文章题目
响应老大生信菜鸟团推文：一起来看文献吧！
菜鸟团一周文献推荐(No.1)
本次阅读文献题目： Five-DNA methylation signature act as anovel prognostic biomarker in patientswith ovarian serous cystadenocarcinoma.
生词
hypermethylation 高甲基化
hypomethylation 低甲基化
目的
The potential clinical significance of methyla-tion biomarkers serving as molecular prognostic markerswas examined using Kaplan–Meier method and receiveroperating characteristic (ROC) analysis.
结果
结果一
Identification of DNA methylation markers associatedwith the OS of patients in the training dataset
单因素cox回归：找到 1630 DNA methylation sites (P< 0.05)
多因素cox回归：找到5个甲基化味道 (cg05254747, cg13652336,cg25123470, cg06038133, and cg04907664)
这5个被认为是预测病人生存期的最佳标记。
同时文章有下面一段描述，是关于risk score的
在《基于TCGA数据库筛选乳腺癌不良预后相关mi+RNAs及风险评估》也看到过，描述如下：
我们使用R，使用Survival程序包的Predict函数，根据公式风险评分 =h0(t)(Exp miRNA1* βmiRNA1+...... +ExpmiRNAn*βmiRNAn)计算出每位患者的风险评分，以 风险评分的中位值将患者分为两组，大于中位值的患者为高风险组，小于或等于中位值的患者为低风 险组。
计算如下图：存活风险评分(SRS)
现在我的疑问是这个值是哪来的，应该就是Survival程序包的Predict函数得来的，需要后面碰到再继续。
关于单因素和多因素cox回归，如下描述
好了，下面这个就让我明白了，那么就是这个AUC到底是什么作用了，不是从万千丛中挑选了几只嘛，那么是根据什么挑选的呢？下面是Materials and methods的描述：
Subsequently, AUC wasused to measure and compare the model performance;the model with a higher predictive performance was eventually selected out. The model could be used to construct a risk score formula that would be helpful topredict survival. The prognostic risk scores for each pa-tient were calculated based on this formula. According to their prognostic risk scores, these patients were ranked and further separated into“low-risk”and“high--risk”groups using the median risk score as the cutoff point. Patients with risk score higher than the medianrisk score were assigned to the high-risk group, whereas patients with lower risk were assigned to the low-risk group. After that, the Kaplan–Meier estimator, a non-parametric statistic, with log-rank test (Mantel–Cox)was used to calculate the cumulative survival time and compare the differences in OS between the two groups.Kaplan–Meier curves were drawn using the“survival”package.
总结上面就是，根据AUC的值来选择模型，哪些基因组成的AUC值最高，那些基因组成的那个模型就是最好的。然后用这个模型。然后对这个模型进行一个风险评分。接下来根据预后分险评分，用median risk score将病人分为高低风险组。接下来用Kaplan–Meier（“survival”包）来对刚刚分出来的高低风险组进行一个生存分析的预测，从而来判断这个模型是否合格。
结果二
Association between five-DNA methylation signature andpatient OS in the training and validation datasets
上面的cox回归，通过Hazard ratios (HRs)可以说明这个five-DNA methylation与病人生存相关。
Hazard ratios (HRs) from the Cox regression analysis indi-cated that the five-DNA methylation signature was signifi-cantly associated with the OS of patients (P<0.001, HR2.72, 95% CI 2.03–3.65).
用 Kaplan–Meier analysis 来评价这个新建立的five-DNA methylation的效能
The Kaplan–Meier analysis wasperformed in the training and validation datasets to deter-mine the potential predictive value of this five-DNA methylation signature in the prognosis.
上面的截图说明这个five-DNA methylation could distinguish high-risk patients fromlow-risk patients, implying its significance in theprognostic prediction of OSC.
同时，高低风险组的这个5个甲基化位点的高低也进行了分析：
high-risk patients exhibited signifi-cantly lower methylation levels for cg05254747 andcg04907664 and significantly higher methylation levelsfor the other three methylation sites in both training
结果三
Evaluation of the predictive performance of the five-DNA methylation signature using ROC analysis
利用ROC分析评估five-DNA methylation特征在预测生存中的敏感性和特异性，进一步评估评估数据集中five-DNA methylation特征的预测准确性。five-DNA methylation 特征的AUC为0.715 (P< 0.001, 95% CI 0.62-0.81)(图2)，表明five-DNA methylation具有较高的敏感性和特异性。因此，该方法可用于预测OSC患者的预后生存期，具有较高的准确性，在临床应用中可能具有重要的意义。
结果四
Predictive performance of the five-DNA methylationsignature based on different regrouping methods
由于预后相关因素有很多，比如 age, stag, histologicgrade, and size of residual tumor after cytoreductivesurgery, and the reproducibility was poor in the prog-nostic markers identified by different groups等。
作者用三个年龄段（小于50、51-60、大于61）的高低风险组，来看生存时间，并且给出了ROC曲线，也就可以看到AUC值。三个年龄段的高低风险组，均具有显著差异，下面只展示第一张结果图。
接下来 作者把其他与生存相关的临床信息进行了以表格形式的展示，来证明他们这个ive-DNA methylation的棒棒！
结果五
Comparison of the five-DNA methylation signature withother known prognostic biomarkers
将本文中的five-DNA methylation与其他文章研究中biomarker进行比较，得到下面的图，同样是通过AUC值的高低来判断谁最棒，可见five-DNA methylation是最棒的。
总结
通过结果可以了解作者这篇生存分析的思路，是我最大的收获！