Inference of biological networks using Bi-directional Random Forest Granger causality

Area of Science:

• Computational biology
• Neuroscience
• Bioinformatics

Background:

• Ordinary least squares Granger causality is standard for time series causal inference.
• High-dimensional data challenges existing Granger causality methods.
• Need for advanced techniques to analyze complex biological networks.

Purpose of the Study:

• To propose a novel Granger causality method for high-dimensional time series data.
• To enhance causal discovery by utilizing time series data bidirectionally.
• To validate the proposed method on simulated and real biological datasets.

Main Methods:

• Developed Bi-directional Random Forest Granger causality.
• Employed random forest regularization for feature selection.
• Reversed time series data to capture additional causal information.

Main Results:

• Demonstrated effectiveness on simulated datasets.
• Successfully applied to functional Magnetic Resonance Imaging (fMRI) data for brain network mapping.
• Applied to HeLa cell data for gene network inference in cancer.

Conclusions:

• Bi-directional Random Forest Granger causality is effective for high-dimensional causal discovery.
• The method provides insights into complex biological systems like brain and gene networks.
• Offers a robust approach for analyzing time series data in various scientific domains.