Library preparation methodology can influence genomic and functional predictions in human microbiome research

Abstract

Observations from human microbiome studies are often conflicting or inconclusive. Many factors likely contribute to these issues including small cohort sizes, sample collection, and handling and processing differences. The field of microbiome research is moving from 16S rDNA gene sequencing to a more comprehensive genomic and functional representation through whole-genome sequencing (WGS) of complete communities. Here we performed quantitative and qualitative analyses comparing WGS metagenomic data from human stool specimens using the Illumina Nextera XT and Illumina TruSeq DNA PCR-free kits, and the KAPA Biosystems Hyper Prep PCR and PCR-free systems. Significant differences in taxonomy are observed among the four different next-generation sequencing library preparations using a DNA mock community and a cell control of known concentration. We also revealed biases in error profiles, duplication rates, and loss of reads representing organisms that have a high %G+C content that can significantly impact results. As with all methods, the use of benchmarking controls has revealed critical differences among methods that impact sequencing results and later would impact study interpretation. We recommend that the community adopt PCR-free-based approaches to reduce PCR bias that affects calculations of abundance and to improve assemblies for accurate taxonomic assignment. Furthermore, the inclusion of a known-input cell spike-in control provides accurate quantitation of organisms in clinical samples.

Keywords: genomics; microbiome; sequencing.

Fig. 1.

One-way ANOVA analysis across library preparation methods. Relative abundance measurements were calculated for the mock community across the four different protocols and analyzed for consistency between library preparations from both technical replicates. Shading in the heat map indicates relative abundance in the mock-community DNA mixture from low (green) to high (red) abundance. Adjusted P values were calculated based on a maximum P value of 0.01. Samples and organisms were clustered based on an uncentered Pearson complete linkage analysis. The letters “A” and “B” indicate technical replicates for each sample preparation.

Fig. 2.

Map of mean GC content and mean relative sequencing depth by library prep method across the genome for R. sphaeroides. (A) The complete genome for the organisms is used, including any known plasmids, and (B) sub division of the genome into 10-kb bins for mean analysis. Outer grey ring depicts the delta from 50% GC content for a sequence bin. The inner 4 colored rings depict the delta of the average sequencing depth for the bin from the average sequencing depth of the whole genome. Maximum and minimum values per ring are given in the legend.