Desenvolvimento de um pipeline de predição e análise de rede de interação proteina-proteina com aplicação no patossistema soja-ferrugem

Abstract

Molecular biological networks are important for providing interactions between elements of a cell or communication between two or more organisms. Different computational approaches can be used to predict and cluster these networks. In this thesis, network prediction and analysis approaches were established to study the interactions between proteins from two different organisms, using integration of different bioinformatics programs. This established approach was applied to study the interaction of a pathogenic fungus with soybeans. Brazil is the world's largest producer of soybeans ($Glycine$ $max$), this crop is also the main contributor to Brazilian GDP associated with agribusiness, worldwide soybean trade generates around 200 billion dollars per year. Asian soybean rust, caused by the biotrophic fungus Phakopsora Pachyrhizi, is the most important disease in this crop, it is estimated that it has already caused losses of around 150 billion dollars worldwide. Proteins play a key role in the biological processes of organisms, including the interaction interface between pathogens and their hosts, promoting processes such as infection. In this way, understanding the interaction between pathogen proteins and host proteins can allow the development of new methods and approaches such as controlling Asian soybean rust. In this project, a protein-protein interaction network approach was applied specifically to the infection of $Glycine$ $max$ by $Phakopsora$ $pachyrhizi$ to identify probable effector proteins of the fungus. Initially, protein sequences predicted based on the reference genome of G. $max$ and P. $pachyrhizi$ were retrieved from public databases and aligned with protein sequences from public portals such as STRING. In-house scripts were produced to identify potential interaction pairs between proteins, using high-confidence interactions described in several portals, followed by calculation of interaction probability index and integration of data from protein subcellular localization predictors and expression data genes generated by RNASeq methodology during a G. max-P. interaction. pachyrhizi. The network will also be expanded with data from other public protein-protein interaction databases. To date, the interactions found have been modeled in the format of complex networks, which allows the application of techniques for mining and identifying key proteins in the infection process. The objective of this project is also the structural modeling of selected proteins based on the network to study the molecular mechanisms associated with the interaction interface.