NETWORK-REGULARIZED HIGH-DIMENSIONAL COX REGRESSION FOR ANALYSIS OF GENOMIC DATA.

Area of Science:

• Genomics
• Biostatistics
• Bioinformatics

Background:

• High-dimensional genomic data analysis often ignores valuable biological network information.
• Existing survival analysis methods lack effective integration of prior network structures.

Purpose of the Study:

• To propose a novel network-based regularization method for high-dimensional Cox regression.
• To leverage prior network information for improved estimation and variable selection in genomic survival analysis.

Main Methods:

• Developed a network-based regularization technique using ℓ1-penalty and quadratic Laplacian penalty.
• Implemented the method with an efficient coordinate descent algorithm.
• Established theoretical guarantees for model selection consistency and estimation bounds.

Main Results:

• The proposed method demonstrates superior variable selection accuracy and stability compared to Lasso and elastic net.
• Theoretical analysis confirms the benefits of incorporating network structural information.
• Identified biologically plausible subnetworks associated with breast cancer metastasis.

Conclusions:

• Network-based regularization offers a powerful approach for high-dimensional genomic survival data.
• This method enhances biological insights by integrating network structures.
• The approach has significant implications for understanding complex diseases like breast cancer.