Multiomics Integration at Single-Cell Resolution Using Bayesian Networks: A Case Study in Hepatocellular Carcinoma

Area of Science:

• Integrative biology
• Computational biology
• Genomics

Background:

• Multiomics data integration is crucial for understanding complex diseases.
• Single-cell sequencing enables simultaneous multiomics analysis within individual cells.
• Hepatocellular carcinoma (HCC) research benefits from advanced data integration techniques.

Purpose of the Study:

• To perform multiomics data integration at single-cell resolution for HCC.
• To apply Bayesian networks (BNs) for analyzing causal and associational relationships in HCC multiomics data.
• To reveal cellular heterogeneity and mechanistic insights in HCC.

Main Methods:

• Integration of RNA-seq, Reduced Representation Bisulfite Sequencing, and copy number variation data.
• Utilized Bayesian networks (BNs) to model conditional dependencies in single-cell multiomics data.
• Employed hidden Markov models for copy number variation estimation and validated models on an independent dataset.

Main Results:

• Identified best-fitted BN models for 295 genes, elucidating HCC heterogeneity at single-cell omics resolution.
• Provided novel insights into multiomics mechanistic relationships within human lymphocyte antigen class I genes in HCC.
• Demonstrated the utility of BNs for single-cell multiomics integration in a complex disease context.

Conclusions:

• This study pioneers the integration of multiomics data using machine learning (BNs) at single-cell resolution for HCC.
• The findings enhance understanding of HCC's cellular heterogeneity and molecular mechanisms.
• This approach offers a powerful framework for future complex disease research using single-cell multiomics data.