Evaluation and prediction of the antioxidant activity of Epimedium from multi-wavelength chromatographic fingerprints and chemometrics

Epimedium herbs are commonly used as herbal medicine ingredients in China. In this study, firstly the total phenolic content and total flavonoid content (TFC) of Epimedium extracts were determined, respectively. The majority of the Epimedium extracts showed strong 2,2-diphenyl-1-picrylhydrazyl (DPPH) scavenging activity, and TFC of Epimedium exhibited highly correlations with DPPH scavenging activity with high correlation coefficient (R2 = 0.9279). Then, high performance liquid chromatography fingerprints of 25 Epimedium samples (5 species) were developed. A data fusion strategy was used to construct multi-wavelength fingerprints by combining 3 profiles with peak-rich wavelengths (230, 270 and 310 nm) into a single vector for each sample. Whittaker smoother, adaptive iteratively reweighted penalized least square (airPLS) and multi-scale peak alignment methods were adopted to preprocess the constructed fingerprint matrix. Variable selection and partial least square were applied to build a calibration model between fingerprints and their antioxidant activities. The potential peaks, which respond for the antioxidant activity, were screened by the selected variable with large regression coefficient. These results suggest that water extracts of Epimedium are a source of antioxidant compounds, and proposed multi-wavelength fingerprints can be used for predicting the antioxidant activity of Epimedium.