Construction and validation of diagnostic models for Alzheimer′s disease based on serum microRNAs
10.3760/cma.j.cn113661-20210131-00061
- VernacularTitle:基于血清小分子核糖核酸的阿尔茨海默病诊断模型构建与验证
- Author:
Zhengluan LIAO
1
;
Heng SU
;
Yan CHEN
;
Enyan YU
Author Information
1. 浙江省人民医院精神卫生科 浙江省心理保健基地 浙江省医学重点学科，杭州 310014
- Publication Type:Journal Article
- Keywords:
Alzheimer disease;
MicroRNAs;
Artificial intelligence;
Diagnostic model
- From:
Chinese Journal of Psychiatry
2021;54(5):331-336
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct and evaluate a predictive diagnosis model of Alzheimer′s disease(AD) based on serum microRNA (miRNA).Methods:Downloaded the serum miRNA chip data of 1 021 AD patients and 288 healthy controls from the gene expression omnibus (GEO) database, a total of 1 309 samples were obtained. Matched AD patients and healthy controls by age, and screened out 494 samples for training and verifying the diagnostic model. Sorted the miRNA expression value from high to low and selected the top 1 000 probes for the next step. According to the ratio of 7∶3, all 494 samples were randomly divided into a training group and a validation group. LASSO regression was used to screen miRNAs, combined with gender, APOE4 genotype and miRNA data, stepwise regression was used to further screen independent variables and construct a multi-factor diagnostic model. Receiver operating characteristic (ROC) curves and calibration curves were used to verify the accuracy of the model in the training group, validation group, training group+validation group, and 1 309 total samples; A Nomogram was drawn for actual case prediction.Results:The area under the Curve of ROC of the multi-factorial diagnostic model in the training group, the validation group, the training+validation group and the total samples of 1 309 cases were 0.870, 0.831, 0.842 and 0.826, respectively, showing high predictive effectiveness. An AD case was used to show an example of applying the method and its predicted results (the total score of the patient was 521, and the probability of predicted NC was 0.022 8, indicating that the sample was AD, which was consistent with the actual diagnosis).Conclusions:The diagnostic model based on 11 serum miRNA, gender, and apolipoprotein E (ApoE4) genotypes can well predict AD and may provide the possibility for clinical prediction of AD.
