Teng Lirong

Optimization of the Fermentation Medium for Epothilones High-Producing Strain Using Response Surface Methodology Coupled with Desirability Function

In this paper, response surface methodology was applied to optimize the fermentation medium for epothilones high-producing strain. Considering the activity of epothilones, desirability function was employed to establish the criterion for the optimization. Suitable carbon sources and nitrogen sources were chosen by single-factor test firstly. Plackett-Burman design combined with the linear model was applied to identify the significant components in the fermentation medium. Based on the results of the Plackett-Burman design, a Box-Behnken design was developed for further optimization. A multivariate quadratic regression model was employed to fit the data of the Box-Behnken design and search for the optimum fermentation medium which was as follows (g·L-1): ammonium acetate 9.00, yeast extraction powder 7.00, sodium carboxymethyl cellulose 0.31, beef extract 4.00, NaCl 4.00, K2HPO4·3H2O 0.40, MgSO4·7H2O 0.20, CaCl2 0.09, FeCl3 0.01. The expected desirability values (Dv) with the optimum fermentation medium was 1.892. Five validation experiments were implemented with the optimum fermentation medium and their average Dv was 1.865. The relative error between the experimental values and the expected value was 1.14%, which indicated that the expected value fits experimental values.

Determination of the Protein Content in Cordyceps militaris by Near Infrared Spectroscopy Quantitative Model Based on Partial Least Squares Method

Partial least squares (PLS) method was applied to establish models with near infrared spectroscopy (NIRS) for quantitative analysis of protein content in Cordyceps militaris. Varies pretreating spectra methods were used for conversion of near infrared spectra in order to remove the noise in spectra. The pretreated spectra were applied to develop PLS quantitative analysis models respectively. Each model was optimized by selecting the most suitable number of factors. The most optimum model was selected depend on the root mean squares of calibration sets calculate by cross-validation (RMSECV) and the root mean squares of predicted sets(RMSEP). RMSECV of the optimum model was 0.0199. Using the optimum model for determination of protein content in Cordyceps militaris, RMSEP was 0.0145.