Partial least squares regression and fourier transform infrared (FTIR) microspectroscopy for prediction of resistance in hepatocellular carcinoma HepG2 cells

Abstract

We evaluated the feasibility of FTIR microspectroscopy combined with partial least squares regression (PLS-R) for determination of resistance in HepG2 cells. Cell viability testing was performed using neutral red assay for the concentration of cisplatin resulting in 50% antiproliferation (IC50). The resistance index (RI) is the ratio of the IC50 in resistant HepG2 cells vs. parental HepG2 cells. Principal component and unsupervised hierarchical cluster analyses were applied and a differentiation of samples of cells (parental, 1.8RI, 2.3RI, 3.0RI, and 3.5RI) was demonstrated (3000–2800cm−1 in the lipid and 1700–1500cm−1 in the protein regions. The FTIR spectra were preprocessed with several treatments to test the algorithm. PLS-R models were built using the 1170 spectra of the HepG2 cells. Cross-validation was used to evaluate prediction of the RI value using this model. PLS-R models—preprocessed with the second derivative FTIR spectra—yielded the best model (R2=0.99, RMSEE=0.095 and RPD=7.98). Most RI values were predicted with high accuracy (91–100%) such that the linear correlation between the actual and predicted RI values was nearly perfect (slope~1). FTIR microspectroscopy combined with chemometric analysis using PLS-R offers quick, accurate, and reliable quantitative analysis of HepG2 cell resistance.