Reverse engineering of natural systems by graph theory

With the advent of high-throughput technology, Biological research widened its horizons in terms of biomarkers and mechanisms of action of several diseases and phenotypes. On the other hand, complex diseases, like diabetes, several neurodegenerative pathologies and cancer, are still orphan of a cause and then of a cure. One of the possible reasons is that these are not strictly monogenic diseases since they result from a global interplay between molecular players and master regulators. In this context, where “the whole is something over and above its parts and not just the sum of them all” (Aristotle 384-322 B.C.), is clear that the Cartesian reductionism cannot completely help understand how a disease arises and develops. Systems Biology comes on the stage here and puts emphasis on whole behavior as being basically indivisible. It sustains the Smuts’s holistic theory according to which whole systems such as cells, tissues, organisms, and populations were proposed to have unique emergent properties and that it was impossible to reassemble the behavior of the whole from the properties of the individual components. Hence, new technologies were necessary to define and understand the behavior of systems. New mathematical models and computational approaches emerged in the past decades. Thereby taking inspiration from the theory of graphs. Aspects of nature that could be explained by the interaction of individual agents were modeled as networks and their properties studied topologically. Speculations on the global structure of biological systems were based on two important assertions: systems have a hierarchical structure, and the structure is held together by numerous linkages to construct very complex networks. In this work, we retrace this path by first reconstructing and studying a complex molecular system made by gene and microRNA expression profiles in patients affected by colorectal cancer. We show how the study of topological properties of the system helped identifying a tiny subnetwork of master-regulator and effectors that, individually, were associated to poor survival rates when extremely expressed. Group-effects were not captured, until the development of Pyntacle, a cross-platform and open source Python suite of high-performance computing algorithms for the discovery of key-players in networks. Pyntacle is introduced and presented in this work and then used proficiently in two other case studies. The former regards ecological food webs and reports on the assessment of their nestedness property, which is an indicator of their global robustness and redundancy. The latter is an exploration of the relationships between sex and ageing process in Drosophila melanogaster, which developed into two computational steps: definition of co-expressing modules of genes and identification of sex independent key-players molecules in male and female flies.