Research On Disease Classification Method Based On Multi-view Clustering

The International Classification of Diseases(ICD)is an internationally used disease classification standard and has been widely used in clinical practice.However,the characteristics of the international disease classification system depend on clinical characteristics and lag behind the current research on the molecular biology of disease.Therefore,it is necessary to combine continuously updated biomedical data to classify diseases to meet the needs of precision medicine.Similarly,for a specific disease,unsupervised clustering methods have been widely used for unbiased biomedical discovery of disease subtypes.But with the deepening of disease biomolecular research,disease molecular expression data continues to increase.Clustering research only through the data of a single view,often because of different data,the disease classification conclusions are different.Therefore,it is necessary to fuse multi-source data to obtain more reasonable disease typing results.Therefore,based on the multi-view clustering algorithm,this paper studies the classification system of the entire disease and the classification of cancer.The main research work of this paper is as follows:(1)This article collects and integrates four kinds of disease relationships,namely,disease international disease classification code,disease gene relationship,disease symptom relationship and disease protein interaction.Through the four sets of data,the similarity relationship between the four types of disease pairs was constructed,and the international disease sub-system was evaluated based on the disease similarity relationship.It verifies the rationality of the classification of the international disease classification system,and also proves the high molecular diversity in the chapter of the international disease classification system.In view of the lack of specific understanding of disease molecules in the international disease classification system,two multi-view disease classification models are constructed in this paper,namely a clustering model based on deep multi-view network fusion and a disease classification model based on multi-view conceptual decomposition algorithm.The two models take the similarity relationship of the four types of diseases obtained above as four data views,and conduct multi-view disease clustering research to obtain a higher-quality disease classification.(2)Aiming at the existing classification studies,the single source data is mostly used to obtain the results of different disease classification results.In this paper,we obtained five kinds of cancer data sets,including expression data and survival data of five kinds of cancer.Each type of cancer expression data includes three types of expression data: gene expression data,DNA methylation data,and mi RNA expression data.The deep multi-view network fusion clustering model is used for disease classification research,and a multi-view clustering algorithm based on graph fusion is also applied to disease classification research.The two methods use three expression data as three data views to conduct multi-view disease typing experiments.The experimental results were verified by using survival data.The survival results were used to verify the experimental results.The results show that the two models are significantly better than the single data type analysis in identifying cancer subtypes,and the classification results obtained have certain clinical significance.