1Centre of Human & Aerospace Physiological Sciences, King’s College London; 2Respiratory Therapy Area Unit, GlaxoSmithKline Medicines Research. Systems biology is ‘an iterative process of computational model building and experimental model revision with the aim of understanding or simulating complex biological systems’ (PMID23849719). In practice, systems biology rests on three pillars: computational modelling, high-throughput ‘omics methods, and systematic perturbation of the biological system of interest. By experimentally disrupting a biological system, by measuring a broad range of targetted (preferably unbiased) variables and by building and refining computational models describing these (and related) variables, the experimenter aims to uncover the fundamental properties of a system without recourse to a hypothesis. Computational models that can accommodate such a paradigm naturally fall into two categories: those that model the data and those that model the mechanism. In both cases, a recursive approach is recommended, leading to the concept of the ‘model as hypothesis’. In this talk I present a mechanistic modelling approach that can encompass many different types of data and that lends itself to recursive model refinement: constraint-based metabolic modelling, using metabolic network reconstructions. I will outline the principles and assumptions that underlie constraint-based modelling. I will also present case studies showing that constraint-based modelling can aid interpretation of high-throughput data (including metabolomics and transcriptomics) and can predict genes whose mutations are likely to be favoured in a human population under strong selective pressure (hypoxia).