Comparison of cancer subtype identification methods combined with feature selection methods in omics data analysis

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate cancer subtype identification is crucial for effective diagnosis and treatment.
• Feature selection is vital for reducing data dimensionality and identifying informative genes for cancer subtyping.
• Combinatorial approaches integrating feature selection and subtype identification methods are underexplored.

Purpose of the Study:

• To identify the optimal combination of feature selection and subtype identification methods for single omics data analysis.
• To evaluate the performance of various feature selection techniques when combined with unsupervised clustering algorithms.
• To provide a guideline for selecting the best methodology based on data characteristics and evaluation criteria.

Main Methods:

• Investigated combinations of six filter-based feature selection methods and six unsupervised subtype identification algorithms.
• Utilized The Cancer Genome Atlas (TCGA) datasets across four cancer types for comprehensive analysis.
• Evaluated method performance using multiple metrics, focusing on p-values and accuracy.

Main Results:

• No single combination consistently outperformed others across all datasets and metrics.
• Consensus Clustering (CC) and Neighborhood-Based Multi-omics Clustering (NEMO) with variance-based feature selection showed promising results (lower p-values).
• Nonnegative Matrix Factorization (NMF) demonstrated improved performance with feature selection, particularly when combined with Similarity Network Fusion (SNF), Monte Carlo Feature Selection (MCFS), and Minimum-Redundancy Maximum Relevance (mRMR).

Conclusions:

• The optimal combination of methods is context-dependent, varying with data type, feature set size, and chosen evaluation metrics.
• A practical guideline is proposed to aid researchers in selecting the most appropriate method combination for their specific cancer subtyping tasks.
• This study highlights the importance of synergistic feature selection and subtype identification strategies in cancer omics data analysis.