Identification of Browning in Human Adipocytes by Partial Least Squares Regression (PLSR), Infrared Spectral Biomarkers, and Partial Least Squares Discriminant Analysis (PLS-DA) Using FTIR Spectroscopy

Abstract

We aimed to identify the browning of white adipocytes using partial least squares regression (PLSR), infrared spectral biomarkers, and partial least squares discriminant analysis (PLS-DA) with FTIR spectroscopy instead of molecular biology. PLSR helps distinguish human beige adipocytes treated with norepinephrine and rosiglitazone. When PLSR was based on the selected regions of 3997–3656 and 1618–938 cm−1, PLSR achieved an R2 of cross-validation of 88.95, a root mean square error of cross validation (RMSECV) of 2.13, and a ratio performance deviation (RPD) of 3.01. Infrared spectral biomarkers [1635 cm−1 (β-sheet amide I), 879–882, 860–3 cm−1 (A-form helix), and 629–38 cm−1 (OH out-of-plane bending)] were identified in human beige adipocytes based on spectral differences between human beige adipocytes and human white adipocytes, principal component analysis-linear discriminant analysis (PCA-LDA) cluster vector, U-test, and Fisher’s score per wavenumber. PLS-DA yielded a useful classification of adipocytes and expression distribution of adipogenesis genes in adipocytes. PLSR, infrared spectral biomarkers, and PLS-DA using FTIR spectroscopy are proposed as effective tools for identifying specific biological activities in a limited environment through features that do not require labeling and are relatively inexpensive in terms of time and labor.