Co-expression network multivariate regression.

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Genetics

Background:

• High-dimensional omics data analysis requires accounting for variable dependence for accurate statistical inference.
• Omics variables often exhibit complex, structured correlation patterns that are not fully addressed by existing methods.
• There is a methodological gap in omics data analysis for explicitly handling structured dependence, particularly in differential expression analysis.

Purpose of the Study:

• To propose a novel method, Coexpression network multivariate Regression (CoReg), that integrates co-expression network structure into multivariate regression.
• To address the challenge of accounting for inter-correlations among omics variables in statistical analyses.
• To improve the accuracy and replicability of statistical inference in omics data.

Main Methods:

• CoReg integrates co-expression network structures into multivariate regression models.
• The method precisely accounts for inter-correlations (dependence) among omics variables.
• Simulations were used to evaluate the performance of CoReg.

Main Results:

• CoReg substantially improves the accuracy of statistical inference in omics data analysis.
• The method enhances the replicability of findings across different studies.
• Simulations demonstrated the effectiveness of CoReg in handling structured dependence.

Conclusions:

• CoReg offers an alternative approach for omics data association analysis with dependence adjustment.
• The method is analogous to mixed-effects models for handling dependence in lower-dimensional settings.
• CoReg addresses a critical methodological gap in omics data analysis, leading to more reliable results.