A means to an end: Validating models by fitting experimental data

Validation of a computational model is often based on accurate replication of experimental data. Therefore, it is essential that modelers grasp the interpretations of that data, so that models are not incorrectly rejected or accepted. We discuss some model validation problems, and argue that consideration of the experimental design leading to the data is essential in guiding the design of the simulations of a given model. We advocate a “models-as-animals” protocol in which the number of animals and cells sampled in the original experiment are matched by the number of models simulated and artificial cells sampled. Examples are given to explain the underlying logic of this approach.