Learning common and specific patterns from data of multiple interrelated biological scenarios with matrix...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput omics technologies generate vast genome-wide data.
• Current integrative and differential analysis methods have limitations in exploring complex biological scenarios.
• Existing methods often fail to identify combinatorial patterns or require matched data dimensions.

Purpose of the Study:

• To develop a flexible framework, CSMF, that integrates common and differential analysis paradigms.
• To simultaneously reveal common and specific patterns from omics data across interrelated biological scenarios.
• To address limitations of current methods in identifying combinatorial differences and handling unmatched data.

Main Methods:

• Developed a flexible framework named CSMF (Common and Specific patterns via Matrix Factorization).
• Applied CSMF to diverse high-throughput omics datasets including ChIP-seq, RNA-seq, and single-cell RNA-seq.
• Utilized matrix factorization for simultaneous integration and differential analysis.

Main Results:

• Demonstrated CSMF's effectiveness on four distinct biological applications.
• Uncovered novel insights into hidden combinatorial patterns within multi-modal omics data.
• Identified common and specific patterns with significant biological implications across different cell lines, cancer types, and developmental stages.

Conclusions:

• CSMF is a powerful and flexible tool for analyzing omics data from interrelated biological scenarios.
• The framework effectively uncovers both commonalities and differences, providing deeper biological insights.
• CSMF advances the exploration of complex biological systems by integrating disparate data types and analytical approaches.