PoLoBag

PoLoBag infers signed, directed gene regulatory interactions and reconstructs gene regulatory networks from gene expression data using an ensemble Lasso regression bagging approach that incorporates polynomial feature expansion to capture higher-order and non-linear relationships.

Key Features:

• Signed Inference: Infers regulatory interactions with direction (regulator to target) and sign (activation or inhibition) for each interaction.
• Ensemble Regression Approach: Uses an ensemble regression strategy within a bagging framework, averaging Lasso weights estimated from multiple bootstrap samples to improve robustness and accuracy.
• Polynomial Features: Incorporates polynomial features into bootstrap samples to model higher-order and non-linear interactions between genes.
• Network Cycles: Infers network cycles to represent reciprocal and feedback regulatory relationships.

Scientific Applications:

• Gene regulatory network reconstruction: Constructs detailed gene regulatory networks (GRNs) from expression data, providing signed and directed interactions.
• Systems biology analyses: Aids analysis of cellular response mechanisms and disease-related regulatory pathways by supplying direction and sign of regulatory effects.
• Benchmarking and validation: Demonstrated accuracy on both simulated and real-world datasets.

Methodology:

Applies Lasso regression to bootstrap samples within a bagging framework and averages Lasso weights across bootstrap replicates; incorporates polynomial feature expansion up to a specified degree (e.g., degree 2) to model non-linear relationships.