UNRAVELLING SPECTRAL SIGNATURES IN BIOPROCESS DEVELOPMENT

Fresh challenges are now arising for the bio-industries as a consequence of advances in Process Analytical Technologies (PAT) and the resulting generation of high dimensional spectral data. In particular, there is a need for analysis algorithms that realise information that will provide new and enhanced insights into bioprocess development. In this paper, a methodology is proposed which involves the application of Independent Component Analysis (ICA) to spectral signatures, following the appropriate pre-processing of the signals. An evolving window approach is considered to identify the number of key components present and serves to indicate critical operational points, i.e. their limitation or appearance within a batch. Also of interest is the use of the technique as a finger printing tool to help identify differences between batches as they evolve. The critical challenge is to make use of such methods in early stage development where typically only limited data is available. To place the technique proposed in context, the methodology is applied to NIR spectra generated from a pilot scale industrial antibiotic fermentation.