Inferring cancer common and specific gene networks via multi-layer joint graphical model

Area of Science:

• Computational biology
• Genomics
• Cancer research

Background:

• Cancer arises from genetic variants, necessitating gene network reconstruction for understanding carcinogenesis.
• High-throughput sequencing enables computational gene network inference, but challenges remain in comparing networks across cancer types.

Purpose of the Study:

• To develop a computational model for jointly estimating gene networks across multiple cancer types.
• To identify shared and unique regulatory mechanisms between different cancer types for precision medicine.

Main Methods:

• Proposed a sparse regularized multi-layer decomposition graphical model.
• The model jointly estimates gene networks from various gene expression data types.
• Decomposes cancer-specific networks into globally shared, partially shared, and cancer-type-unique components.

Main Results:

• Demonstrated model effectiveness on synthetic data for joint gene network estimation.
• Applied the model to real datasets for inferring gene networks in cancer subtypes and cell lines.
• Identified key genes associated with common and specific regulatory mechanisms across cancer types.

Conclusions:

• The developed model effectively explores similarities and differences between cancer gene networks.
• Identified shared and unique network components offer insights into common and specific carcinogenesis mechanisms.
• This approach supports precision cancer diagnosis and treatment by revealing cross-cancer regulatory insights.