Using Bayesian networks to identify control topography between cancer processes and immune responses via metagene constructs

Cancer arises from a deregulation of both intracellular and intercellular control systems. Understanding the architecture of these control systems and how they are changed in diseases could present opportunities for therapeutic targets to restore normal control. However, since intercellular control structures only appear in intact systems, it is difficult to identify how these control structures become altered using in vitro models and it can be difficult to determine if an in vivo model system appropriately replicates what occurs in human disease. In order to overcome this, we use the diversity in normal and malignant human tissue samples from the Cancer Genome Atlas database of human breast cancer to identify intercellular control topology in vivo. To improve the underlying biological signals from the noisy gene expression data, we constructed Bayesian networks using metagene constructs, which represented groups of genes that are concomitantly reported with different immune and cancer states. From these directional, acyclic graphs, we found opposing relationships between cell proliferation and epithelial-to-mesenchymal transformation (EMT) with regards to macrophage polarization. Furthermore, we also found that it was possible to identify the relationship between EMT and macrophage polarization with fewer datasets when the Bayesian network was generated from malignant samples alone, while it was possible to identify the relationship between proliferation and macrophage polarization with fewer samples when the samples were taken from a combination of the normal and malignant samples. When the same technique was applied to other cancers, we found a common result that proliferation was associated with a type 1 cell-mediated anti-tumor immunity and EMT was associated with a pro-tumor anti-inflammatory response. All together, these networks give us an understanding of what relationships are occurring in human cancer progression, and this knowledge can be used to help identify model system that more closely mimic human disease progression