Identification and quantitative analysis of salt-adulterated honeysuckle using infrared spectroscopy coupled with multi-chemometrics

Abstract

The quality of honeysuckle is typically degraded seriously by unscrupulous mixing with adulterating substances, especially salt. This study proposes a model based on fourier-transform mid-infrared (Mid-IR) spectroscopy and diode-array near-infrared (NIR) spectroscopy in combination with chemometric techniques for the identification and quantification of salt-adulterated honeysuckle. For model training, Mid-IR and NIR spectra of 72 sample batches (including 32 batches of salt-free samples and 40 batches of salt-adulterated samples) were collected. After that, honeysuckle sample authentication was performed using the partial least squares discriminant analysis (PLS-DA) model based on NIR spectroscopy data. This model was optimized after processing by 28 types of pretreatment schemes, possessed of 100% correct authentication of salt-free and salt-adulterated sample at 0.5%-20% (W/W) levels. Subsequently, PLS regression models were built for quantify salt adulteration levels. These models were optimized by the synergy interval PLS (SiPLS) and backward interval PLS (BiPLS) algorithms. The SiPLS-optimized mid-IR-based PLS model demonstrated the lowest values of two error metrics, namely, a root-mean-square error of estimation (RMSEE) of 2.38%, and a root-mean-square error of prediction (RMSEP) of 3.14%. Then, optimal PLS models were obtained through applying several variable selection methods, including variable importance for projection (VIP), competitive adaptive reweighted sampling (CARS), the successive projection algorithm (SPA), and uninformative variable elimination (UVE). The results indicated that the SiPLS-VIP multi-dimensional analysis methods showed a maximum RMSEP reduction of 1.02% in comparison with models constructed using the BiPLS or SiPLS algorithms alone. In general, the proposed multi-dimensional analysis methods show a great potential to complement existing methodologies for rapid and effective salt adulteration detection in honeysuckle samples.