Nomogram model for predicting the risk of post-stroke depression based on clinical characteristics and DNA methylation

OBJECTIVE: To construct a comprehensive nomogram model for predicting the risk of post-stroke depression (PSD) by using clinical data that are easily collected in the early stages, and the level of DNA methylation, so as to help doctors and patients prevent the occurrence of PSD as soon as possible. METHODS: We continuously recruited 226 patients with a history of acute ischemic stroke and followed up for three months. Socio-demographic indicators, vascular-risk factors, and clinical data were collected at admission, and the outcome of depression was evaluated at the third month after stroke. At the same time, a DNA-methylation-related sequencing test was performed on the fasting peripheral blood of the hospitalized patients which was taken the morning after admission. RESULTS: A total of 206 samples were randomly divided into training dataset and validation set according to the ratio of 7:3. We screened 24 potentially-predictive factors by Univariate logistic regression and least absolute shrinkage and selection operator (LASSO) regression analysis, and 10 of the factors were found to have predictive ability in the training set. The PSD nomogram model was established based on seven significant variables in multivariate logistic regression. The consistency statistic (C-index) was as high as 0.937, and the area under curve (AUC) in the ROC analysis was 0.933. Replication analysis results in the validation set suggest the C-index was 0.953 and AUC was 0.926. This shows that the model has excellent calibration and differentiating abilities. CONCLUSION: Gender, Rankin score, history of hyperlipidemia, time from onset to hospitalization, location of stroke, National Institutes of Health Stroke scale (NIHSS) score, and the methylation level of the cg02550950 site are all related to the occurrence of PSD. Using this information, we developed a prediction model based on methylation characteristics.