Scientific Modeling Workflows

Terarium supports scientific decision making by helping users organize, refine, and communicate the results of their modeling processes. Users can:

  • Gather existing knowledge.
  • Break down complex scientific operations into separate, easy-to-configure tasks.
  • Create reproducible visual representations of how resources, processes, and results chain together.
Table of contents
  1. Building transparent, reproducible workflows
    1. Templates
  2. Configuring complex scientific tasks

Building transparent, reproducible workflows

The Terarium workspace is a visual canvas for building and capturing modeling processes. Workflows show how resources (models, datasets, and documents) move between different operators to produce results.

Terarium resource panel and a workflow for calibrating an SEIRHD model to LA country COVID data

Each box is a resource or an operator that handles a task like transformation and simulation. They have a title, thumbnail preview, and optional annotations for capturing relevant context. Users can chain the outputs and inputs of resources and operators to:

  • Recreate, reuse, and modify existing models and datasets to suit their modeling needs.
  • Rapidly create scenarios and interventions by configuring, validating, calibrating, and optimizing models.

Templates

Sensitivity analysis template configured to explore the outcomes of an SEIRHDV model

Workflow templates streamline the process of building common modeling workflows. They provide pre-configured and linked resources and operators tailored to user objectives, such as analyzing uncertainty, forecasting potential outcomes, or comparing intervention strategies. Available templates include:

  • Situational awareness: Calibrate a model to historical data to obtain the best estimate of parameters for the present and then forecast into the near future.
  • Sensitivity analysis: Configure a model with parameter distributions that reflect all the sources of uncertainty and then simulate into the near future.
  • Decision making: Simulate a baseline scenario (with no interventions) and various scenarios with intervention policies and then show the relative impact of each policy compared to the baseline.
  • Horizon scanning: Configure a model to represent the extremes of uncertainty for some parameters and then simulate into the near future with different intervention policies and compare the outcomes.
  • Value of information: Configure a model with parameter distributions that reflect all the sources of uncertainty and then simulate into the near future with different intervention policies.
  • Reproduce models from literature: Create models from extracted equations, configure them using extracted values, simulate to reproduce results, and—if multiple models are created—compare them.
  • Calibrate an ensemble model: Simulate and calibrate several models individually and then calibrate the ensemble.

Configuring complex scientific tasks

Resources and operators in the workflow graph summarize the data and inputs/outputs that they represent. Users can drill down to view more details or settings.

Terarium’s operators support various ways for users to configure complex scientific tasks, including:

  • A guided wizard for quickly configuring common settings.
  • A notebook for direct coding.
  • An integrated AI assistant for creating and refining code even if the user doesn’t have any programming experience.

Available operators include:

  • Modeling
    • Create model from equations: Build a model using LaTeX expressions or equations extracted from a paper.
    • Edit model: Modify model states and transitions using an AI assistant.
    • Stratify model: Divide populations into subsets along characteristics such as age or location.
    • Compare models: Generate side-by-side summaries of two or more models or prompt an AI assistant to visually compare them.
  • Config and intervention
    • Configure model: Edit variables and parameters or extract them from a reference resource.
    • Validate configuration: Determine if a configuration generates valid outputs given a set of constraints.
    • Create intervention policy: Define intervention policies to specify changes in state variables or parameters at specific points in time.
  • Simulation
    • Simulate: Run a simulation of a model under specific conditions.
    • Calibrate: Determine or update the value of model parameters given a reference dataset of observations.
    • Optimize intervention policy: Determine the optimal values for variables that minimize or maximize an intervention given some constraints.
    • Simulate ensemble: Run a simulation of multiple models or model configurations under specific conditions.
    • Calibrate ensemble: Extend the calibration process by working across multiple models simultaneously.
  • Data
    • Transform dataset: Modify a dataset by explaining your changes to an AI assistant.
    • Compare datasets: Compare the impacts of two or more interventions or rank interventions.