Detecting Early Warning Signal Of Complex Diseases Based On Dynamical Network Biomarkers

Complex diseases are resulted from the joint action of many factors,whose genetic model is relatively complex.The sudden deterioration of some complex diseases is called critical state.Before the occurrence of critical state,the illness conditions change slowly.However,the condition deteriorates rapidly in a short time after the state,which is extremely hazardous due to its difficult treatment and shorter survival time.Hence,it is particularly important to diagnose the critical state in time and identify the warning signals before the sudden deterioration of diseases.This thesis mainly focused on the research of identifying the critical state of outbreak influenza A in ten countries.Thereinto,dynamical network biomarkers(DNB)was applied in protein sequences to constructed early warning indexes.The influenza subtypes related to the epidemic were captured through the early warning indexes.Starting from the gene expression data of A/H1N1 and H3N2 diseases,the DNB method was improved to screen the dominant gene network and identify the critical state of different influenza subtypes.The main work of this thesis is as follows:(1)Ten protein sequence data of influenza A reported in several countries were selected to sequentially extract the amino acid that first appeared most frequently in all sequences in year y to form a new sequence.The most frequently occurring amino acids were extracted sequentially from all the sequences in each year to form new sequences.The original sequence in year y+1 was compared with the new sequence in year y to further obtain the 0-1sequence of protein data.The 23 dimensional digital characteristics were extracted to construct the dynamic network of ten protein interactions,from which the DNB were found to build the early warning indicators and quantify the outbreak signal of influenza A,so as to identify the outbreak critical point of the epidemic effectively.In addition,the H1N1 and H3N2 subtypes were analyzed based on the significant signal changes captured from the indexes.(2)Based on the study(1),gene expression data of human bronchial epithelial cells infected with influenza A H1N1 virus were selected.Comparing the normal samples with the infected samples,differential expressed genes(DEGs)were filtrated.To prevent errors caused by different gene expression levels,expression data were normalized by special z-score method.Hierarchical clustering was used to classify differential expressed genes into 40 categories,the optimal of which was determined as DNB combined with three indicators SD,PCC and OPCC.At the level of overall gene network,early warning index EWI were constructed based on indicators CV,DIF and ODIF to identify the critical state of H1N1 disease.This plays an important role in preventing the deterioration of influenza A H1N1 disease,and whose dominant gene network provides reference data for studying influenza vaccine.(3)Based on the study(1),gene expression data of peripheral blood of healthy volunteers after inoculation with influenza A H3N2 virus were selected.Different from the data structure in study(2),significance analysis was used by comparing the infected samples at different time points after virus with the gene expression data before inoculation,which obtaineddifferential expressed genes at each time points.According to the Silhouette Coefficient,the optimal classification of each time point which was determined and was analyzed for further K-means clustering.The dominant network of each time point was determined by indexes CV ? PCC and OPCC.The union of all the dominant networks was selected as DNB to construct early warning index EWI to identify the critical state of H3N2 subtype infection.And the gene composition and function in the dominant network were analyzed deeply.