Principal Component Score Modeling for the Rapid Description of Chromatographic Separations

Abstract

This paper describes the use of Principal Component Analysis (PCA) as a tool for modeling chromatographic separations. PCA is an analytical technique developed to extract key information out of large data sets and to develop relationships and correlations. The basis of the proposed model is the use of PCA to correlate experimental chromatographic data across different process variables or scales. The generated correlations are then used to provide for the simulation of additional chromatographic runs not included in the initial dataset. The approach is demonstrated by application to the cation exchange separation of a four protein component feed comprising ovalbumin, ovatransferrin, lysozyme, and myoglobin. A good fit between modeled and experimental data was found, and the ability of the method to model additional chromatographic separations not within the original dataset is demonstrated. The technique has the potential to accommodate changing system variables such as column dimensions as well as process variables including sample volume and salt gradient. It provides a potentially powerful tool for the rapid investigation of scale-up effects and for the minimization of the material inventories needed for such studies.

Growth and productivity impacts of periplasmic nuclease expression in an Escherichia coli Fab' fragment production strain

“…Basic research insights now frequently inform host cell modifications for boosting titre (Tan et al, 2008) and efficiency of downstream processing (DSP) steps such as capture (Clemmitt and Chase, 2000; Nesbeth et al, 2006) and purification (Humphreys et al, 2004). Combining these cell‐engineering approaches with ultra scale down approaches (Chan et al, 2006; Shapiro et al, 2009) and modeling techniques (Chhatre et al, 2008; Edwards‐Parton et al, 2008), provides a powerful means of optimizing and integrating unit operations to improve whole bioprocesses.…”

Improved process analytical technology for protein a chromatography using predictive principal component analysis tools

“…This alignment method was employed for all UV, conductivity, and pH transition curves in our analysis. Although there are many other scaling methods available (Andersson and Hamalainen, 1994; Bro et al, 1999; Bylund et al, 2002; Edwards‐Parton et al, 2008; Li et al, 2004; Malmquist and Danielsson, 1994; Nielsen et al, 1998; Torgrip et al, 2003; Watson et al, 2006), the advantage of this alignment method over others is that the data in each cycle can be aligned independently, which is essential for online detection.…”

“…PCA has been employed for both analytical and preparative scale chromatographic separations (Edwards‐Parton et al, 2008; Larson et al, 2003; Malmquist and Danielsson, 1994; Pate et al, 1999, 2004; Statheropoulos et al, 1996). Malmquist and Danielsson (1994) analyzed the entire chromatogram when carrying out peptide mapping as a quality control for recombinant protein production.…”

“…Pate et al (1999) have used PCA to predict chromatogram shapes at different column scales and have extended this approach to nonlinear, overloaded conditions (Pate et al, 2004). Edwards‐Parton et al (2008) developed PCA models for the rapid description of chromatographic separations to account for scale‐up effects. However, to date there has been no report on the application of PCA for Protein A chromatography and no report on process yield prediction by analyzing multiple transitions in the chromatograms.…”

Integration of stochastic simulation with multivariate analysis: Short‐term facility fit prediction

“…This approach has been adopted to analyze the near‐infrared spectra of multiple raw material lots so as to understand the impact of raw material variability captured by the spectra on cell culture performance . On the downstream front, PCA has been used to analyze the impact of chromatography operating conditions and scales on chromatogram profiles as well as to generate predictive models for chromatographic separations . Most works on MVA utilize historical datasets.…”