Penalized discriminant analysis for the detection of wild-grown and cultivated Ganoderma lucidum using Fourier transform infrared spectroscopy

• A successful combination of FTIR and penalized discriminant method to distinguish wild grown G.lucidum from cultivated one.
• Penalized discriminant models enabled an automatic selection of a small number of informative wavelengths for discrimination.
• The Elastic net model outperformed the PCDA and PLSDA with reduced model complexity and excellent classification accuracy.
• Results provided scientific evidence and interpretation of the major bioactive compounds in G.lucidum regarding its anti-cancer effects.

An effective and simple analytical method using Fourier transform infrared (FTIR) spectroscopy to distinguish wild-grown high-quality Ganoderma lucidum (G. lucidum) from cultivated one is of essential importance for its quality assurance and medicinal value estimation. Commonly used chemical and analytical methods using full spectrum are not so effective for the detection and interpretation due to the complex system of the herbal medicine. In this study, two penalized discriminant analysis models, penalized linear discriminant analysis (PLDA) and elastic net (Elnet),using FTIR spectroscopy have been explored for the purpose of discrimination and interpretation. The classification performances of the two penalized models have been compared with two widely used multivariate methods, principal component discriminant analysis (PCDA) and partial least squares discriminant analysis (PLSDA). The Elnet model involving a combination of L1 and L2 norm penalties enabled an automatic selection of a small number of informative spectral absorption bands and gave an excellent classification accuracy of 99% for discrimination between spectra of wild-grown and cultivated G. lucidum. Its classification performance was superior to that of the PLDA model in a pure L1 setting and outperformed the PCDA and PLSDA models using full wavelength. The well-performed selection of informative spectral features leads to substantial reduction in model complexity and improvement of classification accuracy, and it is particularly helpful for the quantitative interpretations of the major chemical constituents of G. lucidum regarding its anti-cancer effects.

Figure optionsDownload as PowerPoint slide