Causal integration of multi-omics data with prior knowledge to generate mechanistic hypotheses

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Multi-omics datasets offer comprehensive molecular insights but lack systematic methods for hypothesis generation.
• Integrating diverse omics data (phosphoproteomics, transcriptomics, metabolomics) is crucial for understanding complex biological systems.

Purpose of the Study:

• To present COSMOS (Causal Oriented Search of Multi-Omics Space), a novel computational method for integrating multi-omics data.
• To enable the systematic extraction of mechanistic hypotheses from integrated omics datasets.

Main Methods:

• COSMOS integrates phosphoproteomics, transcriptomics, and metabolomics data.
• It leverages prior knowledge of biological networks (signaling, metabolic, gene regulatory).
• Computational methods estimate transcription factor and kinase activities and perform network-level causal reasoning.

Main Results:

• COSMOS successfully captured crosstalk within and between omics layers in clear cell renal cell carcinoma (ccRCC) data.
• The method identified known ccRCC drug targets, demonstrating its biological relevance.
• Mechanistic hypotheses were generated for experimental observations across multi-omics datasets.

Conclusions:

• COSMOS provides a powerful, freely available tool for extracting mechanistic insights from multi-omics studies.
• The method facilitates a deeper understanding of complex diseases like ccRCC by integrating diverse molecular data.
• COSMOS is expected to be broadly useful for researchers analyzing multi-omics data.