Identifying Critical States Of Complex Diseases Based On Single-sample Dynamic Network Biomarkers

Complex diseases have complex genetic patterns and occur under the combined action of many factors,which seriously endanger human health.The development process of complex diseases is usually irreversible.Therefore,for complex diseases such as cancer,early diagnosis and treatment are essential.In recent years,studies have shown that dynamic network biomarkers(DNB)can be used as early warning signals for a variety of complex diseases,and can effectively identify critical states in the development of diseases.However,most DNB-related algorithms are based on patient tissue data and have high computational complexity,so they are usually not implemented in clinical diagnosis.Based on the DNB theory and the construction of individual specific networks,this paper proposes two complex disease critical state detection methods based on three-dimensional landscape dynamic markers and based on simulated annealing algorithms,which can perform personalized diagnosis on a single sample,greatly improving clinical application value.The quantitative analysis of the two time series datasets of HRG-induced MCF-7 breast cancer cell line differentiation process and HCV-induced dysplasia and hepatocellular carcinoma(HCC)verifies the effectiveness of this method and successfully identifies the critical states in the breast cancer cell line differentiation process and HCC disease process.The results of KEGG pathway analysis and a large amount of literature confirm that the DNB molecules identified by this method are indeed related to the development of the disease.The main work of this thesis is as follows:(1)Constructing an individual specific network based on the characteristics of single sample data.It is difficult to construct a molecular network through a single column of biological data,so this paper introduces an individual specific network to solve this problem.First,a sufficient number of control samples is required to construct a reference network,and then a diseased sample is added to the reference sample to construct a new network.The difference between the new network and the reference network can be regarded as the disturbance network between the diseased sample and the normal sample,that is,the individual specific network.In clinical practice,only a certain number of reference samples are required to construct an individual specific network for any single diseased sample data.(2)Based on the study(1),a method to identify the critical state of complex diseases based on three-dimensional landscape dynamic network biomarkers is proposed.This method is a model-free calculation method based on DNB theory and the construction of individual specific networks.According to the three statistical conditions of the DNB molecule in the critical state of the system,the corresponding local score can be calculated for each node(molecule)in the individual specific network.According to a given threshold,the DNB module in each sample can be automatically determined and the sample’s global score can be given.The state of the sample with the highest global score can be confirmed as the critical state of the disease.This method can score a single sample and can effectively detect the state of the disease,but because the partial score needs to traverse all nodes in the network,the complexity is relatively high.(3)Based on the study(1),a method for identifying critical states of complex diseases based on simulated annealing is proposed.This method is based on DNB theory and individual specific networks combined with intelligent algorithms to solve DNB molecules.On the background of constructing an individual specific network,an appropriate objective function is established according to the three conditions of the DNB molecule in the critical state,then the problem of solving the DNB molecule is transformed into a single objective optimization problem.The application of the simulated annealing algorithm can realize the rapid optimization of this problem.After determining the molecules in the DNB group,the score of a single sample can be evaluated.This method is more excellent in terms of DNB size,algorithm running time and biological relevance.Biological relevance refers to the Pearson correlation coefficient between the expression level of DNB determined in a critical state and the phenotype of interest.In addition,this method can not only solve the problem of a single sample in clinical diagnosis,but also describe the statistical characteristics of each disease state.(4)Using(2)and(3)two calculation methods to quantitatively analyze the HRG-induced MCF-7 breast cancer cell line differentiation process and HCV-induced dysplasia and hepatocellular carcinoma standardized gene expression datasets,respectively determine the critical states in the breast cancer cell line differentiation process and HCC disease process.The reliability of the two calculation methods is verified by KEGG path analysis and literature search.The construction of the individual specific network greatly enhances the clinical application value of DNB theory,and is of great significance to the early diagnosis of complex diseases in the biomedical field.