基于复杂网络结合生存分析的中医药治疗肺癌的核心有效处方的发现研究

Effective core formulae for lung cancer based on complex network and survival analysis

有效核心方的研究有助于中医药治疗的经验总结并揭示其内在的组方规律,该研究以探索中医药治疗肺癌的核心有效处方的数据挖掘方法为目的,提出了应用个人用药指纹谱表征患者的阶段用药信息,通过D指数法筛选潜在获益药物群,并在药物的配伍网络的基础上采用复杂网络的极大团算法结合生存分析模型挖掘分析中医药治疗肺癌的核心有效处方,并对龙华医院322例肺癌用药的生存数据进行了挖掘分析,通过用药指纹谱的分析初步显示了用药水平与生存期的关联性,D指数法筛选到了43个潜在获益药物,通过复杂网络结合对数正态分布的参数生存模型分析得到了2个潜在的对患者生存期获益的核心处方,并通过了多因素生存模型的验证。从2个潜在核心有效方的组成看,基本体现了中医药治疗肿瘤的扶正思想,所得结果也基本符合临床实际情况。该研究结果表明,采用复杂网络结合生存分析模型能够对复杂中医肿瘤临床的生存数据进行挖掘,是发现中医药治疗肿瘤的潜在核心有效处方的有效方法之一。

The study on the effective core formulae（ CEF） not only summarized traditional chinese medicine（ TCM） treatment experience,but also helped reveal the underlying knowledge in the formulation of TCM prescriptions. The aim of the present paper was to investigate the method of data mining for the discovery of core effective formulae for lung cancer. In the present study,a prescription fingerprint approach was used to characterize the staged prescription information of patients. The D index was used to screen potential beneficial herbs. Then,based on a herbal compatibility network,the maximal clique searching algorithm（ BK algorithm） and survival analysis were applied to discover CEF for lung cancer,and a mining analysis was made for the 322 cases from Longhua hospital. The correlation between prescriptions and survival time was analyzed by prescription fingerprints. Forty-three potentially beneficial herbs were obtained,and two CEFs were significant for the survival time by a parametric survival model based on lognormal distribution,the results were verified by a multivariate survival model. The rules of combination of the two CEFs basically conform to TCM onco-therapeutic theory of strengthening the body resistance and the actual conditions in clinic. All results showed that the established approach was feasible for discovering the core effective formulae for lung cancer and mining survival data for complex TCM onco-therapy.