Rapid calibration of near-infrared spectroscopic measurements of mammalian cell cultivations.

Near-infrared (NIR) spectroscopy is a flexible method that can be employed to noninvasively monitor the concentrations of multiple nutrients and wastes in mammalian cell bioreactors. Development of suitable calibrations can be a labor- and time-intensive process that must be repeated when process conditions are altered significantly. To address this difficulty, we have produced a new approach for generating NIR spectroscopic calibrations that requires significantly less time compared with standard calibration schemes. This method reduces development time from the present level of several weeks to several hours. A small number of experimentally collected spectra serve as inputs to a computational procedure that yields a large number of simulated spectra, each containing both analyte-specific and analyte-independent information. Such simulated spectra may be employed as a calibration set for quantifying analytes in experimentally collected spectra. Spectroscopic measurements of the concentrations of five components (ammonia, glucose, glutamate, glutamine, and lactate) can be accomplished with levels of error similar to those obtained with full experimental calibrations. A key to this process is the utilization of random numbers, which randomizes the influence of natural variations, present in each experimentally collected spectrum, on the resultant composite spectrum. This approach may increase the feasibility of employing NIR spectroscopy to monitor bioreactors and other biological processes subjected to varying operating conditions.