Premium
Comparative QSAR study on hydroxyl radical reactivity with unsaturated hydrocarbons: PLS versus MLR
Author(s) -
Medven Željka,
Güsten Hans,
Sabljić Aleksandar
Publication year - 1996
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/(sici)1099-128x(199603)10:2<135::aid-cem413>3.0.co;2-v
Subject(s) - quantitative structure–activity relationship , chemistry , homo/lumo , molecular descriptor , linear regression , partial least squares regression , computational chemistry , quantum chemical , quantum chemistry , molecule , thermodynamics , organic chemistry , mathematics , stereochemistry , statistics , supramolecular chemistry , physics
A reliable QSAR model for estimating the reaction rate constants of unsaturated hydrocarbons with hydroxyl radical has been developed by application of the PLS (partial least squares projections to latent structures) multivariate technique and a large set of 18 empirical variables and calculated quantum chemical or structural descriptors. It was found that five calculated quantum chemical and structural descriptors have a pronounced influence on the reaction rate constants of unsaturated hydrocarbons with hydroxyl radical: energy of the highest occupied molecular orbital ( E HOMO ), hardness ( HA ), average number of alkyl substitutents per unsaturated bond ( n Alk), energy of the lowest unoccupied molecular orbital ( E LUMO ) and number of carbon atoms in unsaturated bonds ( n c ). The derived QSAR model was rationalized within the framework of the frontier molecular orbital (FMO) theory and by thermodynamic arguments. In addition, a QSAR model has been derived using multiple linear regression (MLR) and a stepwise procedure with the same set of measured and calculated descriptors in order to assess the advantages and limitations of both statistical methods. Finally, a comparison was made between estimates from this study and the estimates from QSAR models published earlier, i.e. Atkinson's fragment contribution method and the non‐linear QSAR model of Klamt based on calculated quantum chemical descriptors. This comparison has provided important information on the performance of individual models as well as on their advantages and limitations. It can be safely concluded that the PLS model derived in this study is the method of choice for future reliable estimation of log k OH data for unsaturated chemicals.