A group LASSO-based method for robustly inferring gene regulatory networks from multiple time-course datasets

Area of Science:

• Systems Biology
• Bioinformatics
• Computational Biology

Background:

• Gene regulatory networks (GRNs) are crucial for understanding cellular processes.
• Reverse engineering GRNs from gene expression data is challenging.
• Integrating multiple time-course datasets can improve GRN inference.

Purpose of the Study:

• To develop a robust method for inferring GRNs from multiple time-course gene expression datasets.
• To address the challenge of noise and outliers in gene expression data.
• To improve the accuracy and reliability of gene regulatory network inference.

Main Methods:

• Proposed a novel method: Huber group LASSO.
• Developed an efficient algorithm combining auxiliary function minimization and block descent.
• Adapted stability selection for robust network topology identification.

Main Results:

• Huber group LASSO demonstrated superior performance compared to group LASSO.
• Outperformed group LASSO in both Area Under the Receiver Operating Characteristic Curve (AOC) and Area Under the Precision-Recall Curve (APC).
• Showcased robustness to large errors and outliers in gene expression data.

Conclusions:

• The developed algorithm guarantees convergence to the optimal solution.
• Huber group LASSO effectively integrates multiple time-course datasets.
• The method enhances resistance to noise and outliers in gene expression data analysis.