MP06:Partial Least Squares (PLS1) Algorithm for Obtaining Meaningful Concentrations of Cholesterol and Polyunsaturated Fatty Acids in Human Serum
Gerard G. Dumancas1, Mary Muriuki1, Neil Purdie1, and Lisa Reilly2
1Department of Chemistry, Oklahoma State University, Stillwater, OK, USA 74078
2Department of Physical Sciences, Bethany College, Bethany, WV 26032
We have previously exploited various chemometric algorithms for the direct determination of cholesterol and polyunsaturated fatty acid (PUFA) molar concentrations in synthetic mixtures and human serum. The simple colorimetric assay used is rapid, rugged, inexpensive, and specific to the -CH=CH-CH2- group that accomplishes, in a single assay the simultaneous quantitation of cholesterol, ω-3 (methyl esters of linolenic, eicosapentaenoic (EPA) and docosahexaenoic (DHA) fatty acids), and ω-6 (methyl esters of linoleic, conjugated linoleic (CLA), and arachidonic fatty acids). Previously, ridge regression (RR), P-matrix regression (PM), principal component regression (PCR), and partial least squares (PLS2) successfully out-performed the K-matrix regression (KM) approach when applied to the study of prepared mixtures (synthetic sera) in chloroform solutions. In this paper, partial least squares in the form of PLS1 is investigated and applied to obtain meaningful concentrations in actual human serum samples. Results show that PLS1 yielded lesser root mean square errors of prediction in the calibration model and more meaningful concentrations than PLS2 in the actual human serum samples.
|Click [+] for other articles on||The Market Place for Lab Automation & Screening||The Market Place|