The biological state of the cell is characterized by a complex network of interacting genes, gene products, proteins, microRNAs, as well as other molecules. Microarrays and next generation sequencing technologies have been widely applied to study alterations of these molecules in complex diseases such as cancer. However, most of the standard, widely-used methods for bioinformatics analysis treat these various sources of information independently, looking for overlaps in feature sets rather than directly modeling their interactions.
It has become clear that, to understand what drives (complex) disease, we need to integrate multiple types of ‘omics data in a natural way that allows us to gain insight into the molecular interactions that occur in disease development and progression. Gene regulatory network reconstruction algorithms infer such interactions by drawing on large numbers of measured expression samples to estimate an “aggregate” network model, which represents single estimates for the likelihood of molecular interactions. While informative, aggregate models fail to capture the heterogeneity represented in a disease population. In this presentation, I will introduce a computational framework for single-sample network reconstruction that allows us to “extract” individual patient networks from aggregate networks. I will demonstrate the strengths of this method in multiple big 'omics datasets, and will highlight newly identified gene regulatory interactions that play a role in cancer survival.