Joint skeleton estimation of multiple directed acyclic graphs for heterogeneous population

Area of Science:

• Computational biology
• Statistical genetics
• Network inference

Background:

• Directed acyclic graphs (DAGs) model high-dimensional variable interactions.
• Estimating DAG skeletons (undirected graphs) is possible with observational data.
• Real-world data often comprises mixtures of populations, each with its own DAG.

Purpose of the Study:

• To develop a method for jointly estimating DAG skeletons across multiple populations.
• To handle situations where sample population origin is unknown or partially labeled.
• To improve accuracy and robustness compared to separate estimation or hard labels.

Main Methods:

• A two-step approach for joint DAG skeleton estimation.
• Utilizes probabilistic soft labels for sample population origin.
• Evaluates estimation consistency and performance via simulations.

Main Results:

• The proposed method achieves more accurate and robust DAG skeleton estimation.
• Soft labels enhance accuracy compared to hard labels.
• Demonstrated effectiveness across various simulation settings.

Conclusions:

• The developed method offers a powerful tool for multi-population DAG skeleton inference.
• Applicable to complex biological datasets, such as gene expression data.
• Provides a robust framework for analyzing mixed-population data with potential labeling errors.