Detecting Biomarkers Of Alzheimer's Disease Based On Differential Gene Co-expression Network

Alzheimer's disease(AD)is a neurodegenerative disease that causes irreversible damage to parts of the brain and ultimately leads to impaired memory and bodily functions.As the aging rate increases,the number of AD patients will not be underestimated.In addition,the cost of Chinese AD patients is much higher than cancer and cardiovascular disease,and will become an important part of increasing social burden.So far,AD still hasn't clear pathogenic mechanism and effective treatment methods.Most of the researches are related hypotheses,such as amyloid hypothesis and inflammation hypothesis.However,studies have shown that the early diagnosis of Alzheimer's disease has important personal and social benefits.Therefore,in combination with China's national conditions,detecting blood biomarkers that can be used to diagnose AD will have a greater benefit.Although there are already a few biomarkers such as A?-42/A?-40 ratio,concentration of tau protein in CSF and APP gene,there is no clinically validated AD biomarker.In this thesis,differential gene co-expression networks are constructed by different methods based on data characteristics in different brain region data and time series data,and then candidate biomarkers will be detected from the network,after integrating candidate biomarkers and preliminary validating in blood data,the AD biomarkers are finally detected.The main research contents are as follows:(1)For different brain regions data,this study adopts a method of integrating differential gene of six brain regions to construct a differential gene co-expression network,and after obtaining a differential gene co-expression network,different network modules are obtained by clustering the differential co-expression network.The machine learning method screens out the module that contains 44 genes with the best classification ability from the network module as biomarkers of different brain regions.Finally,functional enrichment analysis of the module indicate that module is mainly involved in biological processes such as energy metabolism,transformation and dysfunctional modification.(2)For time series data,this study adopts a method based on gene co-expression network to construct a differential co-expression network,then screens out the differential gene co-expression network.Since the network consists of some modules,the 100 most statistically differential gene pairs in the network are selected.The 189 specific genes are subjected to SVM time series multi-classification evaluation to obtain time series biomarkers.In functional enrichment analysis,biomarkers are mainly involved in biological processes such as programmed cell death,transcription and phosphorylation.(3)Integrating different brain regions and time series biomarkers as candidate biomarkers for the further preliminary validation of blood data.The SVM method is used to evaluate the classification ability of those biomarkers,and the corresponding ROC curve and AUC value are obtained.Among them,the biomarkers of different brain regions correspond to an AUC value of 0.760,and the AUC value corresponding to the integrated biomarkers is 0.920.Considering that the biomarkers composed of 232 genes after integration have better classification ability,they can be used as AD biomarkers detected in this study.