Systems Biology and Artificial Life: Towards Predictive Modeling of Biological Systems

Abstract

Systems biology has been defined as ‘‘the search for the syntax of biological information, that is, the study of the dynamic networks of interacting biological elements’’ [1]. Based on large, quantitative data sets, systems biology aims to generate models that facilitate an integrative understanding of biological systems that goes beyond increasingly detailed knowledge of the system’s elements. According to a classical definition, ‘‘Artificial Life is a field of study devoted to understanding life by attempting to abstract the fundamental dynamical principles underlying biological phenomena, and recreating these dynamics in other physical media—such as computers—making them accessible to new kinds of experimental manipulation and testing’’ [2]. These definitions show that systems biology and artificial life share the same objective: a principled and comprehensive understanding of living systems. Both interdisciplinary fields employ computational and other formal models of biological systems, and apply mathematical tools to analyze models and complex systems. The use of models is partially complementary: Systems biology focuses on analyzing and understanding experimental data using fairly generic modeling techniques, while artificial life considers rather elaborate and specific computational and other formal models as objects of experimentation, aiming to understand general biological features that are not necessarily represented by quantitative data. Using computational models to generate synthetic data is a recurring element in the contributions to this special issue. Ohno et al. present a model-based investigation of ammonia detoxification by the liver, and Ogawa et al. implemented multiple models of the regulatory networks realizing the Drosophila circadian clock for comparative study. Both these studies use the E-cell simulation framework. Van Leemput et al. use the SynTReN simulator to conduct a comparative study of regulatory network inference algorithms. Methods to analyze data and to infer networks complement synthetic data generation. A novel network inference algorithm based on variational Bayes expectation maximization (VBEM) is introduced by Tienda-Luna et al., and Matsubara et al. present extreme signal flows (ESFs) as a technique to analyze signaling networks. Models of biological systems will increase in complexity in the future. The overlap of systems biology and artificial life can be expected to grow in this process, as formal and computational techniques to model levels of biological organization such as tissues, growth, ecology, and evolution are combined with advanced techniques for data-driven model inference. As model complexity grows, the relative amount of molecular biology data and other prior knowledge to validate model inference methods decreases. Therefore, models that incorporate multiple key levels of biological organization will become increasingly important as a source of realistic synthetic test data. It would be desirable to develop a standard for specifying processes of synthetic data generation, as this would facilitate comparable validation of model inference algorithms as well as comparative studies of models. Biological systems typically interact with other levels of organization, and considering such interactions in modeling and analysis is often crucial for understanding a given biological system. In