Recursive regularization for inferring gene networks from time-course gene expression profiles.

Area of Science:

• Systems Biology
• Bioinformatics
• Computational Biology

Background:

• Inferring gene networks from time-course microarray data uses vector autoregressive (VAR) models to identify gene associations.
• This process is a variable selection problem, with elastic net being a promising but sometimes error-prone method.
• Standard elastic net for VAR modeling increases true positives but also leads to more false positives.

Purpose of the Study:

• To propose a novel regularization method, the recursive elastic net, to enhance gene network inference.
• To reduce false positives in VAR model-based gene network estimation.
• To improve the accuracy and true discovery rate of gene network inference.

Main Methods:

• Incorporated relative importance of VAR coefficients into the elastic net framework.
• Developed a recursive elastic net approach for gradual false positive reduction via updated importance.
• Applied the method to time-course gene expression data, including experimental data from MCF-7 breast cancer cells.

Main Results:

• The recursive elastic net drastically reduces false positives while maintaining a high number of true positives.
• Achieved a two-fold or greater increase in true discovery rate compared to competing methods, even with limited time points.
• Demonstrated effectiveness on experimental gene expression data from breast cancer cells.

Conclusions:

• The recursive elastic net is a powerful and effective tool for inferring gene networks.
• This method significantly improves the accuracy of gene network reconstruction from time-course gene expression profiles.
• Offers a robust solution for reducing errors in biological network analysis.