Simulation and identification of gene regulatory networks

Gene regulatory networks are a well-established model to represent the functioning, at gene level, of utterly elaborated biological networks. Studying and understanding such models of gene communication might enable researchers to rightly address costly laboratory experiments, e.g. by selecting a small set of genes deemed to be responsible for a particular disease, or by indicating with confidence which molecule is supposed to be susceptible to certain drug treatments. This thesis explores two main aspects regarding gene regulatory networks: (i) the simulation of realistic perturbative and systems genetics experiments in gene networks, and (ii) the inference of gene networks from simulated and real data measurements. In detail, the following themes will be discussed: (i) SysGenSIM, an open source software to produce gene networks with realistic topology and simulate systems genetics or targeted perturbative experiments; (ii) two state of the arts algorithms for the structural identification of gene networks from single-gene knockout measurements; (iii) an approach to reverse-engineering gene networks from heterogeneous compendia; (iv) a methodology to infer gene interactions fromsystems genetics dataset. These works have been positively recognized by the scientific community. In particular, SysGenSIM has been used – in addition to providing valuable test benches for the development of the above inference algorithms – to generate benchmark datasets for international competitions as the DREAM5 Systems Genetics challenge and the StatSeq workshop. The identificationmethodologies earned their worth by accurately reverse-engineering gene networks at established contests, namely the DREAM Network Inference challenges. Results are explained and discussed thoroughly in the thesis.