 |
Center for Nonlinear Studies |
Models of cellular regulation often consist of large and intricate networks of interactions at the molecular scale. They are hard to interpret. Further, since individual interaction parameters, and the existence of interactions themselves, are usually difficult to measure directly, they are often inferred using statistical fits. An alternative approach is to use less detailed and easier to interpret models, with complexity appropriately matched to the amount of data and the problem in question. I will discuss our recent results on efficient inference of coarse-grained phenomenological models of cellular networks from data. I will argue that such models can make accurate predictions even when microscopic details of the system are not well understood. In fact, they may have a better predictive power than microscopically accurate models. Our approach is tested on synthetic models of various complexities and on experimental data.