Towards Verified Biological Models

  • A Sadot ,
  • Jasmin Fisher ,
  • D Barak ,
  • Y Admanit ,
  • M.J Stern ,
  • E.J.A Hubbard ,
  • D Harel

IEEE/ACM Transactions on Computational Biology and Bioinformatics | , Vol 5: pp. 1-12

The last several decades have witnessed a vast accumulation of biological data and data analysis. Many of these data sets represent only a small fraction of the system’s behaviour, making the visualization of full system behaviour difficult. A more complete understanding of a biological system is gained when different types of data (and/or conclusions drawn from the data) are integrated into a larger-scale representation or model of the system. Ideally, this type of model is consistent with all available data about the system, and it is then used to generate additional hypotheses to be tested. Computer-based methods intended to formulate models that integrate various events and to test the consistency of these models with respect to the laboratory-based observations on which they are based are potentially very useful. In addition, in contrast to informal models, the consistency of such formal computer-based models with laboratory data can be tested rigorously by methods of formal verification. We combined two formal modelling approaches in computer science that were originally developed for non-biological system design. One is the inter-object approach using the language of live sequence charts (LSCs) with the Play-Engine tool, and the other is the intra-object approach using the language of statecharts and Rhapsody as the tool. Integration is carried out using InterPlay, a simulation engine coordinator. Using these tools, we constructed a combined model comprising three modules. One module represents the early lineage of the somatic gonad of C. elegans in LSCs, while a second more detailed module in statecharts represents an interaction between two cells within this lineage that determine their developmental outcome. Using the advantages of the tools, we created a third module representing a set of key experimental data using LSCs. We tested the combined statechart-LSC model by showing that the simulations were consistent with the set of experimental LSCs. This small-scale modular example demonstrates the potential for using similar approaches for verification by exhaustive testing of models by LSCs. It also shows the advantages of these approaches for modelling biology.