Metabolic carbon labelling experiments enable (Vallino and Stephanopoulos, 1993; Varma and Palsson, a large amount of extracellular fluxes and intracellular 1994). Strong emphasis was laid on the description and carbon isotope enrichments to be measured. Since the analysis of bidirectional reaction steps and on the docurelation between the measured quantities and the unmentation and exploitation of biological assumptions known intracellular metabolic fluxes is given by bilinear made on intracellular fluxes. A simulation strategy for balance equations, flux determination from this data set requires the numerical solution of a nonlinear inverse labeling experiments and the corresponding computaproblem. To this end, a general algorithm for flux estimational methods were introduced and some general proption from metabolic carbon labelling experiments based erties of labeling systems were derived. In this contribuon the least squares approach is developed in this contrition, we now concentrate on flux estimation from given bution and complemented by appropriate tools for statisexperimental data and the statistical analysis of the tical analysis. The linearization technique usually applied for the computation of nonlinear confidence regions is achieved results. The problems that have to be expected shown to be inappropriate in the case of large exchange due to the bilinear structure of the carbon labelling fluxes. For this reason a sophisticated compactification balance equations with respect to fluxes and fractional transformation technique for nonlinear statistical analylabelling have already been illustrated in Part I. sis is developed. Statistical analysis is then performed by computing appropriate statistical quality measures like output sensitivities, parameter sensitivities and the pa-Available Measurement Data rameter covariance matrix. This allows one to determine the order of magnitude of exchange fluxes in most practi-The experimental details of metabolic carbon labeling cal situations. An application study with a large data set experiments are described in Anderson (1983), Wiechfrom lysine-producing Corynebacterium glutamicum demonstrates the power and limitations of the carbon-ert (1996b), and Wiechert (1996a). The set of measured labelling technique. It is shown that all intracellular fluxes data obtained with such experiments is always subdiin central metabolism can be quantitated without asvided into two parts: sumptions on intracellular energy yields. At the same time several exchange fluxes are determined which is 1. Extracellular metabolite fluxes between the cell inteinvaluable information for metabolic engineering.