A Regularized Cox Hierarchical Model for Incorporating Annotation Information in Predictive Omic Studies

Area of Science:

• Bioinformatics
• Statistical Genetics
• Computational Biology

Background:

• High-dimensional omics data often includes valuable meta-features like biological pathways and functional annotations.
• These meta-features can enhance prediction accuracy for time-to-event outcomes.
• Integrating external summary statistics from similar studies is crucial for improving predictive models.

Purpose of the Study:

• To introduce a regularized hierarchical framework for integrating meta-features.
• To improve prediction and feature selection performance for time-to-event outcomes.
• To handle high-dimensional omics data effectively by incorporating external information.

Main Methods:

• A hierarchical framework was developed to incorporate meta-features.
• Regularization was applied to both omics and meta-features to manage high dimensionality.
• The hierarchical Cox model was efficiently fitted using iterative reweighted least squares and cyclic coordinate descent.

Main Results:

• The regularized hierarchical model substantially improved prediction performance compared to standard Cox regression when meta-features were informative.
• Applications to breast cancer and melanoma survival data demonstrated improved prediction by incorporating meta-features.
• The model facilitated the discovery of important omics feature sets through sparse regularization at the meta-feature level.

Conclusions:

• The regularized hierarchical regression model effectively integrates external meta-feature information for time-to-event outcomes.
• The framework demonstrates improved prediction performance with informative meta-features and robust performance with uninformative ones.
• The model is valuable for both developing predictive signatures and for discovery applications aimed at identifying key outcome-associated features.