Detecting The Early Warning Signals Of Disease Deterioration Based On Sample-specific Temporal Differential Networks

The development of diseases is generally divided into normal state,pre-disease state and disease state.Pre-disease state is a critical stage of the disease state,and patients in this state may return to the normal state as long as they receive reasonable and effective treatment.Therefore,it is of great significance for patients to detect the pre-disease state.In this paper,an algorithm is developed,and a composite variable to detect the critical point of the system is proposed based on the sample-specific temporal differential networks which are established by individual single sample,and the composite variable can be used to detect the early warning signals of disease deterioration effectively and identify the pre-disease state.The validity of this method is verified by a numerical simulation experiment,prostate cancer data(GSE 5345)and breast cancer data(GSE13009).By analyzing the obtained dynamic differential network biomarkers with KEGG enrichment and survival analysis,it is found that most of the genes get involved in the development of cancer.The main work and results of this paper are as follows:Firstly,by establishing a complex system which is made up of 8 gene nodes with differential equations,we not only detect the critical point of this system effectively,but also identify 4 dynamic network markers of this complex system,and the feasibility of the critical point detection method for complex systems based on SSDNs is verified.Secondly,the gene expression data of GSE 5345 and GSE 13009 is downloaded from GEO database,processed and screened.On the basis of gene expression data,the molecular interaction network is established by pearson correlation coefficient between genes,and then,the observed data of genes at each time point is transformed into the samplespecific temporal differential networks.Finally,by using the MATLAB R2016 a software,we obtain that the critical mutation time points of prostate cancer and breast cancer are respectively 24 h and 1.5h,and we also identify 202 dynamic differential network biomarkers of prostate cancer,120 and172 dynamic differential network biomarkers of breast cancer stimulated by EGF and HRG respectively,then,24 common biomarkers among them of breast cancer which were stimulated by these two ligands are selected.In this paper,we have successfully detected the critical points of the deterioration of prostate cancer and breast cancer by using the characteristics of biological networks.And most of the genes in the dynamic differential network biomarkers can be used as target genes for cancer treatment.