Optimizing data integration improves gene regulatory network inference in Arabidopsis thaliana.

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Gene regulatory network (GRN) inference traditionally relies on gene expression data.
• Integrative strategies enhance GRN inference using complementary prior data.
• Current methods lack robust evaluation of prior data integration intensity.

Purpose of the Study:

• To develop a novel approach for optimizing the integration of prior data in GRN inference.
• To address the challenge of determining the optimal strength of prior data integration for regression-based models.
• To improve the accuracy and robustness of GRN inference.

Main Methods:

• Applied weighted random forest (weightedRF) and weighted LASSO models to Arabidopsis thaliana root nitrate induction data.
• Integrated transcription factor binding motifs as prior information.
• Developed DIOgene, a method using prediction error and a null hypothesis to optimize gene-specific data integration intensity.

Main Results:

• Demonstrated significant diversity in optimal integration intensities across genes.
• Achieved strong performance in minimizing prediction error and retrieving known interactions.
• DIOgene outperformed state-of-the-art methods in GRN inference.
• Successfully identified master regulators of nitrate induction.

Conclusions:

• DIOgene provides a hypothesis-driven, gene-specific approach to optimize prior data integration in GRN inference.
• The method enhances the accuracy and reliability of inferred GRNs.
• Optimizing integration strength is crucial for robust GRN discovery, particularly for complex biological processes like nitrate induction.