NIR and MIR comparative study of Chilean Brown Seaweeds: Partial Least Square regression as a tool for direct determination of polyphenols and antioxidant capacity

Seaweeds have diverse compounds which have been used on food, cosmetic and pharmacological industry. Among these compounds exist polyphenols, a group of molecules which have acquire a huge importance on the last years because of their antioxidant capacity. This property is used to elaborate ad-value products. Because of its long coastline, Chile, has a great amount of seaweeds. It has been demonstrated that brown seaweeds (Phaeophyceae) are the species with the highest content of polyphenols, particularly, phlorotannins, molecules composed by oligomers of phloroglucinol. Nowadays, determination of polyphenols and antioxidant capacity implies extraction methodologies and spectrophotometric assays; these techniques are time-consuming and susceptible to systematic errors.
Infrared (IR) spectroscopies techniques are fast, easily reproducible and do not require big amounts of samples or complicated pretreatments. IR spectroscopy combined with multivariate calibration methods allowed a fast determination which can be on line with industrial processes. The aim of this work is to evaluate the use of near infrared (NIR) and mid infrared (MIR) and Partial Least Squeares regression (PLSr) for the direct quantification of polyphenolic content and antioxidant capacity of dried Macrocystis pyrifera. Individuals were separated according to their morphological structures, this samples were dried and milled. NIR spectra were acquire by a FT-NIR Bruker, MPA on diffuse reflectance mode with a resolution of 8 cm-1 in a wavenumber range of 12500-4000 cm-1. MIR measurements were carried out by a FT-IR Perkin Elmer, Spectrum BX on transmittance mode using KBr disks. The spectra were measured with a resolution of 8 cm-1 in a wavenumber range of 4000-400 cm-1; CO2 region (2400-2284 cm-1) was eliminated for the analysis. Polyphenols extracts were obtained on semi-continue extraction with acetone:water (70:30 v/v). Polyphenolic content and antioxidant capacity were quantified by Folin-Ciocalteau and radical cationic ABTS*+ method respectively. Preproccesing, transforming and PLSr were realized on Infometrix, Pirouette 4.5 software. PLS models were validated with cross-validation and external-sample prediction, comparing the values of their root mean square errors (RMSEV and RMSEP), regression coefficients (rCAL and rVAL) and residual prediction deviation (RPDc and RPDv). Obtained models showed good results, in case of antioxidant capacity with RMSEV and RMSEP values under 0,04; RPDc and RPDv over 4,0; rCAL>0,99 and rVAL>0,97. For polyphenols content PLS models there were values of RMSEV and RMSEP under 1,3; RPDc and RPDv over 4,0; rCAL>0,99 and rVAL>0,97.
Both NIR and MIR generated similar results with slightly better models generated by NIR. In spite of the closed similarity of the models generated by both techniques, we consider that NIR is a better tool for industry because it is easier and faster than MIR, which allowed a better adaptation for on line industrial processes.