Design of new algorithms for gene network reconstruction applied to in silico of biomedical data

Resumen

The root causes of disease are still poorly understood. The success of current therapies is limited because persistent diseases are frequently treated based on their symptoms rather than the underlying cause of the disease. Therefore, biomedical research is experiencing a technology-driven shift to data-driven holistic approaches to better characterize the molecular mechanisms causing disease. Using omics data as an input, emerging disciplines like network biology attempt to model the relationships between biomolecules. To this effect, gene co- expression networks arise as a promising tool for deciphering the relationships between genes in large transcriptomic datasets. However, because of their low specificity and high false positive rate, they demonstrate a limited capacity to retrieve the disrupted mechanisms that lead to disease onset, progression, and maintenance. Within the context of statistical modeling, we dove deeper into the reconstruction of gene co-expression networks with the specific goal of discovering disease-specific features directly from expression data. Using ensemble techniques, which combine the results of various metrics, we were able to more precisely capture biologically significant relationships between genes. We were able to find de novo potential disease-specific features with the help of prior biological knowledge and the development of new network inference techniques. Through our different approaches, we analyzed large gene sets across multiple samples and used gene expression as a surrogate marker for the inherent biological processes, reconstructing robust gene co-expression networks that are simple to explore. By mining disease-specific gene co-expression networks we come up with a useful framework for identifying new omics-phenotype associations from conditional expression datasets.In this sense, understanding diseases from the perspective of biological network perturbations will improve personalized medicine, impacting rational biomarker discovery, patient stratification and drug design, and ultimately leading to more targeted therapies.