Development and validation of a decision tree classification model for the essential hypertension based on serum protein biomarkers.

Background: Essential hypertension (EH) is a key risk factor for cardiovascular disease. However, the etiology of EH is complex and unknown. So far, there is no good protein biomarker for screening EH. The purpose of this study was to discover potential biomarkers for EH by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) and establish a decision-tree classification model. Methods: A total of 108 patients with clinically confirmed EH and 105 HC were enrolled in the present study from September 2020 to April 2021 and were randomly divided into the training group and the blind-test group. The serum protein expression profiles were performed using MALDI-TOF MS combined with magnetic beads with weak cation exchange (MB-WCX). The training group, which comprised 54 EH patients and 53 HC, was used to screen the statistically differential protein peaks by SPSS 19.0 and construct a decision-tree classification model by C5.0 algorithms of SPSS Modeler 18.0. All protein peak intensities of samples in the blind-test group, which comprised 54 EH patients and 52 HC, were used to verify the diagnostic capabilities of the model by classification model. Results: EH patients had higher age, systolic and diastolic blood pressures than HC group. The intensities of 60 protein peaks differed significantly between the EH patients and HC. An optimal decision-tree classification model of EH was successfully established with mass-to-charge ratios of 1,326.7, 1,785.3, 4,228.0, and 8,963.8 as differential protein peaks by the software analysis. The decision-tree classification model was able to distinguish between EH patients and HC and had a sensitivity of 94.44%, a specificity of 94.33%, an accuracy of 94.39%, and an area under the receiver operating characteristic (ROC) curve of 0.96. The blind-test results indicated a sensitivity of 87.04%, a specificity of 88.46%, an accuracy of 87.74%, and an area under the ROC curve of 0.928. Conclusions: MALDI-TOF MS combined with MB-WCX can be used to screen for serum differential protein expression profiles in EH patients. The decision-tree classification model based on mass-to-charge ratios of 1,326.7, 1,785.3, 4,228.0, and 8,963.8 could provide a new and reliable method for screening and identifying EH with high sensitivity and specificity.