The use of EEM fluorescence data and OPLS/UPLS-DA algorithm to discriminate between normal and cancer cell lines: a feasibility study

Abstract

Excitation emission matrix (EEM) fluorescence spectroscopy combined with the OPLS method has been investigated as a promising tool to discriminate between normal and cancer cell lines in two datasets: (i) using several types of normal and cancer cells (including 3T3, ARPE, HEK, HepG2, HeLa, HT-29 and 786-0 cells); (ii) considering the expression of matrix metalloproteinase-2 and -9 (MMP-2 and MMP-9) in suspensions of HEK and 786-0 cell lines. Partial Least Squares-Discriminant Analysis (PLS-DA) using the score matrix from PARAFAC (Parallel Factor Analysis), UPLS-DA (Unfolded Partial Least Squares with Discriminant Analysis) and orthogonal projection to latent structures (OPLS) were used as the bases for the discrimination models. UPLS-DA presented relevant performance for cancer cells in both datasets, with 100% and 66.7% correct prediction for first and second cases, respectively, and poor discrimination relative to normal cells in the first dataset (25%). By using the OPLS, we achieved 75% correct prediction for normal cells and maintained 100% concordance for cancer objects. On applying OPLS to the second dataset, we obtained 100% correct prediction in both classes (normal and cancer) for calibration and prediction sets. These results suggest that EEM fluorescence spectroscopy combined with chemometrics could be used as a clinical tool for cancer cell detection based on intrinsic biomolecular signatures.