Predicting how a system behaves under changing conditions is an essential component of science and engineering. The ability to make accurate predictions about the system indicates that it is well understood and provides the opportunity to simulate the response to conditions that would be empirically difficult or impossible to test. In the life sciences, the term systems biology was introduced to articulate the notion that the molecular and phenotypic response of a cell or organism to perturbations is the result of interplay of a multitude of molecules. The ability to predict the behavior of such complex molecular systems remains challenging and inevitably requires the involvement of different types of models and data that support them. In this article, we discuss a range of data-driven models that have proven particularly useful for predicting the behavior of biological systems at different levels of complexity and the matching data generation methods that support them. We specifically focus on predictions based on protein or proteome data generated by mass spectrometry. We describe three case studies that represent frequently encountered situations in systems biology.