ASV vs OTUs clustering: Effects on alpha, beta, and gamma diversities in microbiome metabarcoding studies.

Abstract

In microbial community sequencing, involving bacterial ribosomal 16S rDNA or fungal ITS, the targeted genes are the basis for taxonomical assignment. The traditional bioinformatical procedure has for decades made use of a clustering protocol by which sequences are pooled into packages of shared percent identity, typically at 97%, to yield Operational Technical Units (OTUs). Progress in the data processing methods has however led to the possibility of minimizing technical sequencers errors, which were the main reason for the OTU choice, and to analyze instead the exact Amplicon Sequence Variants (ASV) which is a choice yielding much less agglomerated reads. We have tested the two procedures on the same 16S metabarcoded bacterial amplicons dataset encompassing a series of samples from 17 adjacent habitats, taken across a 700 meter-long transect of different ecological conditions unfolding in a gradient spanning from cropland, through meadows, forest and all successional transitions up to the seashore, within the same coastal area. This design allowed to scan a high biodiversity basin and to measure alpha, beta and gamma diversity of the area, to verify the effect of the bioinformatics on the same data as concerns the values of ten different ecological indexes and other parameters. Two levels of progressive OTUs clustering, (99% and 97%) were compared with the ASV data. The results showed that the OTUs clustering proportionally led to a marked underestimation of the ecological indicators values for species diversity and to a distorted behaviour of the dominance and evenness indexes with respect to the direct use of the ASV data. Multivariate ordination analyses resulted also sensitive in terms of tree topology and coherence. Overall, data support the view that reference-based OTU clustering carries several misleading disadvantageous biases, including the risk of missing novel taxa which are yet unreferenced in databases. Since its alternatives as de novo clustering have on the other hand drawbacks due to heavier computational demand and results comparability, especially for environmental studies which contain several yet uncharacterized species, the direct ASV based analysis, at least for prokaryotes, appears to warrant significand advantages in comparison to OTU clustering at every level of percent identity cutoff.