Development of machine learning approaches based on biological networks for prediction and determination of rules governing the emergence of phenotypes of interest

Abstract

The reductionist approach has been effective in explaining the functioning of several biologicalprocesses. However, these processes have been shown to be extremely complex and to have emergent properties that cannot be explained, or even predicted, by such approach. To overcome these limitations, researches have adopted the systems biology approach, a new biology field that investigate how properties emerge from the nonlinear interaction of multiple components of biological processes. These interactions can be represented by a mathematical object called graph or network, where interacting elements are represented by nodes and interactions are represented by edges connecting pairs of nodes. Based on systems biology principles, we propose in this project (i) to construct a Saccharomyces cerevisiae integrated network of gene nteractions simultaneously containing experimentally verified protein physical interactions, metabolic interactions and transcriptional regulation interactions, (ii) to calculate the centrality measures of this network and (iii) to use these centrality measures for predicting and describing, by means of machine learning approaches, the essentiality of each S. cerevisiae's gene and gene interaction in different growth conditions and the associations between gene interactions and phenotypes or biological processes of interest. Moreover, we also propose to develop a machine learning and network-based method to build signaling pathway networks related to biological processes of interest in S. cerevisiae.