Data Mining And Early-warning For The Sudden Deterioration Of Some Cancers

In this paper,data mining and identifying early-warning signals of prostate cancer and liver cancer are by dynamic network biomarkers based on multi-samples and single-samples,and the results are in agreement with the experiments.Enrichment analysis and survival analysis of the obtained dynamical network biomarkers showed that most of these genes in biomarkers are closely related to some cancers.In addition,the reliability of the dynamic network biomarkers to determine the critical points and early warning signals of the sudden deterioration of complex diseases is verified by using leave-one-out cross validation and bootstrap.The first chapter is the introduction,which mainly introduces several algorithms and some research background in mining critical points and early warning signals for sudden deterioration in complex diseases.The second chapter is preparatory knowledge,mainly introducing the concept of critical point,bifurcation of dynamic system and properties of stochastic disturbance system.The third chapter introduces how to mine the critical point and early warning signal of the complex system by dynamic network biomarkers method based on multi-samples and single-samples.The two methods are applied to two real high-throughput biomolecular time series datas(prostate cancer samples and liver cancer samples),and the critical points and early warning signals of their malignant mutations are excavated,which agree with the experiments.The MATLAB2012 Rb is used to simulate the trend of malignant mutations of complex diseases,the whole process of malignant mutation of prostate cancer samples was simulated by Cytoscape also.The fourth chapter analyzes the obtained dynamic network biomarkers by KEGG enrichment firstly,we found that these dynamic network biomarkers are indeed associated with some cancers.18 abnormally expressed genes are found in each disease sample after survival analysis respectively,according to some concerned reference literatures,we know that the 36 genes with abnormal expression are closely related to some cancers.Finally,the reliability of the dynamic network biomarkers method is verified by leave-one-out cross validation and bootstrap.