Multivariate tests and multiple test-procedures for abundance data of microorganisms in consideration of information from phylogenetic sequences

微生物群落组成分析是农业、医学和生态学研究的重要组成部分。如今，首选的方法是直接基于DNA样本，因此不依赖于实验室条件下微生物的繁殖。从分析的电泳法到专门的微阵列，再到现代下一代测序技术，如焦解测序或Illumina MiSeq测序，发展的进步增加了检测到的OTU(操作分类单位)的数量，并增加了在样品内和样品之间直接解释和比较它们的能力。焦磷酸测序的结果是对OTUS出现次数的量化，它在物种规模上是独立的，作为副产品提供了相应序列的信息，也可以用来量化OTUS的系统发育相似性。目前，国际上的研究小组将这种补充信息纳入统计分析。建立了计算成本高昂的生态距离方法，将丰度和系统发育距离的信息结合起来。加权Unifrac距离和DPCoA(双主坐标分析)这两个最流行的变种的共同理论基础已于去年发表。这项应用的第一个目标是调整我所开发的多变量方法以使用额外的系统发育信息，并将结果与已发表的方法进行比较。我们研究所的方法是基于距离度量的排列或旋转检验，其主要目的是将这些方法与单变量检验结合使用，作为多重检验过程中的构建块，尽可能地将高维数据划分为更小的变量集(例如，在更高的分类水平上)，甚至划分为单变量(OTU)，同时仍在较强意义上控制家族错误率。为了实现这一点，我们研究所(或与我们研究所一起)开发的几个多重测试程序将被调整以使用系统发育距离。我们将与位于奎德林堡/布伦斯韦格的联邦栽培植物研究中心Julius Kühn-Institut的合作伙伴密切合作。

Analysis of the composition of microbial communities is an important component of agricultural, medical and ecological research. Nowadays the preferred methods are based directly on DNA samples and therefore are independent of the reproduction of microbial organisms under laboratory conditions. The developmental progress from electrophoretic methods of analysis to specialized microarrays and furthermore to modern next generation sequencing techniques as pyrosequencing or sequencing by Illumina MiSeq increased the number of detected OTUs (operational taxonomic units) and the ability to directly interpret and compare them within and between samples. The result of pyrosequencing is a quantification of the occurrences of OTUs which is independent in scale for species and as a byproduct provides information of the corresponding sequences which can be used to quantify the phylogenetic similarity of OTUs, too.Currently, international groups of researchers focus on incorporating this supplementary information in statistical analysis. Computationally expensive methods for ecological distances were established that combine information from abundance and phylogenetic distances. A common theoretical basis for the two most popular variants, the weighted Unifrac distance and the DPCoA (double principal coordinate analysis), has been published last year.The first target of this application is to adjust multivariate methods that were developed in our institute to the use of additional phylogenetic information and to compare the results against published methods. The methods from our institute are permutation or rotation tests based on distance measures.The main target is to use these methods afterwards together with univariate tests as building blocks in procedures for multiple testing which partition high dimensional data as far as possible into smaller sets of variables (e.g. on a higher taxonomical level) or even into single variables (OTUs) while still controlling the familywise error rate in the strong sense. To achieve this, several multiple test procedures that were developed in (or together with) our institute will be adjusted to the use of phylogenetic distances. We will work closely together with our partners from the Julius Kühn-Institut, a Federal Research Centre for cultivated plants in Quedlinburg/Braunschweig.