ACCURATE DETERMINATION OF DOUBLE-BOND POSITION IN MONOUNSATURATED STRAIGHT-CHAIN FATTY-ACID ETHYL-ESTERS FROM CONVENTIONAL ELECTRON-IMPACT MASS-SPECTRA BY QUANTITATIVE SPECTRUM STRUCTURE MODELING

An objective approach is given for selecting the most significant frag ments for mass spectrum-structure correlation. The method is based on partial least-squares regression (PLSR) on normalised spectral intensi ties as independent variables, and the carbon-carbon double bond posit ion as response. Such a correlation between a structural descriptor an d the mass spectra of pure compounds has not hitherto been reported in the literature. The procedure is demonstrated using conventional elec tron impact (70 eV) mass spectra of twelve straight-chain mono-unsatur ated fatty acid ethyl esters, ranging in carbon-carbon double bond pos itions from 5 to 11 (except 7), incorporating both cis- and trans-isom ers and chain lengths from C-16 to C20. On the basis of the relative i ntensity of thirteen fragments, a quantitative model relating these to the response was developed. At a 95% confidence interval, all the res iduals from the cross-validation fall within the confidence interval g iven by [true value +/- 0.46] carbon position units. Thus, the model p ractically always gives an unambiguous estimate of the carbon-carbon d ouble bond position. Possible further developments of this method are discussed for the use of gas chromatography-mass spectrometry and gas chromatography-Fourier transform infrared spectroscopy in combination.