Study of quantitative structure-mobility relationship of the peptides based on the structural descriptors and support vector machines

摘要: Support vector machines (SVM), as a novel learning machine, was used to develop the non-linear quantitative structure-mobility relationship model of peptides based on calculated descriptors for first time. The molecular representing structural features compounds included constitutional and topological by CODESSA program, which can be obtained easily without optimizing structure molecule, CPSA (charged partial surface area) SYBYL software. MLR method select responsible electrophoretic mobility linear model. prediction result SVM (epsilon = 0.04, gamma 0.002 C 100) is much better than that method. RMS error training set, test set whole 0.0569, 0.0553, 0.0565 correlation coefficient 0.925, 0.912 0.922. respectively. results are in agreement with experimental values. This paper provided new effective predicting behavior peptide some insight into what related peptides. Moreover, it also offered an idea about dealing optimization obtaining their id biomacromolecules. (c) 2005 Elsevier B.V. All rights reserved.