关键词: 抗原肽, 定量序效模型, 高维特征, 支持向量回归, 多肽疫苗

Abstract:

Determining highly active epitopes of the cytotoxic T lymphocyte (CTL) is essential for the computational design of peptide vaccines for tumors. In this study, we characterized each residue in the restricted CTL epitopes using 531 physicochemical properties. We selected 18 descriptors with clear meanings from 531×9 descriptors for each peptide of length nine using the binary matrix shuffling filter and worst descriptor elimination multi-round methods. Most of the 18 selected descriptors were the hydrophobic and steric properties of the residues. Among the 18 descriptors, 10 descriptors were related to the second, fourth, and ninth residues, which is consistent with the known facts. We then constructed a support-vectorregression-based quantitative sequence activity model (QSAM) using 18 selected descriptors. The values of the accuracies of fitting (R²), leave-one-out cross validation (Q_cv²), and extra-sample prediction (Q_ext², RMSE_ext) were 0.957, 0.708, 0.818, and 0.366, respectively. These results, which were tested on HLAA^* 0201 data, showed that our QSAM was superior to those reported in the literature. Finally, we predicted the activities of peptides of all possible combinations of the nine residues. Several peptides were found with higher affinity activities than those of previously reported epitopes. Our study improves the understanding of the relationship between the compositional residues and the affinity activity of the peptide, which provides a valuable guideline for the design of highly active peptide vaccines. Our predicted high affinity peptides are potential candidates for further experimental verification.

Key words: Antigenic peptide, Quantitative sequence activity model, High-dimensional descriptor, Support vector regression, Peptide vaccine