Identification And Validation Of Hub Protein For Potential Biomarkers In Cerebrospinal Fluid With Alzheimer’s Disease By Bioinformatics And Experiments

Alzheimer’s disease(AD)is a heterogeneous neurodegenerative disease with complex pathologies,including the deposition of amyloid Aβ,the tangles of nerve fibers caused by Tau phosphorylation,inflammation and neuro loss.At present,the study of its pathology and clinical biomarkers is still limited,among which,cerebrospinal fluid(CSF)plays an important role in the study of biomarkers,and the study of cerebrospinal fluid biomarkers can reflect the pathology of AD.With the development of proteomics technology,the increasing amount of data indicates that more effective biomarkers can be screened out.Compared with traditional functional screening of AD biomarkers,large-scale data-driven identification of potential biomarkers for accurate diagnosis of AD has a wide range of applications.Moreover,the machine learning feature screening methods can effectively screen out the features with the best value for the classification task from the large-scale data.Therefore,we raise the corresponding scientific question:can we use some effective feature screening methods to identify specific potential biomarkers for AD in the cerebrospinal fluid proteome data for future basic research and clinical diagnosis?To answer this scientific question,we downloaded proteomic data of CSF from 4 cohorts of patients with cognitive unimpaired and AD.Through the combined differential expression analysis of several cohorts,we screened out 29 significantly different proteins shared by cohorts.Then,through the online enrichment analysis website David,functional annotation of the screened proteins was conducted,we found that these proteins were mainly enriched in a series of metabolic and glycolysis related pathways.In addition,six key proteins(YWHAZ,SMOC1,ALDOA,PKM,CHI3L1,SPP1)which played key roles in the task of classifying AD group and normal group were selected by feature importance screening from the differential proteins in each cohort using LASSO regression and random forest methods.These proteins were significantly upregulated in the cerebrospinal fluid of AD patients compared with cognitively unimpaired control patients.In addition,compared with other neurodegenerative diseases,SMOC1,ALDOA,and PKM in CSF of AD patients were significantly upregulated.Then,two groups of Aβ42 positive(A+)and Aβ42 negative(A-)and two groups of p-Tau positive(T+)and p-Tau negative(T-)were generated by Gaussian mixture model from Aβ42 and p-tau data in cohort.Correlation analysis and difference analysis of these 6 proteins in these groups showed that these proteins were highly correlated with Aβ and tau pathology In AD.Then,we constructed protein prediction model and generated ROC curves to capture the predictive power of these proteins for AD.Finally,we used a clinical cohort to verify the expression levels of proteins with strong diagnostic ability in cerebrospinal fluid,blood and brain tissues of patients with Alzheimer’s disease and other patients.In conclusion,these key proteins screened out from our research may become targets for basic AD research and potential biomarkers for clinical research in the future.