The application of the unscented Kalman filter to control starvation-induced programmed cell death-apoptosis-in Chinese hamster ovary cells was investigated. Neural network-based sensitivity analysis identified glutamine and asparagine as two major amino acids that play a key role in the suppression of apoptosis. Dynamic equations that accounted for the dependence of apoptotic cells on the concentrations of viable cells, glutamine, and asparagine were derived. These state equations were highly nonlinear and included nine state variables. An oxygen mass balance was written in the liquid phase. It served as the output equation for the unscented Kalman filter. Using the oxygen uptake rate as the observer, it was possible to estimate the states. A model predictive controller was then implemented once the apoptotic cells in the bioreactor approached a concentration of 1.5 X 10(4) cells/mL, taking into account the operating range of the flow cytometer and measurement error. The manipulated variables were the flow rates of glucose, glutamine, and asparagine. Simulation results showed that the controller was able to keep the apoptotic cells at a concentration of 1.5 x 10(4) cells/mL.