In Silico QSAR-Based Predictions of Class I and Class II MHC Epitopes - Immunoinformatics - page 72

Fig. 1. Relative contributions of positionwise amino acids at each binding position 1 to 9 for
the H2-D b (black bars), H2-K b (white bars) and H2-K k (striped bars) alleles. The contribution
made by different individual amino acids at each position of the 9mer H2-D b , H2-K b and
H2-K k binding peptide. The contribution is equivalent to a positionwise amino acid regression
coefficient obtained by PLS regression (as described in the text).
4.3.2 Iterative Self-Consistent (ISC) Algorithm - Class II Alleles
The generated models ( n =44-185), as shown in Table 4, have an acceptable level of
predictivity: LOO-CV statistical terms, SEP and q 2 , ranged between 0.418 and 0.816
and 0.649 and 0.925, respectively. The non-cross-validated statistical terms NC,
SEE, and r 2 ranged between 4 and 8, 0.051 and 0.180, and 0.967-0.999, respectively.
Convergence ranged between the 4 th and 17 th iteration. An extended motif, as defined
by the class I models, is summarized in Table 5 showing anchor and nonanchor resi-
dues related to strong and weak binding residues.