Algorithms for gene regulatory networks reconstruction

Gene regulatory network (GRN) inference is a central problem in systems biology, that has been attracting a lot of research efforts recently. Perturbation experiments, in which the activity of some genes is suppressed, are one of the main tools to tackle the problem. Modern technologies allow biologists to collect large amounts of data about perturbation experiments on molecular level. This data can be used to infer the structure of the real biological networks, that can not be accessed directly, which is essential for the understanding of the mechanism of genes interactions. Discovering these interactions is not only interesting from the fundamental point of view, but is also applied in the pharmaceutical industry to create new medicines. In this work we analyse the previous methods, highlight their advantages and drawbacks, and propose improvements to enhance the effectiveness of GRN inference. In addition to the improvement of the existing techniques, we present REHub (Reverse Engineering with Hubs) a novel method, distinguishing the influential genes (hubs) and promoting the genes these hubs interact with. Our approach combines statistical and graph algorithmic methods that exploit the scale-free nature of the GRNs, using the clustering techniques. The combination of the state of the art and novel techniques is implemented in the form of an original framework. To evaluate our approach a series of experiments are performed over in silico data. A de facto standard benchmark for the GRN inference algorithms DREAM4 shows that the REHub-based methods outperform the best state of the art methods for the network reconstruction. The achieved DREAM score is the highest reported in the literature.