Reverse engineering of gene networks with LASSO and nonlinear basis functions

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Understanding gene regulatory networks (GRNs) is crucial for deciphering cellular functions.
• Reverse engineering GRNs from experimental data presents significant computational challenges.
• Existing methods often struggle to accurately infer causal relationships and network structures.

Purpose of the Study:

• To develop and evaluate a robust algorithm for reverse engineering gene regulatory networks.
• To integrate time-series and steady-state data for comprehensive network inference.
• To assess the algorithm's performance against established benchmarks.

Main Methods:

• An algorithmic pipeline combining ordinary differential equations (ODEs) and parameter estimation via least angle regression.
• Utilizing cross-validation for determining in-degrees and selecting nonlinear transfer functions.
• Employing a bootstrap procedure to score inferred network edges.

Main Results:

• The algorithm successfully generated a complete directed gene regulatory network.
• Inferred network edges were assigned confidence scores using bootstrapping.
• The algorithm demonstrated superior performance in inferring a directed gene-to-gene network in the DREAM2 challenge.

Conclusions:

• The proposed algorithmic approach offers a powerful tool for gene regulatory network reconstruction.
• Integration of multiple computational techniques enhances the accuracy of network inference.
• The algorithm's success in a competitive setting highlights its potential for biological discovery.