Multi-view based integrative analysis of gene expression data for identifying biomarkers

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Publicly available gene expression datasets from microarray technology are abundant but present analysis challenges.
• High noise, high dimensionality, batch effects, and low biomarker reproducibility hinder effective gene expression data analysis.
• Integrative analysis offers potential but current methods have limitations.

Purpose of the Study:

• To develop a novel integrative framework, MVIAm, for analyzing multiple gene expression datasets.
• To improve the integration, classification, and biomarker identification from microarray data.
• To address the inherent challenges in gene expression data analysis for enhanced clinical applications.

Main Methods:

• Designed MVIAm (Multi-View based Integrative Analysis of microarray data) framework.
• Applied multiple cross-platform normalization methods for multi-view dataset aggregation.
• Utilized Multi-View Self-Paced Learning (MVSPL) for robust gene selection in cancer classification.

Main Results:

• Demonstrated MVIAm's capabilities using simulated data and real-world breast and lung cancer datasets.
• MVIAm effectively addresses noise, dimensionality, batch effects, and reproducibility issues.
• The framework shows flexibility and effectiveness in gene expression data analysis.

Conclusions:

• MVIAm provides a systematic approach to integrative microarray analysis.
• The proposed model enhances the identification of significant biomarkers.
• MVIAm expands the application range of microarray technology in biological systems and clinical settings.