Pattern fusion analysis by adaptive alignment of multiple heterogeneous omics data

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Integrating multi-omics data is crucial for understanding complex diseases.
• Heterogeneous data from different platforms and reliability levels pose significant challenges.
• Existing methods struggle to extract consistent information across noisy, multi-view datasets.

Purpose of the Study:

• To develop a novel framework, Pattern Fusion Analysis (PFA), for effective multi-omics data integration.
• To automatically align information and correct biases in heterogeneous omics data.
• To fuse local sample-patterns into a global sample-pattern for comprehensive disease analysis.

Main Methods:

• Pattern Fusion Analysis (PFA) framework.
• Automated information alignment and bias correction.
• Optimal adjustment of data type effects to identify significant sample-patterns.

Main Results:

• PFA successfully captures intrinsic sample clustering structures in synthetic datasets, outperforming state-of-the-art methods.
• The framework provides an automatic weighting scheme for data type and sample contributions.
• PFA reveals shared and complementary sample-patterns in CCLE datasets and identifies distinct cancer subtypes in TCGA datasets.

Conclusions:

• PFA offers a robust approach for multi-omics data integration, handling noise and data heterogeneity.
• The method enhances the discovery of disease-related patterns and subtypes.
• PFA outperforms existing methods in capturing underlying biological structures and clinical relevance.