Bioprocess modeling
In this example multivariate modeling of gene expression data from
yeast fermentation is presented. Data from VTT (Anne Huuskonen, Heikki
Vuokko, Virve Vidgren, John Londesborough and Jari J. Rautio) was
originally gathered from samples of very high gravity wort
fermentations using Saccharomyces
pastorianus (combining S.
cerevisiae and S. bayanus genes).
Samples
were analysed using the transcript analysis with aid of
affinity capture (TRAC) method. TRAC can be used to create a dynamic
expression picture along the physiological states of observed
cultivations. The expression of selected genes relevant to wort
fermentation at high frequency from several days fermentations were
monitored. Changes in expression during the first hours of
fermentations for several genes affecting maltose metabolism,
glycolysis and ergosterol synthesis seemed to be remarkable.
To find out more about gene interactions during different metabolic
states, multivariate modelling was carried out using PCA and PLS
methods. Score plots formed trajectories from the first hours through
different metabolic states. Gene expression could be used to monitor
fermentation phase changes and product quality. PLS modelling of
fermentation sugars and apparent extract (carbohydrate conversion) are
shown here.
