Bioprocess development requires parameters such as yield and purity to be quantitatively linked to operating conditions. The determination of such parameters can involve substantial analytical efforts resulting in bottlenecks. Here, we demonstrate two approaches, based on data reduction techniques, to circumvent such bottlenecks for challenging protein mixtures. In the first approach, a multivariate analysis technique, based on partial least squares applied to absorbance spectra, is employed as a highly parallelized assay and is deployed routinely for the analysis of multicomponent adsorption isotherms. However, this approach lacks the resolution to quantify closely related product variants, which is often vital in commercial contexts. Even traditional assays with high resolution may not yield the required results. For such cases, we demonstrate a second approach based on a problem-specific deconvolution technique with exponentially modified Gaussian functions. This is applied to a model protein with multiple variants, representing a difficult separation, and we show how it can be used as an analytical tool yielding the exact composition of fractions in preparative chromatographic separations. This approach proved to be substantially better than standard automated deconvolution using computer packages. Both approaches represent viable solutions for different areas within the typical operating space for bioprocessing.

Facilitating the analysis of complex samples using data reduction techniques: Case studies in downstream processing

251st National Meeting (2016)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.