Plawss: power law semantic similarity metodologia data-driven baseada em lei de potência para o cálculo de similaridade semântica GO

• Ano de defesa: 2020
• Autor(a) principal: Ito, Eric Augusto
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Cornelio Procopio; Brasil; Programa de Pós-Graduação em Bioinformática; UTFPR
• Palavras-chave em Português: Computação semântica; Bioinformática; Ontologia; Semantic computing; Bioinformatics; Ontology; CNPQ::CIENCIAS EXATAS E DA TERRA; Engenharia/Tecnologia/Gestão
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/30171

Genes are widely studied by the scientific community due to their importance in various researches, many of them related to health. Because of this, many methods were developed to calculate the SS between genes. Semantic Similarity has been used in several researches such as network inference and validation, protein folding, among others. Initially proposed by Wang el al. (WANG et al., 2007) and which was added to the GOGO tool (ZHAO; WANG, 2018), the methodology presented by Wang and GOGO is not limited to using IC to calculate the semantic similarity. Wang proposed a hybrid method that calculates similarity from the topology of the directed acyclic graph GO. GOGO in turn proposed to use the number of child terms as a substitute for IC, since GOGO noticed the inverse correlation between IC and the number of child terms, so even without using the IC, GOGO manages to take advantage of methods based on IC along with Wang’s hybrid method. However, GOGO proposes a method that depends on variables that do not fit the data of ontologies, on the other hand Wang is limited to weighing the terms GO only depending on the type of connection between the terms GO. This work presents a new method named PLAWSS for calculating semantic similarity in genes using a hybrid model to calculate semantic similarity using Genetic Ontology, which is data-driven adapting to ontology data using power law to weigh each GO term, and that in addition to the type of connection, this work also takes into account the number of children of the ancestor to identify the specificity of the GO term. Six data sets composed of metabolic pathways were clustered using the semantic similarities calculated between each pair of genes, the clusters formed from the molecular functions and biological processes showed the best results, being 83.33% and 66.67% respectively, corroborating for the testing of the proposed method.