A Protocol for Model Validation and Causal Inference from Computer Simulation

Abstract

The philosophical literature on modelling is increasingly vast, however clear formal analyses of computational modelling in systems biology are still lacking. We present a general, theoretical scheme which (i) visualizes the development and repeated refinement of a computer simulation, (ii) explicates the relation between different key concepts in modelling and simulation, and (iii) facilitates tracing the epistemological dynamics of model validation. To illustrate and motivate our conceptual scheme, we analyse a case study, the discovery of the functional properties of a specific protein, E-cadherin, which seems to have a key role in metastatic processes by way of influencing cell growth and proliferation signalling. To this end we distinguish two types of causal claims inferred from a computer simulation: (i) causal claims as plain combinations of basic rules (capturing the causal interplay of atomic behaviour) and (ii) causal claims on the level of emergent phenomena (tracing population dynamics). In formulating a protocol for model validation and causal inference, we show how, although such macro-level phenomena cannot be subjected to direct causal tests qua intervention (as, e.g., formulated in interventionist causal theories), they possibly suggest further manipulation tests at the basic micro-level. We thereby elucidate the micro-macro-level interaction in systems biology.