PhD thesis defense: Pau Bellot

Pau Bellot, defends his PhD thesis entitled Study of Gene Regulatory Networks Inference Methods from Gene Expression Data Monday May 8th, 10h30, Aula Telensenyament, B3 building, first floor

Dissertation summary:

A cell is a the basic structural and functional unit of every living thing, it is protein-based an that regulates itself. The cell eats to stay alive, it grows and develops; reacting to the environment, while subjected to evolution. It also makes copies of itself. These processes are governed by chain of chemical reactions, creating a complex system. The scientific community has proposed to model the whole process with Gene Regulatory Networks (GRN). The understanding of these networks allows gaining a systems-level acknowledgment of biological organisms and also to genetically related diseases.

This thesis focused on network inference from gene expression data, will contribute to this field of knowledge by studying different techniques that allows a better reconstruction of GRN. Gene expression datasets, are characterised by having thousands of noisy variables measured only with tens of samples. Moreover, these variables presents non-linear dependencies between them. Therefore, recovering a model that is capable of capturing the relationships contained in this data, constitutes a major challenge.

The main contribution of this thesis is a set of fair and sound studies of different GRN inference methods and post-processing algorithms. First, we present a novel approach for inferring gene networks and we compare it with other methods. It is inspired by the concept of “variable importance” in feature selection. However, many algorithms can be proposed to infer GRNs, so there is a need to assess the quality of these algorithms. Secondly, and motivated by the fact that the previous comparison was not informative enough, we introduce a new framework for in silico performance assessment of GRN inference methods. 

This work has led to an open source R/Bioconductor package called NetBenchmark. Finally, and thanks to this tool we have corroborated that inferring gene regulatory networks from expression data is a tough problem. The different algorithms have some particular biases and strengths, and none of them is the best across all types of data and datasets. Therefore, we present a framework for evaluating and standardising network consensus methods to aggregate various network inferences