基因芯片数据预处理方法(LnMR和RAln)的评估和比较  

Evaluation and comparison of GeoChip data pre-processing methods: LnMR and RAln

作　　者：郑乔舒 岳淏伟 杨云锋[1] 

机构地区：[1]清华大学环境学院环境模拟与污染控制国家重点联合实验室,北京100084

出　　处：《微生物学通报》2015年第5期817-825,共9页Microbiology China

摘　　要：【目的】评估并比较两种基因芯片数据预处理方法(Ln MR和RAln)。【方法】以西藏高寒草甸草原夏季放牧实验和中国东部农田土壤移栽与玉米种植交互作用实验的两套基因芯片数据为例,利用等级-丰度曲线、均匀度指数、单因素方差分析、Q-Q图、α多样性和响应比等统计方法评估预处理效果。【结果】两种方法均能有效缩小极值和信号差异,改善信号分布,减小随机误差,提高数据正态性,增强实验结果的趋势,使芯片数据更适于进一步统计分析。两种预处理方法各有特点,Ln MR法更适合检测不同处理间微生物结构差异,RAln法可以一定程度上消除基因芯片测定的系统误差。【结论】Ln MR法和RAln法是两种行之有效的基因芯片预处理方法,在实际分析中研究者应根据研究需要合理选择。[Objective] To evaluate and compare two GeoChip data pre-processing methods, LnMR and RAIn. [Methods] The rank-abundance curve, evenness indice, one-way ANOVA, Q-Q plot, a diversity indice and response ratio were used to evaluate the pre-processing methods of GeoChip data from two recently published studies, a summer grazing experiment in the Tibetan grassland and a field study on the mutual effects of soil transplant and maize cropping. [Results] Both methods are efficient in removing or diminishing extreme values, optimizing data distribution, reducing random errors, improving data normalization and manifesting experimental differences, which makes GeoChip data more suitable for further statistical analysis. In particular, LnMR is more suitable for detecting subtle differences of microbial community compositions among different treatments, whereas RAIn is more efficient in removing systematic errors. [Conclusion] LnMR and RAIn are two powerful GeoChip data pre-processing methods, and should be applied with caution.

关 键 词：基因芯片 数据预处理 微生物功能基因 微生物生态 LN MR RAln 

分 类 号：Q811.4[生物学—生物工程] Q93