Interpretation of large scale biological data facilitated by curated causal biological network models

We have previously shown that a semi-automated knowledge extraction workflow, featuring a text mining pipeline as well as a curation interface, provides an efficient workflow for knowledge extraction and the building of causal biological network models. The network models describe biological processes from upstream events to downstream measureable and quantifiable nodes, i.e., the expression of genes (Big Data) measured with microarray experiments. We now demonstrate the relevance of one of these curated network models for the interpretation of complex molecular data. Our example includes an atherosclerotic plaque destabilization network model that was created and curated with the knowledge extraction workflow as well as multiple transcriptomics datasets in a cardio vascular disease context of aortic tissue. Once the network is overlaid with experimental data, reverse causal reasoning and network perturbation amplitude algorithms allow the quantification of the impact an exposure/treatment/disease has on the experimental system. We also compare and discuss the results against a more conventional approach for transcriptomics data interpretation based on gene sets enrichment analysis. Our results show that by using curated causal biological network models with large datasets and perturbation quantification algorithms, the interpretation becomes more efficient, objective, and interpretable, as biological processes are described more holistically and are based on causal relationships and relevant context information. In summary, the semi-automated knowledge extraction workflow facilitates the construction of causal biological network models describing disease specific and complementary biological processes that can be fully contextualized and used for the interpretation of large scale biological data. In the context of Big Data, this toolset allows the organization of data in a more meaningful and accessible way as well as the simplification to incorporate new data.