Multi-omics protein-coding units as massively parallel Bayesian networks: Empirical validation of causality...

Area of Science:

• Genomics
• Systems Biology
• Computational Biology

Background:

• Understanding gene regulation requires integrating diverse biological data.
• Modeling complex gene interactions is crucial for deciphering cellular functions.

Purpose of the Study:

• To construct fine-grained models of human protein-coding units.
• To investigate conditional independence relationships within these units using multi-omics data.

Main Methods:

• Utilized high-throughput epigenomics, transcriptomics, and proteomics data.
• Developed Bayesian network models representing protein-coding units as directed acyclic graphs (DAGs).
• Analyzed conditional independence relationships implied by DAG structures.

Main Results:

• Successfully modeled over 4000 human genes, integrating transcript isoforms and chromatin accessibility.
• Demonstrated that a significant portion of predicted conditional independencies are supported by empirical data.
• Identified transcript annotation and network structural balance as key factors verifying these independencies.

Conclusions:

• The constructed protein-coding unit models accurately reflect biological conditional independencies.
• Transcript isoform structure and network properties are critical determinants of gene expression relationships.
• This approach provides a robust framework for analyzing gene regulatory networks.