Information Technology Reference
In-Depth Information
binding, and to explore and define preferred amino acids within each pocket. The
explanatory power of such a 3D-QSAR method is considerable, not only in its direct
prediction accuracy but also in its ability to map advantageous and disadvantageous
interaction potentials onto the structures of the peptides being studied. The data are
highly complementary to the detailed information obtained from crystal structures of
individual peptide-MHC complexes.
4.2 Methodology
4.2.1 Peptide Database
Th
e information and data based on the peptide sequences and their binding affinities
were obtained from the AntiJen database, a development of JenPep (Blythe,
Doytchinova,
and Flower
2002; McSparron, Blythe, Zygouri, Doytchinova and
Flower 2003) [URL: http://www.jenner.ac.uk/AntiJen]. Compilations of quantitative
affinity measures for peptides binding to class I and class II MHCs were carried out
with known binding affinities (IC
50
). These include human class I (HLA-A*0101,
HLA-A*0201, HLA-A*0202, HLA-A*0203, HLA-A*0206, HLA-A*0301, HLA-
A*1101, HLA-A*3101, HLA-A*6801, HLA-A*6802, HLA-B*3501), mouse class I
(H2-K
k
, H2-K
b
, and H2-D
b
), human class II (HLA-DRB1*0101, HLA-DRB1*0401,
and HLA-DRB1*0701), and mouse class II (I-A
b
, I-A
d
, I-A
k
, I-A
s
, I-E
d
and I-E
k
). For
class I, only nonameric peptides were included, with the exception of H2-K
b
and
H2-K
k
, where octameric peptides were also examined. For each set of class II alleles,
peptide lengths of 10 to 25 were obtained from the AntiJen database. Several QSAR
methodologies have been applied to both the class I and class II alleles and their
procedures are described as follows. All QSAR and molecular modeling calculations
were carried out on a Silicon Graphics octane workstation using the SYBYL 6.9
molecular modeling package (Tripos Inc., USA).
4.2.2 Additive Method - Class I and Class II Alleles
Extracted IC
50
values were first converted to log[1/IC
50
] values (or -log
10
[IC
50
]
or pIC
50
) and used as the dependent variables in a QSAR regression. pIC
50
can be
related to changes
IC
50
. The
values were predicted from a combination of the contributions (
p
) of individual
amino acids at each position of the peptide and used as the dependent variables
in a QSAR. The binding affinities were originally assessed by a competition
assay based on the inhibition of binding of the radiolabeled standard peptide to
detergent-solubilized MHC molecule (Ruppert, Sidney, Celis, Kubo, Grey, and
Sette 1993; Sette, Sidney, del Guercio, Southwood, Ruppert, Dalberg, Grey, and
Kubo 1994a).
We developed a program to transform the nine-amino-acid (aa) peptide
sequences into a matrix with elements 1 and 0. An element is 1 when a certain
amino acid at a certain position or a certain interaction between two side chains
in
the
free
energy
of
binding:
Δ
G
bind
=
-
RT
ln
