Identifying key multifunctional components shared by critical cancer and normal liver pathways via SparseGMM

Area of Science:

• Genomics
• Systems Biology
• Biostatistics

Background:

• Developing statistical models for complex diseases with genetic underpinnings using multimodal data is challenging.
• Understanding gene regulatory networks is crucial for deciphering disease mechanisms.

Purpose of the Study:

• To present SparseGMM, a novel statistical approach for gene regulatory network discovery.
• To apply SparseGMM to multiomic data for identifying gene regulatory relationships in liver cancer.

Main Methods:

• SparseGMM utilizes latent variable modeling with sparsity constraints to learn Gaussian mixtures from multiomic data.
• A Bayesian framework combined with coexpression patterns quantifies regulator confidence and target gene assignment uncertainty via gene entropy.
• The method was applied to liver cancer and normal liver tissue data, with validation on an independent single-cell RNA sequencing (scRNA-seq) dataset.

Main Results:

• SparseGMM identified PROCR, PDCD1LG2, and HNF4A as key regulators involved in angiogenesis, immune response, and blood coagulation in liver cancer.
• A significant increase in gene entropy was observed in cancer tissues compared to normal liver tissues.
• High-entropy genes included critical multifunctional components within pathways like p53 and estrogen signaling.

Conclusions:

• SparseGMM provides a robust framework for gene regulatory network discovery from multiomic data.
• The findings highlight specific gene regulators and pathways implicated in liver cancer pathogenesis.
• Increased gene entropy in cancer suggests altered regulatory dynamics and identifies potential therapeutic targets.