Simplified Fourier-transform mid-infrared spectroscopy calibration based on a spectra library for the on-line monitoring of bioprocesses.

In order to significantly reduce the time involved in mid-infrared spectroscopy calibrations, a novel approach based on a library of pure component spectra was developed and tested with an aerobic Saccharomyces cerevisiae fermentation. Instead of the 30-50 standards that would have been required to build a chemometric model, only five solutions were used to assemble the library, namely one for each compound (glucose, ethanol, glycerol, ammonium and acetate). Concentration profiles of glucose, ethanol and ammonium were monitored with a fair accuracy, leading to standard error of prediction (SEP) values of 0.86, 0.98 and 0.15 g L(-1). Prediction of the two minor metabolites, acetate and glycerol, was less accurate and presented a detection limit of around 0.5 g L(-1). The overall performance of the library-based method proved to be very similar to a 49-standard chemometrics model. The model was shown to be very robust and uncorrelated, since it was able to predict accurately the concentration changes during a spiking experiment. Even though simple, this method allows more advanced calculations, such as determination of the explained variance and detection of unexpected compounds using residuals analysis.