A Systematic Approach to Obtain Validated Partial Least Square Models for Predicting Lipoprotein Subclasses from Serum NMR Spectra Velitchka V. Mihaleva,* ,†,‡ Danië l B. van Schalkwijk, § Albert A. de Graaf, § John van Duynhoven, ‡,∥,⊥ Ferdinand A. van Dorsten, ‡,∥ Jacques Vervoort, †,‡ Age Smilde, ‡,@ Johan A. Westerhuis, ‡,@ and Doris M. Jacobs ‡,∥ † Laboratory of Biochemistry, Wageningen University, Dreijenlaan 3, 6703 HA Wageningen, The Netherlands ‡ Netherlands Metabolomics Centre, Einsteinweg 55, 2333 CC Leiden, The Netherlands § TNO, Microbiology and Systems Biology, Utrechtseweg 48, 3700 AJ Zeist, The Netherlands ∥ Unilever R&D, Olivier van Noortlaan 120, 3133 AT Vlaardingen, The Netherlands ⊥ Laboratory of Biophysics, Wageningen University, Dreijenlaan 3, 6703 HA Wageningen, The Netherlands @ Swammerdam Institute for Life Sciences, Universiteit van Amsterdam, Science Park 904, 1098 XH Amsterdam, The Netherlands * S Supporting Information ABSTRACT: A systematic approach is described for building validated PLS models that predict cholesterol and triglyceride concentrations in lipoprotein subclasses in fasting serum from a normolipidemic, healthy population. The PLS models were built on diffusion-edited 1 H NMR spectra and calibrated on HPLC-derived lipoprotein subclasses. The PLS models were validated using an independent test set. In addition to total VLDL, LDL, and HDL lipoproteins, statistically significant PLS models were obtained for 13 subclasses, including 5 VLDLs (particle size 64−31.3 nm), 4 LDLs (particle size 28.6−20.7 nm) and 4 HDLs (particle size 13.5−9.8 nm). The best models were obtained for triglycerides in VLDL (0.82 < Q 2 <0.92) and HDL (0.69 < Q 2 <0.79) subclasses and for cholesterol in HDL subclasses (0.68 < Q 2 <0.96). Larger variations in the model performance were observed for triglycerides in LDL subclasses and cholesterol in VLDL and LDL subclasses. The potential of the NMR-PLS model was assessed by comparing the LPD of 52 subjects before and after a 4-week treatment with dietary supplements that were hypothesized to change blood lipids. The supplements induced significant (p < 0.001) changes on multiple subclasses, all of which clearly exceeded the prediction errors. L ipoprotein particles are supramolecular protein−lipid aggregates that act as bloodstream carriers of lipids between the different organs in our body. On the basis of their density and size, the lipoprotein particles can be divided in four fractions: Chylomicrons (CM), very low density lip- oproteins (VLDL), low density lipoproteins (LDL), and high density lipoproteins (HDL). Sometimes also intermediate density lipoproteins (IDL) are defined to describe the transition from VLDL to LDL. CMs are large particles that carry lipids from the intestines to major organs and typically circulate only in the nonfasting state. Triglycerides and cholesterol are transported by VLDL from the liver to various tissues. The triglycerides in VLDLs are hydrolyzed by lipase generating IDL and LDL. HDLs are involved in reverse cholesterol transfer from the tissues back to the liver. Total LDL and HDL cholesterol are routinely used in the risk assessment of cardiovascular disease (CVD). Elevated LDL cholesterol is positively correlated with a high risk of CVD, whereas for HDL cholesterol this correlation is negative. 1 There is evidence that not only the cholesterol concentration but also the particle sizes are important for CVD risk assessment. 2,3 For example, small dense LDL particles are considered to be more atherogenic compared to larger, relatively less dense LDL particles. 2,3 The inverse correlation is proposed for the HDL particles. 2,3 In addition, information about the metabolic precursor VLDL is present in lipoprotein particle distributions (LPD) but currently often not utilized. Therefore, LPDs have a high potential for improving the diagnostics of CVD and metabolic disorders. Furthermore, LPDs can help to move from diagnostics toward early Received: August 13, 2013 Accepted: December 9, 2013 Published: December 9, 2013 Article pubs.acs.org/ac © 2013 American Chemical Society 543 dx.doi.org/10.1021/ac402571z | Anal. Chem. 2014, 86, 543−550