Network-constrained Random Lasso for biologically interpretable gene network inference across unequal sample sizes.

Area of Science:

• Systems Biology
• Computational Biology
• Genomics

Background:

• Gene regulatory network inference is crucial for understanding complex diseases.
• Existing methods struggle with high-dimensional, noisy, and imbalanced sample data.
• Accurate inference is vital for identifying disease mechanisms and therapeutic targets.

Purpose of the Study:

• To develop a novel computational framework, netRL, for robust gene regulatory network inference.
• To address limitations of existing methods in handling imbalanced sample sizes and high dimensionality.
• To identify phenotype-specific molecular networks in COVID-19 patients.

Main Methods:

• netRL integrates Random Lasso with prior biological network knowledge.
• A bootstrap strategy stabilizes key regulatory gene selection.
• Network-informed penalization uses centrality measures (hubness, betweenness).
• A hypergeometric test assesses edge significance for enhanced reliability.

Main Results:

• netRL outperforms conventional methods in network estimation and gene selection via simulations.
• Distinct gene networks were identified for asymptomatic and critical COVID-19 cases.
• Asymptomatic networks were dense (ribosomal proteins); critical networks were sparse with key hubs (NFKBIA, B2M, CXCL8, FOS).

Conclusions:

• netRL provides biologically interpretable and statistically robust network inference.
• Findings reveal distinct molecular signatures associated with COVID-19 severity.
• Enhancing asymptomatic markers may offer insights into mitigating disease severity.