Principal Component Analysis and Target Testing of Substituent Effect Using Carbonyl Stretching Frequency and 13C NMR Chemical Shift Data Matrices

Abstract

Principal component analysis technique has been applied to analyse the substituent effect on carbonyl stretching frequency and 13C NMR chemical shifts. The general formula for the investigated molecules is X-G-Y, where X represents the set of substituent (OMe, Me, F, Cl, Br, CN and NO2), Y is the probe site and G is benzene ring. According to the indicator function two significant components are responsible for the substituent effect. The validity of several substituent parameters have been investigated by target testing technique. Invariabily substituent parameters derived by iterative multiple linear regression analysis viz. σR (Reynolds), σF (Reynolds) and σR (NMR) have lower SPOIL values when compared with other substituent parameters. Model designing f IR and 13C NMR data matrices separately have shown that models which incorporate σR (Reynolds) and σF (Reynolds) or σR (NMR) and a substituent field parameters have the lowest root mean square error RMSE. Substituent effect on several properties are better correlated with Reynolds' σR and σF than with other commonly used substituent parameter(s). The orthogonality of substituent parameters used in the model can be achieved by including the methyl group in the substituent set.