中文摘要：

目的探究差异表达驱动的微阵列数据之降维策略。方法本文提出permutation检验和99.99%的下单侧可信区间相结合的策略用于“维度粗筛”,HotellingT2检验结合逐步筛选的策略用于寻求组间差异表达的“局部变量组合”。结果以“正常成年男子和精子运动能力低下者精子蛋白表达差异研究”之实例展示了该降维策略的实际应用效果,结果发现了“十个蛋白组合”在组间差异表达。结论本文提出差异表达驱动的微阵列数据降维策略是实用可行的。

英文摘要：

Objective To explore the strategy of dimension reduction driven by differential expression analysis for microarray data. Methods Permutation test followed by a lower 99.99 % confidence interval, was used for feature pre - selection, Hotelling T2 test with a stepwise search strategy was used to find subset of variables with the lowest p - values. Resuits With the above strategy, a subset of 10 proteins was identified as differentially expressed between asthenozoospermic patients and norm male adults. Conclusion Our strategy of dimension reduction driven by differential expression analysis for microarray data is feasible and practical.