Analysis of metagenomic data containing high biodiversity levels

In this paper we have addressed the problem of analysing Next Generation Sequencing samples with an expected large biodiversity content. We analysed several well-known 16S rRNA datasets from experimental samples, including both large and short sequences, in numbers of tens of thousands, in addition to carefully crafted synthetic datasets containing more than 7000 OTUs. From this data analysis several patterns were identified and used to develop new guidelines for experimentation in conditions of high biodiversity. We analysed the suitability of different clustering packages for these type of situations, the problem of even sampling, the relative effectiveness of Chao1 and ACE estimators as well as their effect on sampling size for a variety of population distributions. As regards practical analysis procedures, we advocated an approach that retains as much high-quality experimental data as possible. By carefully applying selection rules combining the taxonomic assignment with clustering strategies, we derived a set of recommendations for ultra-sequencing data analysis at high biodiversity levels.