Research On Liver Cancer Subtype Classification Based On Omics Data

Liver cancer is a serious threat to the safety of human life,and it is urgent to find a way to treat liver cancer.In the field of liver cancer research,subtype classification has become one of the research hotspots.Subtype classification of liver cancer can help patients to carry out precise treatment,thus improving the prognosis of patients,and also help to understand the pathogenesis of liver cancer on the molecular level.With the development of the times,the emergence of high-throughput sequencing technology makes it easier to obtain omics data of liver cancer,and at the same time has brought challenges and opportunities to the research and treatment of liver cancer.Small sample size,high dimension and high noise are the characteristics of liver cancer multi-omics data,and there is complementary information among different types of omics data.It is of great significance for the research of liver cancer how to mine the information related to subtype classification based on liver cancer omics data.The main research content of this thesis is to construct subtype classification method of liver cancer based on omics data.The main work is summarized as follows:1.An integrated method for liver cancer subtype classification and survival prediction based on multi-omics data was proposed.The method used autoencoder to reduce the dimensions of each type of omics data and extract the features,at the same time,univariate Cox regression model was used to select survival related features.Next,based on the features after dimensionality reduction,the univariate Cox regression model was used again to select survival related features,then,the twice selected features were stacked,so it can form a characteristic matrix,three types of omics data can form the corresponding three characteristic matrices.The data integration method based on radial basis function kernel similarity network fusion is used to fuse the three characteristic matrices into a similarity matrix.Finally,spectral clustering method were used to identify the subtypes of liver cancer based on the similarity matrix.The method also trained multiple data types machine learning classification models to predict subtypes.The results showed that the method obtained clinically significant liver cancer subtypes and prognosis analysis results.2.A method for liver cancer subtype classification and gene expression profiles analysis based on omics data was proposed.The method used the PyClone method to quantify tumor heterogeneity of liver cancer based on omics data,subtypes of liver cancer were classified by linear segmental regression method,then,based on gene expression profiles,the method analyzed the prognosis and survival of patients with these two liver cancer subtypes.The results show that the method for subtype classification of liver cancer has important clinical significance for survival and prognosis of patients.