Deep sequencing artificially inflates estimates of microbial diversity

Sequencing artifacts challenge accuracy and reproducibility when quantifying microbial diversity. To track error propagation in microbiome analyses, we analyze no-diversity amplicons, which are amplified from host genes with limited genetic diversity or from synthetic spike-ins. We find that sequencing at greater than 104 reads exponentially increased no-diversity amplicon sequence variant (ASV) richness, with hundreds of ASVs observed per sample. This striking pattern was shared with microbial amplicons (16S rRNA, ITS, gyrB, rpoB), which revealed inflated Shannon diversity with an increase in read counts for both the community and within taxa. Comparing sequencing error profiles between no-diversity and microbial amplicons showed that truncating reads to shorter lengths and use of the AVITI Element platform can mitigate, but not abolish, the impacts of artificial inflation; we recommend caution when read depths vary orders of magnitude between samples. Overall, utilizing no-diversity amplicons can help optimize parameters to improve estimates of microbial diversity.