Research On The Heterogeneity Of Mild Cognitive Impairment And Pathogenesis Of Alzheimer’s Disease

Alzheimer’s disease(AD)is a neurodegenerative disease.The mainly clinical characteristics of AD is memory loss and cognitive impairment.As of the end of 2020,there are more than 30 million AD patients in the world.In the past century,clinical and basic studies on AD have been intensified from the early clinical observations to the combination of neuroimaging,genomics,molecular biology,and bioinformatics,and these studies greatly deepened the understanding about pathogenesis,clinical manifestations and pathological characteristics of AD.However,many problems remain ambiguous.Due to the lack of effective therapeutic drugs,AD cannot be cured,and its progression is difficult to reverse.Moreover,the onset of AD is insidious.Hence,the patients usually show obvious clinical symptoms when they are diagnosed.Currently,there are several methods to aid AD diagnosis,such as EEG,cerebrospinal fluid biomarkers,magnetic resonance imaging(MRI),positron emission tomography(PET)and computer tomography(CT).However,these methods have difficulties on achieving early clinical diagnosis and prediction.Mild cognitive impairment(MCI)is a neurocognitive disorder which involves cognitive impairments beyond those expected based on an individual’s age,but which are not significant enough to interfere with instrumental activities of daily living.Symptoms of AD typically begin with MCI.MCI can be regarded as the early stage of AD,and it may become a breakthrough to delay the development of AD.In addition,potential biological changes may be the underlying cause that affects clinical symptoms.With the high-performance computing tools,prior studies have found several risk genetic variations related to AD.However,we still know little about the gene expression changing and its underlying regulatory mechanisms in AD.Highthroughput sequencing technique provides new opportunities to further explore the molecular mechanisms of AD.If it is possible to define the different phenotypes and predict the progression in the MCI stage,it may provide information for clinical diagnosis and monitor the patient’s condition.In addition,the study of the molecular mechanisms in AD gene expression may provide a new perspective for delaying or treating AD.Based on this,we analyzed the data of MCI patients,AD patients and age-and gendermatched healthy controls from the ADNI and SYNAPSE databases.Using brain imaging and high-throughput sequencing,combined with artificial intelligence algorithm,we explored the clinical prediction,heterogeneity,progression and molecular mechanism of MCI and AD.The main results are as follows:1.Based on the MRI image features of the cerebral cortex,the heterogeneity of MCI was analyzed.We identified four imaging subtypes in MCI.We used 12 clinical neuropsychological scales to evaluate the subtypes.The results showed that the four MCI subtypes have different severity and conversion rates.Moreover,through genome-wide association analysis and enrichment analysis,differences in molecular pathways related to the four subtypes were found.These results indicated that distinct genetic dysfunctions may be related to different subtypes with different imaging features.The study proved that neuroimaging can not only be used as an aid tool for clinical diagnosis,but also help to reveal the heterogeneity in the disease.2.We contrasted a multi-factor prediction model that can personally predict the conversion of MCI to AD by using image features of cerebral cortex,clinical neuropsychological scale,and cerebrospinal fluid amyloid measurement.The C-index of the prediction model is 0.978 in the training cohort and 0.956 in the validation cohort,which showed a good predictive ability.This study proposed a visual tool to personalize predict the conversion probability of MCI to AD.Furthermore,we studied the relationship between clinical characteristics,imaging features,pathological features and gene expression.3.This study conducted a preliminary exploration to use the pseudotime to simulate the molecular progression of AD.The results showed that the structure learning based on the reverse graph embedding can distribute the AD samples on the pseudotime.When evaluating the model using pathological and clinical scores,we found that the pseudotime value increased with the worse cognitive and memory status of the patient.In addition,this study used a pseudotime-based gene regulatory network to explore the regulatory relationship between genes in the AD progression.The result showed that the calsyntenins3 exhibits regulatory relationships with several genes in the progression of the AD.Furthermore,deconvolution algorithm was used to infer the relative proportions of brain cell types in the progression of AD.It was found that in the early stage of the AD,the proportions of each cell type did not differ much,but as the AD progressed,neurons and oligodendrocyte progenitor cells apoptotic rapidly.The relative proportion of brain cell types changed in the late stages of AD.4.We explored the alternative polyadenylation(APA)and its biological influence in AD by using single-cell sequencing data.The results found that APA events showed differences between the normal aging group and the AD group;there were different APA events in different cell types,which might affect the binding sites of microRNA and regulatory sequence motifs.In addition,our study found that some genes with APA events were differentially expressed between the normal aging group and the AD group.Part of these genes were also correlated with pathological and clinical measures.In summary,this paper explored the heterogeneity,clinical prediction,disease progression and underlying molecular mechanisms of MCI and AD by using medical imaging data,transcriptome sequencing data and clinical information.Combined with radiomics,cluster analysis,structural learning,and various bioinformatics tools,we hope this paper can contribute to the further understanding of MCI and AD.