Multi-omics data integration using time-to event endpoint and supervised Cox penalized regression: a comprehensive...

Area of Science:

• Biomedical Informatics
• Computational Biology
• Genomic Medicine

Background:

• Multi-omics data integration (genomics, transcriptomics, epigenomics, proteomics) offers comprehensive insights into complex diseases like cancer.
• Multi-omics prognostic models enhance patient stratification and personalized prognostication.
• High dimensionality, heterogeneity, and correlations in omics data present significant challenges for predictive modeling, especially in time-to-event analyses.

Purpose of the Study:

• To review and synthesize current methodologies for variable selection and regularization in high-dimensional omics data.
• To focus on the application of these methods to survival outcomes in complex diseases.
• To discuss the trade-offs between interpretability, computational efficiency, and predictive performance of different approaches.

Main Methods:

• Exploration of global penalty approaches (LASSO, Ridge, Elastic Net) for model complexity control.
• Analysis of parallel regression methods for independent omics layer analysis.
• Examination of group regularization (Group LASSO, OSCAR) and hierarchical regression (Priority LASSO, IPF-LASSO) for multicollinearity and prior knowledge integration.
• Review of kernel-based methods (KEN-COX) for nonlinear relationships and dimensionality reduction.

Main Results:

• Various methods offer distinct advantages and disadvantages regarding interpretability, computational efficiency, and predictive performance.
• Global penalty methods control complexity; parallel methods offer robustness but may miss correlations.
• Group and hierarchical methods enhance interpretability and handle multicollinearity, while kernel methods address nonlinearity.

Conclusions:

• Tailored approaches are needed to balance interpretability, efficiency, and performance in multi-omics survival modeling.
• Model transparency and clinical applicability are crucial for successful implementation.
• Future research should refine techniques to better capture the complex interplay of omics data in disease progression and survival.