Models, simplified representations of reality, are THE core means for generation of Systems Biology insight, both via theoretical and experimental research. Ideally, models should be i) simple, ii) reproducible, and iii) relevant. Simplicity lies at the heart of the model definition, and it has beneficial effects for ease of maintenance and cost, in terms of resources, specialist skills, time requirements, and research ethics. Reproducibility entails standardised designs that can be re-created with minimal training elsewhere. Relevance requires model selection such that, in spite of all the simplification and standardisation, those key aspects of the biological system are retained that are relevant for the investigated systems’ behaviour [1]. These three parameter dimensions apply, regardless of whether a model is conceptual, computational, experimental, or clinical. As a rule, models should be as simple as possible, yet as complex as necessary to address the question at hand [2]. For Systems Biology applications, in particular those relevant for bio-medical research and development, it is necessary to reach across levels of structural complexity, from molecule to organism, in order to explore causal relationships underlying relevant behaviour. This approach necessarily and consciously combines reduction and integration. It can start at any level (there is no privileged spatial scale, although the cell often is a productive first base [3]), as long as there is sufficient data/insight to conduct hypothesis-driven research. Information from neighbouring levels of structural complexity is crucial in guiding input parameters, in setting boundary conditions, and for subsequent validation of predictions [4]. This approach can be illustrated using an example such as cardiac modelling, say to explore the effect of an ion channel blocker on whole heart electrophysiology or patient ECG properties. The necessarily multi-level approach calls for modular model components, as in computational studies, so in experimental research (e.g. to link insight from single channel patch clamp recordings to whole cell, culture or tissue studies, and on to isolated heart or patient patho-physiological behaviour). A challenge is to define tools for efficient integration across these levels of structural complexity. Often, experimental tools are limited by their given spatial resolution, and many ‘translational’ tools, such as pharmacological probes to modify ion channel behaviour, suffer from limited efficacy and/or specificity (as well as side-effects, often on other organ systems containing similar target proteins). Modelling can allow one to cross these barriers by selectively modifying parameters (whether singly or in controlled groups, with user-defined timing and strength) that may be inaccessible in experimental work. In this context, much insight can be gleaned from simple two-dimensional or gross-anatomically representative three-dimensional models of the heart. However, for mechanistic investigations into the effects of tissue heterogeneity (such as during fibrosis or infarction), additional information on the alignment and micro-structural arrangement of cardiac cells of different populations is needed. Similarly, for many applications it is sufficient to treat individual cardiomyocytes as point-sources of electrical and mechanical activity, but when one wishes to explore molecular mechanisms (say the effect of a drug on ion transport and distribution), intra-cellular heterogeneity and compartmentalisation are key to understanding function. Whole heart models with para-cellular resolution span a spatial range from 10-1 to 10-5 m. The challenges associated with this are similar to those of nanoscopic reconstruction of single cells, going from 10-5 to 10-9 m. Both call for an integrated approach that brings together biological experimentation, structural and functional imaging, image analysis and discretisation, (co-)registration and mesh generation, simulation and visualisation. This lecture will illustrate this process on the example of a recent BBSRC-funded Technology Development Research Initiative focussed on individualised whole heart modelling [5,6], demonstrate pilot work on transferring the expertise acquired to individualised single cell reconstruction [7], and highlight how this paves the way towards novel and predictive approaches for physiology.