DeepProg: an ensemble of deep-learning and machine-learning models for prognosis prediction using multi-omics data.

Area of Science:

• Computational biology
• Genomics
• Machine learning

Background:

• Multi-omics data offer valuable prognostic information but pose integration challenges.
• Accurate patient survival prediction is crucial for personalized cancer treatment.

Purpose of the Study:

• To introduce DeepProg, an ensemble framework for robust multi-omics data integration and patient survival subtype prediction.
• To evaluate DeepProg's performance against existing multi-omics integration methods.

Main Methods:

• Developed DeepProg, an ensemble framework combining deep learning and machine learning.
• Applied DeepProg to multi-omics datasets from liver and breast cancer.
• Performed pan-cancer analysis to identify common genomic signatures associated with survival outcomes.

Main Results:

• DeepProg robustly predicts patient survival subtypes, identifying two optimal subtypes in most cancers.
• Achieved superior risk-stratification compared to other multi-omics integration methods.
• Demonstrated high predictive accuracy in liver (C-index 0.73-0.80) and breast cancer (C-index 0.68-0.73) datasets.
• Identified genomic signatures related to extracellular matrix modeling, immune deregulation, and mitosis in poor survival subtypes.

Conclusions:

• DeepProg offers a powerful approach for integrating multi-omics data to predict cancer patient survival.
• The framework enhances risk stratification and provides insights into the molecular mechanisms underlying cancer progression.
• DeepProg is publicly available, facilitating its application in cancer research and clinical settings.