Benchmarking a 16S rRNA sequencing protocol for microbiome analysis in low-moisture grain environments

Microbial amplicon sequencing studies are an important tool in food and biomedical research. However, accurate interpretation of the 16S rRNA gene survey requires specialized software and an algorithm to convert raw sequencing data into reliable taxonomic profiles. Given the existence of multiple bioinformatics pipelines varying in sequence aggregation strategies, reference databases, and filtering parameters, there is little to no consensus on best practices for LMF processing systems. In this study, we systematically assessed discrepancies in taxonomic composition, alpha diversity, and beta diversity across 32 combinations of bioinformatics workflows, based on eight widely used 16S rRNA pipelines and four taxonomic databases, applied to 16S rRNA gene sequences extracted from wheat milling environments (n = 160). Weighted composite scores were used to select the top 10-performing workflow combinations for downstream analysis. Taxonomic assignments were broadly similar across workflows at the family and genus levels; however, genus-level diversity metrics were more sensitive to workflow choice. At the family level, diversity metrics were conserved across pipeline-database combinations (Chao1: 22.97 ± 2.20-24.92 ± 2.04; Shannon: 2.59 ± 0.19-2.74 ± 0.18; InvSimpson: 10.63 ± 1.25-11.27 ± 1.06; Bray-Curtis: 0.528-0.556; Jaccard: 0.557-0.582), whereas at the genus level both alpha and beta diversity exhibited wider ranges and larger dispersion (Chao1: 45.27 ± 5.68-50.20 ± 5.64; Shannon: 2.54 ± 0.24-2.73 ± 0.22; InvSimpson: 10.37 ± 1.4-11.03 ± 1.05; Bray-Curtis: 0.79-0.82; Jaccard: 0.79-0.80). Furthermore, ASV vs. OTU workflows were comparable across the evaluated metrics; however, ASVs showed numerically higher values for some genus-level measures than OTUs because they can resolve variation down to the single-nucleotide level, thereby retaining low-abundance features important for LMF safety. This work paves the way toward using bioinformatics and 16S pipelines to characterize sparse, low-density, and uneven samples in low-moisture environments.