肺纤维化风险预测的临床生物化学模型

A functional laboratory biochemical model for the prediction of pulmonary fibrosis risk

目的应用数据挖掘技术分析肺纤维化(pulmonary fibrosis,PF)患者及对照组血清生物化学数据,为疾病的早期诊断提供有益线索。方法本研究收集PF患者(n=29)和健康对照(n=55)的生物化学检测数据,采用Z-计分和Log2转换方式将数据进行归一化处理,用主成分分析与贝叶斯回归分析方法提取特征指标,构建PF患者血清生物化学指标的风险预测模型,并进行受试者操作特征分析。结果与正常对照相比,PF患者血清中的α-羟丁酸脱氢酶、乳酸脱氢酶、白蛋白、白蛋白与球蛋白比值、前白蛋白及钙浓度差异有统计学意义(P<0.05)。其中α-羟丁酸脱氢酶是一种有效预测潜在PF风险的生物化学检测指标。结论本研究对PF血清生物化学数据进行挖掘分析,成功构建了PF血清生物化学预测模型。

Objective To determine promising indices for early diagnosis of pulmonary fibrosis(PF)by data extraction and analysis.Methods Biochemical data from patients with PF(n=29)and healthy controls(n=55)were collected and normalized by Z-score indexation and Log2 transformation,followed by principal component and Bayesian probit regression analyses.Signature parameters for PF were identified and used for discriminative function modeling and receiver operating characteristic analysis.Results Theα-hydroxybutyric dehydrogenase(HBDH),lactic dehydrogenase,albumin,albumin:globulin ratio,prealbumin,and calcium parameters were significantly different between PF and healthy control samples(P<0.05),and discriminant functions of PF and health were constructed.HBDH was found to be the efficient parameter that could indicate the potential risk of PF.Conclusion This study is attempt to investigate PF serum biochemical characteristics by mining of clinical biochemical data.Discriminant functional model for predicting PF with signature biochemical parameters were constructed successfully.