Research On Disease-related Knowledge Prediction Method Based On Knowledge Graph

Various omics studies in biomedicine have shown that human diseases and abnormal phenotypes are related to disease biomarkers such as coding genes,non-coding RNA,proteins,metabolites,and these associations are important for understanding the pathophysiological mechanism of diseases from the molecular system level.Both research and disease treatment programs play a very critical role.With the development of science and technology and the continuous update and iteration of sequencing technology,various types of disease-related omics data have appeared in large numbers.For these data,biomedical databases are constructed manually through expert review or automatically constructed through computer technology such as machine learning.However,at present,such databases have problems such as small data volume,single data types,slow update speed,and data redundancy and scattered among various databases.In response to the above situation,this article integrates various disease-related biomedical libraries to construct a disease knowledge graph,and uses the rich information in the knowledge graph to study the prediction methods of disease-related knowledge:First,according to the current characteristics of disease-related biomedical databases and the general norms of medical vertical fields,this paper proposes a top-down method of integrating biomedical databases to construct a disease knowledge map.First construct each ontology class and attribute according to the characteristics of the data source and biomedical specifications,then extract the entities in the biomedical database according to the ontology information,and finally implement the alignment algorithms based on similarity and rules for diseases and related knowledge entities.The integration of the same type of data sources,and the integration of different types of data sources with the help of DO,Me SH and other databases to map diseases.Secondly,this paper uses the rich information in the constructed disease knowledge map to propose a disease-related knowledge prediction method based on graph embedding.This method uses the k-mer frequency encoding of the RNA sequence to extract the characteristic information of nc RNA,and uses the disease semantics and knowledge map The relationship between the disease entity and other omics data extracts disease feature information,and extracts deep feature information through SAE and reduces dimensionality.At the same time,GCN using multi-relational combination realizes simultaneous learning of neighbor nodes and edge type information of heterogeneous graphs of multi-relation diseases.Experiments show that the graph embedding prediction method based on the knowledge graph can achieve better prediction results by introducing RNA sequence,multiple association relationships and relationship type information.Finally,in order to facilitate the researcher’s application of the disease knowledge graph data,this paper builds a search platform,which can search and display the entities and associated relationship data in the knowledge graph.