Evaluation and comparison of multi-omics data integration methods for cancer subtyping

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Computational integrative analysis is crucial for data-driven biological research.
• Evaluating cancer subtyping methods is challenging due to the absence of gold standards.
• The impact of data type selection on integrative study performance requires further investigation.

Purpose of the Study:

• To comprehensively evaluate ten representative cancer subtyping integration methods.
• To assess the influence of different omics data types and their combinations on subtyping performance.
• To identify optimal multi-omics data combinations for cancer subtyping.

Main Methods:

• Constructed benchmarking datasets for nine cancers using four multi-omics data types and all eleven combinations.
• Evaluated ten integration methods based on accuracy (clustering and clinical significance), robustness, and efficiency.
• Analyzed the impact of individual and combined omics data on cancer subtyping.

Main Results:

• Contrary to intuition, integrating more omics data can negatively impact cancer subtyping performance.
• Performance varied significantly depending on the omics data types and cancer types studied.
• Identified several effective multi-omics data combinations beneficial for most cancers.

Conclusions:

• The optimal number and type of omics data for cancer subtyping are context-dependent.
• Researchers should carefully select omics data combinations rather than assuming more is always better.
• Findings provide practical guidance for selecting effective multi-omics data integration strategies in cancer research.