We also left out sequence reads less than 100 bp in length, or with one or more ambiguous nucleotides (N) in order to use only good quality sequences in further analysis [24]. The sequences that passed the initial quality control were analysed with Mothur [25]. Bacterial

and archaeal sequences were aligned to SILVA alignment database [26]. Aligned sequences were preclustered, distance matrices were prepared and the sequences were clustered to operational taxonomic units (OTUs) using average neighbor algorithm. Rarefaction curves PD0332991 ( Additional file 1) and ACE [27] and Chao1 [28] indices (Table 3) were calculated to estimate the community richness, and Simpson and Shannon indices [29] were used in assessing the diversity present in samples. We also calculated Venn diagrams and dendrograms describing the shared OTUs within samples and similarity between the structures of communities, respectively. The dendrograms were constructed using the Yue & Clayton similarity value, θYC[30]. Fungal sequences were aligned and distance matrix was prepared using Mothur pairwise.seqs command. Clustering and

other selleckchem downstream analyses were carried out as with Bacteria and Archaea. Taxonomic affiliations were determined with BLAST [31] Liproxstatin1 and Megan [32]: sequence reads were queried against the NCBI nucleotide database (nr/nt) [33] and the results were analysed using Megan. Fungal sequences affiliated Molecular motor to Plantae or Animalia were removed from the dataset.

We applied Ribosomal Database Project’s Classifier [34] to determine the bacterial and archaeal groups present in samples. The sequences have been deposited in the Sequence Read Archive (SRA) at EBI with study accession number ERP000976. The most abundant microbial groups are presented in Figure 2. Figure 2 Overview of microbial diversity in AD samples. Barplots showing relative sequence numbers of most common microbial groups in samples M1, M2, M3 and M4. Statistical methods Redundancy analysis (RDA) ordination technique [35, 36] was used to explore the relationships between microbial community composition and variation in physical and chemical parameters. Microbial composition data from both sequencing and microarray were used as dependent variables and six selected physico-chemical parameters as constraints. Only the 12 most abundant microbial classes from sequencing and 12 strongest microarray probes were included in the analysis. Correlation coefficients were used as inertia in the model and plotting. Three different constraining variables were used per analysis because the number of constraining variables is restricted to n-1 (n referring to the number of observations; here M1-M4). Analyses were done using R-software package vegan v. 1.17-12 [37].