Data from enzyme production experiments were analysed using different multivariate methods. The data set comprised of 12 objects (3 fungi (¤Aspergillus oryzae, Aspergillus fumigatur, Trichoderma reesei¤) grown on 4 substrates (lenzing and/or wet-oxidisedzylan)) and 12 variables (pH, biomass, 7 enzyme activities (xylanase, zylosidase, arabinosidase, cellulase, acetyl zylan esterase, glucuronidase, feroyl esterase) and 3 hydrolysis efficiencies (reducing suggars at 3 different enzyme loadings)). Principalcomponent analysis (PCA) proved to be an efficient method to obtain an overview of the structure in the data - possibly combined with analysis of variance (ANOVA). Partial least squares regression (PLSR) showed a clear connection between the two differentdata matrices (the fermentation variables and the hydrolysis variables). Hence, PLSR was suitable for prediction purposes. The hydrolysis method at low enzyme level was more suitable for prediction purposes than at higher enzyme levels. Cluster analysiscould almost completely separate the objects according to type of fungi. The fungi grown on lenzing xylan gave problems regarding the clustering according to the type of fungi in the different classification methods (including different distance measures)as well as in PCA and PLSR. Due to the limited size of the data set, a classification methodology was not as appropriate as PCA and PLSR.
Planteproduktion og stofomsætning; Risø-R-1139; Risø-R-1139(EN)