Microbiome meets classical microbiology: quantifying sample CFU using 16S rRNA gene sequencing data

Abstract

Background Next-generation sequencing (NGS) has been extensively employed to perform microbiome characterization worldwide. As a culture-independent methodology, it has allowed high-level profiling of sample microbial composition. However, most studies are limited to sample information regarding relative bacterial abundances (sample proportions), ignoring scenarios in which sample microbe biomass can vary widely. Here, we develop an equivolumetric protocol for amplicon library preparation capable of generating NGS data responsive to input DNA, recovering proportionality between observed read counts and absolute bacterial abundances within each sample. Within a determined range, we show that the estimation of sample colony-forming units (CFU), the most common unit of bacterial abundance in classical microbiology, is challenged mostly by resolution and taxon-to-taxon variation. We propose Bayesian cumulative probability models to address such issues.Results Observed read counts were consistently proportional to input DNA, total microbial load, and bacterium-specific sample abundances, although a saturation tendency was observed as abundances increased. Using Bayesian cumulative probability models, predictive errors in sample CFU estimation varied constantly below one order of magnitude - as measured by the mean absolute log10-ratio (MALR). For total microbial load, observed MALR was no greater than 0.2 during both cross-validation and validation on a test dataset. For observed bacteria, estimation of taxon-specific CFU showed MALR values of at most 0.5. We also performed leave-one-group-out cross-validation to assess predictive performance for previously unseen bacteria. While most bacteria showed MALR no greater than 1, such a threshold was exceeded only by Bacillus cereus.Conclusions Being able to estimate sample CFU in a high-throughput fashion has a wide range of applications, from the study of built environments to public health surveillance. This study shows that equivolumetric protocols along with cumulative probability models allow sample CFU estimation from microbiome datasets. Further, our approach has the potential to generalize to previously unmodeled bacteria, an important feature in high-throughput settings. Lastly, it remains clear that NGS data are not inherently restricted to sample proportions only, and microbiome science can finally meet the working scales of classical microbiology.