运动员与体力劳动者代谢组学判别模型的建立

不同职业的人群健康状态不同,需要不同的健康管理方法,根据各类人群的体质特征建立健康状态的评估方法有助于开展个性化的健康指导.招募运动员(Athlete,n=31)和体力劳动者(Labour,n=42)共73人,分别收集两组志愿者的晨尿.运用一维核磁共振(1D NMR)技术检测尿液中的代谢产物.建立主成分(PCA)及正交偏最小二乘判别分析(OPLS-DA)模型筛选2类人群间的差异代谢标志物.通过可接收操作特征曲线(ROC)评价代谢标志物的假阳性特征,t-test检验代谢标志物的显著性.利用代谢标志物建立两类人群的偏最小二乘判别分析(PLS-DA)预测模型.模型的有效性通过内部交叉、置换检验和外部预测检验确认.结果显示2类人群之间差异的代谢物有24个,通过其中20个代谢标志物建立的预测模型最优(曲线下面积AUC=0.998).内部交叉验证的误判率(FDR)分别为3.2%和0.内部置换检验的p=3.34×10~(–5).外部预测检验误判率为0.这为不同职业人群健康预测模型的建立提供了思路.

Urinary Metabonome Differentiates Athletes and Labor Workers

Under the concept of personal-based health care, different health management strategies are needed for different populations. To achieve this goal, the first step is to characterize the health-related differences among different populations. To this end, we recruited a total of 31 athletes and 42 labor workers to exam population-level differences in their urinary metabonome. First morning urine was collected and stored at-80℃ until use. 1H NMR spectra of the urine samples were collected on a 600 MHz spectrometer. The data collected were then used to build supervised and unsupervised pattern recognition models (PCA model and OPLS-DA model) to differentiate the two populations. Metabolites contributing significantly to the population difference in urinary metabonome were identified byVIP plot, among which false positives were discovered by receiver operating characteristic curve (ROC) andt-test. Predictive PLS-DA model was built, and validated by internal cross-validation, permutation tests and external prediction. The results showed that a PLS-DA model built upon 20 discriminating metabolites had the best predictive accuracy (AUC = 0.998), and the most significant level (p= 3.34×10–5). In addition, all samples from the external prediction set were classified correctly, suggesting that the PLS-DA model built upon 20 discriminating metabolites had high sensitivity and specificity.