Models represent our knowledge, observations and hypotheses in a testable digital framework. Because models are digital, it should be easy to reuse models and reproduce simulation results. However, many dynamic biochemical models are not reusable and many simulation results are not reproducible, including models and simulation results reported in standard formats such as the Systems Biology Markup Language (SBML) and the Simulation Experiment Description Markup Language (SED-ML). This irreproducibility limits the impact of modeling by inhibiting researchers from reusing models and simulation results for additional studies and combining models of individual biological processes into meta-models of entire biological systems.
Currently, models are hard to reuse and simulations are hard to reproduce because (a) few researchers report the metadata needed to reproduce simulations, (b) there are many incompatible simulators, (c) there is no simulation results repository, (d) there is no standard for reducing simulation results, (e) there is no standard for describing results visualizations, and (f) there are inadequate tools for visualizing simulation results.
To address these problems, we will develop novel tools and public servers for (a) using existing simulators to reproducibly simulate a wide range of models and (b) storing and (c) visualizing simulation results:
- We will build a database for storing models, simulation experiments, their results, and their metadata which will mint DOIs and support queries over simulation results. The system will help researchers share and retrieve simulation results and apply big data analytics to simulation results. In turn, the system will help researchers reuse simulation experiments and reproduce simulation results.
- We will build a simulation system which provides a common interface to multiple simulators that each support individual simulation algorithms and modeling domains. This will make it easy for researchers to reuse models and reproduce simulations without having to install domain-specific simulators.
- We will build a web-based system for using the simulation system and simulation results database to interactively simulate and visualize models in a browser. This will enable researchers to retrieve deposited simulation results, request new simulations, and visually analyze simulation results.
To ensure our tools advance biomodeling, we will develop our tools in conjunction with several CPs and SPs which will provide model repositories and journals web-based tools for interactively simulating and visualizing reported models. These CPs will push us to develop user-friendly tools, and we will pull the CPs to require model authors to annotate their simulation experiments so they are reproducible.
To help researchers use our software, we will work with TR&Ds 1 and 2 to combine our software into a reproducible modeling workflow. We will also extensively document our software and distribute it open-source. In addition, as part of the Training and Dissemination Core, we will develop tutorials and organize workshops.