Modeling languages

The metaphor underlying our research is that biological systems can be modeled by a set of interacting programs. Following this inspiration, we design new primitives and languages that allow us to easily program (algorithmically model) biological systems and execute (not just simply solve) them. Our research aims at extending the definition of the primitives and formalisms we have already defined to incorporate the basic features of biology that have emerged through the application of the current developments to real case studies.

We focus on enablers of modular and compositional modeling processes to effectively construct models of systems characterized by interactions of multiple biological networks, in which sub-models can be abstracted/refined at various levels of detail. We design and include language primitives to explicitly define dynamic changes in modules that result from the evolution/mutation of biological elements. Moreover, we consider spatial modeling by defining constructs that specify the position and movement of biological elements along 2D/3D space grids (lattices), compartments and more general structures such as graphs. We explicitly represent dynamical processes that depend on topological aspects, such as those of tissue biology and cell morphogenesis.

The resulting modeling language will possess a rich set of features out of which biologists will select those required to fully define their models. This modeling language will also allow for user-definable extensions to further customize the language to the specific characteristics of the class of problems at hand.

Model analyses

Systems biology studies the dynamics of very complex systems, from multi-level networks to genome-wide systems, from organisms to populations. The state-based abstractions commonly used to handle such complexity stretch the capabilities of simulation tools; the passive exploration of the trajectory space with stochastic approaches as well as the single solution obtained through deterministic methods may not be sufficient to provide satisfactory answers for the complete characterization of the dynamics of a system under changing environmental conditions.

Verification techniques developed in computer science for concurrent and distributed systems can be exploited in various ways to check the properties of biological systems. We develop approaches for biological systems to address logical reasoning, causal analysis and analytical solutions of models. Furthermore, we address the analysis of synthetic data produced by our in-silico experiments to infer new knowledge on the modeled phenomena.

We exploit analysis techniques to define abstractions on which to determine whether certain properties hold of the modeled systems and to extract the bits of information that are essential for providing answers to biological questions not affected by the statistical uncertainty incurred by simulation. The capabilities of model analyses will be coupled with the sensitivity analyses of models in order to limit as much as possible the need for time-consuming, multi-run simulations used by brute force approaches to parameter optimization.

Model simulation

Simulation plays a major role among the currently adopted solution techniques of biological models. Both the numerical integration of systems of ordinary differential equations and the event-driven generation of state space trajectories of stochastic models provide powerful tools for exploring system dynamics.

We compare various approaches to simulation to determine their respective merits and applicability to the study of complex biological systems, also taking the spatial dimension of system dynamics into account. We define algorithms that take advantage of the temporal/spatial hierarchies of biological systems to shorten the computation time of simulations. To reach the same goal, we explore parallel simulation algorithms onto multi-processors computing platforms, so that larger instances of system models can be dealt with.

The simplest cases of models, whose dynamical evolution is primarily confined in the temporal dimension, provide the basis for evaluating the relative merits of the various simulation approaches. Since spatial information is becoming crucial in biological investigations, we define the necessary extensions to simulation algorithms, so as to allow the explicit addition of diffusion and movement processes. Complex biological systems characterized by the interaction of multiple components at different temporal/spatial scales will be used to draw requirements for the hierarchical simulation approaches.

Platform interfaces

The availability of a number of user and data interfaces that allow simplifying and, where possible, automating the modeling tasks is mandatory for our platform to become an effective tool for studying in-silico biological systems. The users of the platform will be provided with a coherent set of functionalities that can be exercised on the basis of the specific needs of the system being investigated.

We define interfacing mechanisms that allow model definition through a friendly syntax user-language that includes abstractions close to the typical narrative semi-formal style with which biologists define systems, still ensuring that a formal semantic of models exists for unambiguous interpretation and execution. We also design and implement the import/export of data and models to/from standard repositories, the access to aggregation and statistical manipulation of data applications to infer new knowledge, as well as to those that automate the generation and execution of multiple model evaluations. Flexible and intuitive interfaces to command the display of model evaluation results (raw and aggregate) are part of the platform.