High-throughput sequencing is great, but don’t forget the importance of analysis!

High-throughput sequencing has revolutionised microbial ecology; however, the method used to analyse and interpret the large datasets produced is extremely important.  This was highlighted by Cunning et al. (2017) when studying Symbiodinium metacommunities of corals. While this paper contained a number of interesting findings relating to coral Symbiodinium communities, I will focus on the authors comparison of analysis methods on high-throughput sequencing datasets.

Cunning et al. (2017) extracted genomic DNA from coral tissue samples and sequenced the gene marker ITS2 using Illumina MiSeq. Three different methods of clustering were used to group these sequences into operational taxonomic units (OTUs); clustering at 100% identity, clustering at 97% identity across samples and clustering at 97% identity within samples. Using three clustering methods allowed the authors to investigate the most appropriate method to analyse Symbiodinium OTUs generated from high-throughput sequencing, which is especially important to generate accurate assessments of diversity when using ITS2 due to its high intragenomic variability.

Clustering at 100% identity was found to overestimate diversity as every ITS2 sequence is treated as a different Symbiodinium type, therefore clustering at 97% similarity is necessary to minimise intragenomic variation. When clustering at 97% identity, across sample clustering often allocated the same dominant OTU to samples containing different dominant sequence variants; however, clustering within samples assigned different OTUs. These results show that clustering across samples minimises interspecific variation and so underestimates diversity, while clustering within samples more accurately represents the true diversity of Symbiodinium and so was chosen to analyse the dataset.

However, there are limitations to this method as it relies on two assumptions; a single dominant Symbiodinium type is found on most coral colonies and different variants of ITS2 sequences identify different Symbiodinium types, even when there is only one nucleotide difference. However, these assumptions are not always met. Coral colonies often contain more than one Symbiodinium type; however, when this occurs they are often from different clades and so their ITS2 sequences should be different enough to be identified as distinct OTUs. As a result, problems are only likely if very closely related Symbiodinium types occur in the same sample. Another possible limitation occurs when multiple co-dominant sequences are present in a coral colony, as different OTUs could be assigned to colonies that contain similar sequence assemblages with only minor differences in abundance. However, this problem could be avoided by using more complex criteria to assign taxonomy or using other markers (such as microsatellites).    

The authors suggest further research is needed in the area to confirm their support for this method and make improvements where necessary. Overall, despite the potential limitations, 97% identity within sample clustering appears to be the best method of analysing datasets from ITS2 high-throughput sequencing. The findings of this paper will allow more accurate analysis of high-throughput sequencing and so improve understanding of Symbiodinium communities; it has also highlighted the importance of selecting the best method of analysis which is vital in all fields of research where high-throughput sequencing is used. In my opinion, the main take home message of this paper is that while high-throughput sequencing has undoubtedly improved our ability to understand marine microbial ecology is it important to determine the most appropriate method of analysis to ensure that the huge datasets produced are interpreted accurately and truly reflect the systems being studied.

Reviewed Article:

Cunning, R., Gates, R. D., & Edmunds, P. J. (2017). Using high-throughput sequencing of ITS2 to describe Symbiodinium metacommunities in St. John, US Virgin Islands (No. e2925v1). PeerJ Preprints. https://peerj.com/articles/3472/