Moon, T Chi, MW Park, SJ Yoon, CN 2024-01-21T08:06:54Z 2024-01-21T08:06:54Z 2022-02-17 2003-10 1082-6076 https://pubs.kist.re.kr/handle/201004/138153 The quantitative structure-retention relationships (QSRR) studies on polycyclic aromatic hydrocarbons (PAHs) show that the multiple linear regression (MLR) models achieved in training sets have to be validated. The MLR models derived by using one or two descriptors showed high linear correlations between input descriptor(s) and high-performance liquid chromatography (HPLC) retention times, but showed very poor predictivity in test sets for one-descriptor models, whereas the models derived by using two descriptors (molecular connectivity and dipole moment) showed good predictivity. These results suggest that one-descriptor models are not sufficient to explain the retention time in spite of high r(2) values in training sets. In addition, the predictivity was affected by the solvent. High predictive r(2) values were obtained in conditions in which methanol was used as a solvent. English Marcel Dekker Inc. Prediction of HPLC retention time using multiple linear regression: Using one and two descriptors Article 10.1081/JLC-120025413 1 Journal of Liquid Chromatography and Related Technologies, v.26, no.18, pp.2987 - 3002 Journal of Liquid Chromatography and Related Technologies 26 18 2987 3002 N scopus 000186395900002 2-s2.0-0242441026 Biochemical Research Methods Chemistry, Analytical Biochemistry & Molecular Biology Chemistry Article POLYCYCLIC AROMATIC-HYDROCARBONS CHROMATOGRAPHIC RETENTION MOLECULAR CONNECTIVITY BIOLOGICAL ACTIVITY CHEMICAL STRUCTURE INDEXES HPLC retention time QSRR MLR PAHs