| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 999.12 KB | Adobe PDF |
Advisor(s)
Abstract(s)
A Medicina Personalizada é um modelo de prática médica que utiliza o
perfil fenotípico e genotípico do indivíduo para melhorar a precisão do
diagnóstico, a eficácia terapêutica ou a prevenção de doenças. Neste
sentido, a enorme quantidade de dados gerados ao longo dos últimos
anos na área biomédica tem contribuído para uma melhor compreensão
dos determinantes genéticos de várias patologias e, consequentemente,
para a implementação de práticas de Medicina Personalizada em
várias áreas, por exemplo na Oncologia e no âmbito das doenças raras.
No entanto, ainda subsistem desafios significativos, nomeadamente no
que diz respeito à integração de dados biomédicos oriundos de fontes
heterogéneas e na obtenção de informação clinicamente relevante.
Este trabalho descreve uma abordagem que usa métodos de aprendizagem
automática aplicados a um Grafo de Conhecimento (GC) biomédico
como um meio para integrar informação armazenada em bases de dados
diversas. Este GC contém relações entre genes, doenças e outras entidades
biológicas, extraídas de três bases de dados: Ensembl, DisGeNET
e Gene Ontology. Neste trabalho exploramos o potencial dos métodos
de aprendizagem automática em grafos para produzir informação clinicamente
relevante e descrevemos a aplicação desta metodologia à previsão
de associações gene-doença. Mostramos ainda que as principais
associações gene-doença previstas por esta abordagem podem ser confirmadas
em bases de dados externas ou já foram previamente identificadas
na literatura.
Personalized Medicine is a model of medical practice that uses an individual's phenotypic and genotypic profile to improve diagnostic accuracy, therapeutic efficacy or disease prevention. The huge amount of data generated over the last few years in the biomedical area has contributed to a better understanding of the genetic determinants of various pathologies and, consequently, to the implementation of Personalized Medicine practices in various areas, for example in Oncology and in the field of rare diseases. However, significant challenges remain, namely with regard to the integration of biomedical data from heterogeneous sources to obtain clinically relevant information. This work describes an approach that uses machine learning methods applied to a biomedical Knowledge Graph (KG) as a means to integrate information stored in different databases. To build the KG, three databases were used: Ensembl, DisGeNET and Gene Ontology. This KG contains relationships between genes, diseases, and other biological entities. In this work, we explore the potential of automatic graph learning methods to produce clinically relevant information and describe the application of this methodology to the prediction of gene-disease associations. We show that the main gene-disease associations predicted by this approach can be confirmed in external databases or have been previously identified in the literature.
Personalized Medicine is a model of medical practice that uses an individual's phenotypic and genotypic profile to improve diagnostic accuracy, therapeutic efficacy or disease prevention. The huge amount of data generated over the last few years in the biomedical area has contributed to a better understanding of the genetic determinants of various pathologies and, consequently, to the implementation of Personalized Medicine practices in various areas, for example in Oncology and in the field of rare diseases. However, significant challenges remain, namely with regard to the integration of biomedical data from heterogeneous sources to obtain clinically relevant information. This work describes an approach that uses machine learning methods applied to a biomedical Knowledge Graph (KG) as a means to integrate information stored in different databases. To build the KG, three databases were used: Ensembl, DisGeNET and Gene Ontology. This KG contains relationships between genes, diseases, and other biological entities. In this work, we explore the potential of automatic graph learning methods to produce clinically relevant information and describe the application of this methodology to the prediction of gene-disease associations. We show that the main gene-disease associations predicted by this approach can be confirmed in external databases or have been previously identified in the literature.
Description
Keywords
Medicina Personalizada Perfil Fenotípico e Genotípico Diagnóstico Genético Aprendizagem Automática Grafos de Conhecimento
Pedagogical Context
Citation
Boletim Epidemiológico Observações. 2021 mai-ago;10(30):57-61
Publisher
Instituto Nacional de Saúde Doutor Ricardo Jorge, IP
