A general model-based causal inference method overcomes the curse of synchrony and indirect effect

Area of Science:

• Systems Biology
• Computational Biology
• Dynamical Systems

Background:

• Model-free methods like Granger Causality struggle to differentiate direct causation from synchrony and indirect effects.
• Existing model-based methods are limited to specific mechanistic models, restricting their application.
• Accurate causal inference is crucial for understanding complex biological and ecological networks.

Purpose of the Study:

• To develop a broadly applicable method for inferring causation in monotonic dynamical systems described by ordinary differential equations (ODEs).
• To create a user-friendly computational package, General ODE-Based Inference (GOBI), for causal inference.
• To overcome the limitations of existing model-free and model-specific inference techniques.

Main Methods:

• Derived a general, easily testable condition for ODE models to reproduce time-series data.
• Developed the General ODE-Based Inference (GOBI) computational package.
• Applied GOBI to various molecular and population-level networks with positive and negative regulations.

Main Results:

• GOBI successfully inferred positive and negative regulations in diverse biological networks.
• The method demonstrated superior accuracy compared to existing model-free inference approaches.
• GOBI is applicable to a wide range of monotonic systems described by ODEs.

Conclusions:

• GOBI provides an accurate and broadly applicable method for causal inference in complex dynamical systems.
• This tool enhances the understanding of regulatory mechanisms in molecular and population dynamics.
• GOBI represents a significant advancement over traditional causal inference techniques.