As water is a strong absorber in the near infrared (NIR) region, the most efficient way to analyze single molecules in complex fluids is to bind the analyte of interest with high efficiency and selectivity to a well-designed carrier material. This strategy enables on one side the direct measurement and on the other side an increase in sensitivity due to the appearing pre-concentrating effects. In the present contribution, a strategy for the analysis of low and high density lipoprotein (LDL and HDL) in human serum applying NIR spectroscopy and multivariate calibration techniques is described. During method development it is useful to evaluate the feasibility of NIRS for classifying and identifying different analytes by establishing a qualitative principal component analysis (PCA) based cluster model. In case of LDL and HDL analysis, titanium oxide (TiO₂) beads offer an efficient material for selective immobilization, including incubation and a washing step. For quantification, a principal component regression (PCR) model of standards in a range from 500–3000 ppm (clinical value is 1500 ppm) and a partial least squares regression (PLSR) model of HDL standards in a range from 100–1000 ppm (clinical value is 400 ppm) are highly efficient. Wavenumber region selection allowed gaining main spectral information between 4000 and 7240 cm⁻¹. For the analysis of real samples it is necessary to analyze HDL and LDL in chronological order by employing precipitation. It is demonstrated that this NIRS method is a highly useful potential alternative or even supplementary clinical method for the fast determination of single molecules in complex biological fluids.