Data-driven learning how oncogenic gene expression locally alters heterocellular networks

Area of Science:

• Computational Biology and Bioinformatics
• Cancer Research and Immunology
• Systems Biology

Background:

• Mechanistic modeling and simulation are crucial for drug development, aiding in understanding oncogenesis, plasticity, and immunity.
• Current hand-curated cell-level models can introduce bias in data interpretation and drug target prioritization.
• Developing unbiased, data-driven approaches for mechanistic cell-level modeling is essential.

Purpose of the Study:

• To develop a novel computational framework for generating causal cell-level models from bulk transcriptomic data.
• To investigate the impact of increased Cell Communication Network factor 4 (CCN4) on the tumor microenvironment.
• To validate computational predictions using experimental models.

Main Methods:

• Integration of digital cytometry with Bayesian network inference to build causal models.
• Analysis of bulk transcriptomic datasets to link oncogenic gene expression with stromal and immune cell alterations.
• Utilizing patient data from breast cancer and melanoma, and testing predictions in immunocompetent mouse models.

Main Results:

• Successfully generated causal cell-level models predicting tumor microenvironment alterations.
• Identified specific impacts of increased CCN4 on stromal and immune cell subsets.
• Experimental validation in mouse models confirmed the computational predictions.

Conclusions:

• The combined digital cytometry and Bayesian network approach offers an unbiased method for mechanistic cell-level modeling.
• CCN4 plays a significant role in modulating the tumor microenvironment.
• This integrated computational and experimental strategy advances drug target discovery and understanding of cancer biology.