motifs are a less-than-
adequate representation of the underlying mechanism of binding. As we have shown
elsewhere (Doytchinova, Walshe, Jones, Gloster, Borrow, and Flower 2004; Flower
2003), the whole of a peptide contributes to binding, albeit weighted differently at
different positions. At least for class I, it is even possible to generate high-affinity
