dCCA: detecting differential covariation patterns between two types of high-throughput omics data.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Multimodal omics data offers deep biological insights but joint analysis is complex due to high-dimensional variable interactions.
• Interaction patterns in omics data can differ across clinical groups, indicating disease-specific biological processes.

Purpose of the Study:

• To introduce Differential Canonical Correlation Analysis (dCCA) for identifying differential covariation patterns between multivariate data across clinical groups.
• To develop computational tools for sparse selection of variable pairs that maximize differential covariation.

Main Methods:

• Proposed Differential Canonical Correlation Analysis (dCCA) to capture group-specific multivariate covariation.
• Developed algorithms for sparse selection of paired variables to maximize differential covariation.
• Utilized an R package for implementing dCCA, available on GitHub.

Main Results:

• dCCA effectively identifies differential covariation patterns between two sets of multivariate variables across clinical groups.
• Simulation studies confirmed dCCA's superior performance in variable selection and differential correlation recovery.
• Applied dCCA to TCGA Pan-Kidney cohort, revealing differential covariations between noncoding RNAs and gene expressions.

Conclusions:

• dCCA provides a powerful approach for analyzing multimodal omics data to uncover group-specific biological interactions.
• The method facilitates the discovery of disease-related molecular mechanisms by highlighting differential covariation patterns.