Pattern-centric transformation of omics data grounded on discriminative gene associations aids predictive tasks in...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• The increasing volume of omics data presents challenges in understanding complex disease mechanisms.
• High dimensionality and feature interplay in omics data hinder effective analysis and prediction.
• Existing methods often require custom algorithms for extracting meaningful biological insights.

Purpose of the Study:

• To develop a novel data transformation method for omics data analysis.
• To shift focus from individual molecules to statistically relevant co-expression patterns (modules).
• To improve the interpretability and predictive power of omics data analysis.

Main Methods:

• Proposed a transformation from molecule-centric to pattern-centric data space.
• Utilized biclustering to identify statistically relevant co-expression patterns.
• Generated new, interpretable variables based on these patterns.

Main Results:

• Achieved significant dimensionality reduction, up to 99%.
• Demonstrated increased predictive performance across various classifiers.
• Validated the approach across multiple omics layers (e.g., transcriptomics, proteomics).

Conclusions:

• Omics data transformation into a pattern-centric view enhances both predictive tasks and human interpretation.
• This approach holds significant potential for advancing precision medicine applications.
• The method effectively addresses the complexity inherent in large-scale omics datasets.