NetMIM: network-based multi-omics integration with block missingness for biomarker selection and disease outcome...

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Integrative multi-omics analysis offers deeper insights into complex diseases than single-platform approaches.
• Existing frameworks often ignore biological feature dependencies and mishandle missing data, reducing statistical power.

Purpose of the Study:

• To develop a network-based integrative Bayesian framework for biomarker selection and disease outcome prediction using multi-omics data.
• To address limitations in current methods by incorporating biological pathway information and handling missing data.

Main Methods:

• Utilized a Dirac spike-and-slab variable selection prior for identifying key biomarkers.
• Incorporated gene pathway information to improve feature selection interpretability.
• Employed a data augmentation approach, inspired by the full Bayesian model with missingness (FBM), to handle block missingness in multi-omics data.

Main Results:

• The proposed framework demonstrated more interpretable feature selection results.
• Achieved more accurate predictions for disease outcomes compared to existing methods.
• Successfully integrated gene pathway information and included subjects with incomplete DNA methylation data.

Conclusions:

• The network-based integrative Bayesian framework provides a robust approach for multi-omics data analysis.
• This method enhances biomarker discovery and disease prediction accuracy by leveraging biological knowledge and effectively managing missing data.
• The framework offers improved interpretability and statistical power in complex disease research.