摘要
目的由于疾病,特别是肿瘤的识别模型,其分型准确度对疾病的治疗和预后具有重要意义,因而,本研究探讨了基于基因表达谱的疾病分型识别模型建模方法。方法结合白血病基因表达谱数据分析,利用偏最小二乘判别分析(PLS-DA)对利用基因微阵列数据予以建立白血病分型模型,并与Golub等提出的建模方法相对照,比较它们的判别效果。结果基于偏最小二乘判别分析的白血病识别模型的拟合准确度和预测准确度均达到100%。结论研究表明,基于偏最小二乘判别分析的模型明显提高了白血病的分型正确率,无论是拟合精度,还是预测精度,均高于Golub等提出的方法。
Objective To discuss the modeling methods in classification of diseases based on gene expression profile, for the classification accuracy of diseases, especially tumor, is much important to improve the treatments and prognosis of diseases. Methods Partial least squares discrimination analysis （PLS-DA） was utilized to model the classification of leukaemia, based on gene expression microarray data of leukaemia. And compared with the method proPosed by Golub, PLS-DA obtains the optimum classification accuracy of diseases. Results The accuracy of PLS-DA leukaemia model, included goodness-of-fit and prediction precision. Conclusion In modeling the classification of diseases, PLS-DA has better goodness-of-fit and prediction precision than Golub＇s method.
出处
《中国卫生统计》
CSCD
北大核心
2007年第4期372-374,共3页
Chinese Journal of Health Statistics
基金
国家自然科学基金项目(30271157)
关键词
基因表达谱
偏最小二乘判别分析
肿瘤分型
Gene expression profile
Partial least squares discrimination analysis
Tumor classification
