BCC2020

➞ Abstract
The development of sequencing technologies to evaluate bacterial microbiota composition has allowed insight into the role of the microbiome in human health. However, the variety of methodologies used to prepare and analyze samples for microbiota composition can introduce artifacts, including errors and biases. These artifacts alter our perception of bacterial diversity and our final interpretation of microbiota differences among samples. Using a mock bacterial microbiota of known composition and abundance, we performed a translational bioinformatic pipeline evaluation of various PCR conditions, amplicon library preparation methods, and bioinformatic analyses to gain insight into methodological sources of artifacts.
Genomic DNA was extracted from pure cultures of individual mock bacterial isolates (n = 43), quantified, and then pooled. To compare the effects of PCR on the development of artifacts, we performed all possible permutations of three polymerases, three alternative primer pairs targeting varying regions of the 16S rRNA gene, two barcoding approaches, five elongation times, two annealing temperature offsets, and two amplicon cleanup methods. All individual PCR reactions were sequenced on an Illumina MiSeq platform. Bioinformatic analysis was performed with three different microbiome analysis pipelines, including DADA2, mothur, and QIIME2. Resulting sequence variants were classified as expected or unexpected, and missing members of the mock community were identified. Unexpected reads were further identified as an artifact representative of either chimeras using DECIPHER, mock community sequences containing mismatches or indels, primer dimers, 16S rRNA contamination, or non-16S rRNA off-target amplification.
We found that primer choice accounted for a significant amount of discord between the mock community and sequence output. Additionally, longer amplicon fragment lengths negatively impacted the quality of sequencing reads. Polymerase choice, annealing temperature, and elongation time negligible impacts on sequencing results. QIIME2 and DADA2 performed similarly using standard pipelines and produced the most accurate results. The use of mothur was associated with a high number of operational taxonomic units which were classified as contamination and increased the interpretation of community diversity.
By employing a defined mock community, this full factorial experiment allowed us to gather insight into methodological sources of pipeline artifacts and allow us to identify a methodology that results in an optimized workflow for improved examination of microbiota composition. This workflow enables full-circle analysis of samples with superior precision in comparison to current workflow standards.
This poster will be presented live at BCC West.