The predictive capacity of a novel population-balance model to simulate aggregation kinetics of attachment-dependent cells at the resolution of one-cell increments has been evaluated. Using spheroid assembly of DU 145 human prostate cancer cells as a representative system, the mathematical model proved to be robust in simulating aggregation over a 5-fold range of surface densities from 5x10(3) to 2.5x10(4) cells/cm(2) with a single matrix of rate constants. For cultures at 1x10(5) cells/cm(2), more than 75% of simulated aggregate concentrations are within the standard deviation of measured concentrations. For the two extreme densities, at least two-thirds of model predictions are within 35% of the mean for experimental data. Error in model predictions is attributed to uncertainty in measurements and intrinsic changes in aggregation. The model has application to the rational design of spheroids in tissue engineering and bioseparation processes in pharmaceutical manufacturing.