PCA, followed by two-dimensional nonlinear mapping and cluster analysis, versus multilinear regression in QSSR

Annamaria Jakab, Gábor Schubert, Miklos Prodan, Esther Forgacs

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

Retention parameters of 45 different barbituric acid derivatives were determined on an amide embedded RP silica column (Discovery RP-AmideC16) using non-buffered tetrahydrofuran-water mixtures as eluents. Linear correlations were calculated between the lgk values and the tetrahydrofuran concentration in the eluent. Chromatographic parameters of barbituric acid derivatives were correlated with different conventional and quantum chemical structural descriptors in QSRR study (the different parameters were: intercept (lgk0) and slope (b) values of the linear, the combined retention parameter (lgk0/b), asymmetry factor (AF) and theoretical plate values (N)). Multilinear regression analysis (SRA) and principal component analysis (PCA), followed by two dimensional nonlinear mapping and cluster analysis techniques, were used to determine the retention behavior of barbituric acid derivatives. Different mathematical-statistical methods indicated that the hydrophobic characteristics of the solutes have marked influence on the retention behavior of barbituric acid derivatives on this amide embedded RP silica column in tetrahydrofuran-water eluent systems. The significant effect of the hydrophobic characteristics of the analytes in the retention behavior indicated that the effects of the interaction between the analytes and the residual silica silanol groups are negligible.

Original languageEnglish
Pages (from-to)1-16
Number of pages16
JournalJournal of Liquid Chromatography and Related Technologies
Volume25
Issue number1
DOIs
Publication statusPublished - Feb 18 2002

ASJC Scopus subject areas

  • Analytical Chemistry
  • Biochemistry
  • Pharmaceutical Science
  • Clinical Biochemistry

Fingerprint Dive into the research topics of 'PCA, followed by two-dimensional nonlinear mapping and cluster analysis, versus multilinear regression in QSSR'. Together they form a unique fingerprint.

  • Cite this