Causal and structural modeling

These methods combine observational or intervention data with directed prior knowledge to fit and analyze structural models. They are applicable beyond signaling, including gene regulatory networks and other directed biological systems.

• Introduction to structural equation models in CORNETO
  • Set up the notebook
  • Define the structural equations
  • Generate observations
  • Edit the baseline model
  • Define parent relationships with a CORNETO graph
  • Build a linear model
  • Simulate hard and shift interventions
  • Store simulated data in CORNETO Data
  • Simulate a custom mechanism intervention
  • From simulation to fitting
• Discovering a linear causal signaling network from perturbation data
  • When is this method useful?
  • What CORNETO adds
  • Example: separating the AKT and ERK branches downstream of EGFR
    • Define the candidate biology
    • Define the perturbation experiment
    • Mark interventions in Data
  • Fit one linear DAG to all conditions
  • Which PKN edges were selected?
    • Compare the candidate network with the inferred model
    • Inspect fitted edge metadata
  • Reuse the fitted model for held-out prediction
  • Inspect the in-sample equation reconstructions
  • Check what was actually fitted
  • Adapting the workflow to your experiment
  • Advanced options and formulation
    • Model choices and their exact effects
    • What does min_commodity_coverage mean?
  • Method details
    • Linear structural equations
    • Relationship to SEMs and SCMs
    • How the connectivity constraints work
  • Interpretation and limitations