Premium
QSAR Studies on Dipeptides Based on a Combinatorial MHDV‐GA‐MLR Method
Author(s) -
Liu ShuShen,
Yin ChunSheng,
Wang XiaoDong,
Wang LianSheng
Publication year - 2002
Publication title -
journal of the chinese chemical society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.329
H-Index - 45
eISSN - 2192-6549
pISSN - 0009-4536
DOI - 10.1002/jccs.200200157
Subject(s) - chemistry , quantitative structure–activity relationship , molecular descriptor , correlation coefficient , linear regression , feature selection , set (abstract data type) , selection (genetic algorithm) , biological system , artificial intelligence , stereochemistry , statistics , mathematics , computer science , biology , programming language
A combinatorial method for estimating and predicting the biological activities of two sets of dipeptides, a set of 48 compounds and another set of 58, was developed. The molecular holographic distance vector (MHDV) was employed to characterize the structures of the peptide molecules. Preliminary selection of the MHDV descriptors was performed based on the number of the molecules having non‐zero MHDV values. The final optimal descriptors were completed by a genetic algorithm‐based variable selection procedure. Then the optimal descriptors are used to relate to the biological activities of the peptides using the multiple linear regression (MLR) method. For two panels of dipeptides, the correlation coefficient of estimations ( r ) are respectively 0.9651 for 48 peptides and 0.936 for 58 peptides, and the correlation coefficient of leave‐one‐out predictions ( q ) are respectively 0.9452 and 0.9075.