Adaptive Calibration Scheme for Quantification of Nutrients and Byproducts in Insect Cell Bioreactors by Near-Infrared Spectroscopy

Abstract

Spectroscopic methods are gaining in popularity in biotechnology because of their ability to deliver rapid, noninvasive measurements of the concentrations of multiple chemical species. Such measurements are particularly necessary for the implementation of control schemes for cell culture bioreactors. One of the major challenges to the development of spectroscopic methods for bioreactor monitoring is the generation of accurate and robust calibration models, particularly because of the inherent variability of biological processes. We have evaluated several methods of building calibration models, including synthetic calibrations and medium spiking methods. The approach that consistently produced reliable models incorporated samples removed from a bioreactor that were subsequently altered so as to increase the sample variation. Several large volume samples were removed from a bioreactor at varying time points and divided into multiple aliquots to which were added random, known amounts of the analytes of interest. Near-infrared spectra of these samples were collected and used to build calibration models. Such models were used to quantify analyte concentrations from independent samples removed from a second bioreactor. Prediction errors for alanine, glucose, glutamine, and leucine were 1.4, 1.0, 1.1, and 0.31 mM, respectively. This adaptive calibration method produces models with less error and less bias than observed with other calibration methods. Somewhat more accurate measurements could be attained with calibrations consisting of a combination of synthetic samples and spiked medium samples, but with an increase in calibration development time.

Online Measurement of Urea Concentration in Spent Dialysate during Hemodialysis

“…That the values are nonzero is primarily a consequence of the partial correlation among the analytes in the calibration data set, which is inherent in a data set such as this. The orthogonality of the calibration model could be enhanced in future experiments by randomly adding analytes to calibration samples to artificially increase the concentration ranges of creatinine, glucose, and lactate (18 ). It is also possible to mathematically force the calibration spectrum to be orthogonal to the interfering analytes by use of measured single-component absorbance spectra.…”

Nondestructive near‐infrared spectroscopic measurement of multiple analytes in undiluted samples of serum‐based cell culture media

“…The analytical utility of NIR spectroscopy is established for measuring millimolar levels of various biological species in simulated matrices Burmeister et al, 1998;Burmeister and Arnold, 1999) and natural matrices from clinical chemistry (Hazen et al, 1998), fermentation (Cavinato et al, 1990;Ge et al, 1994;Hall et al, 1996;Yano et al, 1998), and cell culture Lewis et al, 2000;McShane and Cote , 1998;Riley et al, 1997Riley et al, , 1998aRiley et al, , 1998bZhou et al, 1995) systems. The chemical information originates from overtone and combination transitions associated with fundamental C±H, O±H, and N±H vibrations.…”

“…In addition, proper calibration models demand that analyte concentrations are not correlated with concentrations of any other species within the chemical matrix (ASTM Practice E 1655-94). Samples collected during cell cultivation will naturally possess a metabolism-induced concentration correlation between cellular substrates and metabolic waste products (Riley et al, 1998a). Indeed, the well-known relationship between glucose consumption and lactate production generates a concentration correlation that is unacceptable for building PLS calibration models.…”

“…Indeed, the well-known relationship between glucose consumption and lactate production generates a concentration correlation that is unacceptable for building PLS calibration models. The implications of cell metabolism on method calibration have been examined in detail elsewhere (Lewis et al, 2000;McShane and Cote , 1998;Riley et al, 1997Riley et al, , 1998aRiley et al, , 1998b. Brie¯y, the calibration samples are best collected at various timepoints during cultivation in order to account for the global chemical changes in the samples.…”

“…An adaptive calibration procedure has been proposed to destroy concentration correlations within a set of samples (Riley et al, 1997(Riley et al, , 1998a(Riley et al, , 1998b. This procedure involves spiking samples with known and random amounts of standard materials, thereby adapting the collected samples so they may be used to generate NIR spectra suitable for training multivariate calibration models.…”