To further understand the effect of sequencing errors on PCA, we performed procrustes analysis with the original datasets vs. datasets with simulated base error rates of 1% (Additional file 1: Figure S4). All pair-wise comparisons show that sequencing errors did not greatly affect the

PCA based on the Jaccard distance, in support of our conclusions detailed above. Microbial composition and biomarker determination The two datasets showed significantly Selleck LY2109761 different community structures (Figure 3a). Although the gut flora of all subjects consisted primarily of Firmicutes, Bacteroidetes and Proteobacteria, the relative abundance of these microbes MK-4827 mw varied significantly. Compared to the V6F-V6R dataset, the V4F-V6R dataset identified higher levels of Bacteroidetes and lower levels of Firmicutes (Figure 3c). Interestingly, the categories of genera identified by the two primer sets were similar to each

other, while the relative abundance of the genera differed (Figure 3b). We suggest that both the primer bias and sequencing errors contributed to these differences, but the former may have contributed more because sequencing errors usually occur buy CUDC-907 at a very low frequency and do little to change the overall relative abundance. Several studies have compared microbial community structures using different primer sets [11, 21]. These studies usually found significant primer biases in the evaluation of microbial ecology. However, here we demonstrated for the first time that PCA using the Jaccard distance was minimally affected by primer bias and differences in sequencing quality, suggesting the feasibility of performing meta-analysis for sequences obtained from different sources. Figure 3 Microbial structure at phylum

and genus level. (a) Microbial structures new of each individual determined at the phylum level by the two primer sets. (b) Microbial structures of each individual determined at the genus level by the two primer sets. (c) Relative abundance of Firmicutes and Bacteroidetes determined by the two primer sets. We used LEfSe for the quantitative analysis of biomarkers within different groups (Figure 4 and Additional file 1: Figure S2). This method was designed to analyze data in which the number of species is much higher than the number of samples and to provide biological class explanations to establish statistical significance, biological consistency, and effect-size estimation of predicted biomarkers [16]. To simulate a simple meta-analysis, we compared the microbiomes of four individuals two at a time (e.g., A vs. C and B vs. D). The results demonstrated that when the data from the two individuals came from the same dataset, their biomarkers were generally similar.