PoLoBag: Polynomial Lasso Bagging for signed gene regulatory network inference from expression data.

Area of Science:

• Systems Biology
• Bioinformatics
• Computational Biology

Background:

• Inferring gene regulatory networks (GRNs) from expression data is crucial for understanding cellular mechanisms.
• Existing GRN inference algorithms often struggle to simultaneously determine edge direction and sign (activation/inhibition), and detect network cycles.

Purpose of the Study:

• To develop a novel algorithm for signed GRN inference that accurately determines edge direction, sign, and network cycles.
• To address the limitations of current GRN inference methods.

Main Methods:

• Proposed Polynomial Lasso Bagging (PoLoBag), an ensemble regression algorithm utilizing a bagging framework.
• Averaged Lasso weights from bootstrap samples incorporating polynomial features to capture higher-order interactions.

Main Results:

• PoLoBag demonstrated superior accuracy in signed GRN inference compared to state-of-the-art algorithms.
• The algorithm successfully inferred edge directions, signs, and network cycles in both simulated and real-world expression datasets.

Conclusions:

• PoLoBag offers a robust and accurate solution for signed GRN inference, enhancing the understanding of gene regulatory mechanisms.
• The developed algorithm overcomes key limitations of existing methods, providing more comprehensive network information.