Using scores of amino acid topological descriptors for quantitative sequence-mobility modeling of peptides based on support vector machine

Scores of amino acid topological descriptors (SATD) derived from principle components analysis of a matrix of 1262 structural variables related to 23 amino acids were employed to express the structure of 125 peptides in different length. Quantitative sequence-mobility modelings (QSMMs) were constructed using partial least square (PLS) and support vector machine (SVM), respectively. As new amino acid scales, SATD including plentiful information related to biological activity were easily manipulated. Better results were obtained compared to those obtained with PLS, which indicated that SVM presented robust stability and excellent predictive ability for electrophoretic mobilities. These results show that there is a wide prospect for the applications of SATD and SVM regression in QSMMs.