Manufacture of recombinant proteins from. mammalian cell lines requires the use of bioreactor systems at scales of up to 20,000 L. The cost and complexity of such systems can prohibit their extensive use during the process to construct and select the manufacturing cell line. It is therefore common practice to develop a model of the production process in a small scale vessel, such as a shake-flask, where lower costs, ease of handling, and higher throughput are possible. This model can then be used to select a small number of cell lines for further evaluation in bioreactor culture. Here, we extend our previous work investigating cell line construction strategies to assess how well the behavior of cell lines in such a shake-flask assessment predicts behavior in the associated bioreactor production process. A panel of 29 GS-CHO cell lines, all producing the same antibody, were selected to include a mixture of high and low producers from a pool of 175 transfectants. Assessment of this panel in 10 L bioreactor culture revealed wide variation in parameters including growth, productivity, and metabolite utilization. In general, those cell lines which were high producing in the bioreactor cultures had also been higher producing in an earlier shake-flask assessment. However, some changes in rank position of the evaluated cell lines were seen between the two systems. A potential explanation of these observations is discussed and approaches to improve the predictability of assessments used for cell line selection are considered. (C) 2010 American Institute of Chemical Engineers Biotechnol. Prog., 26: 1446-1454, 2010