Microbial communities are commonly studied using culture-independent methods, such as 16S rRNA gene sequencing. However, one challenge in accurately characterizing microbial communities is exogenous bacterial DNA contamination, particularly in low-microbial-biomass niches. Computational approaches to identify contaminant sequences have been proposed, but their performance has not been independently evaluated. To identify the impact of decreasing microbial biomass on polymicrobial 16S rRNA gene sequencing experiments, we created a mock microbial community dilution series. We evaluated four computational approaches to identify and remove contaminants, as follows: (i) filtering sequences present in a negative control, (ii) filtering sequences based on relative abundance, (iii) identifying sequences that have an inverse correlation with DNA concentration implemented in Decontam, and (iv) predicting the sequence proportion arising from defined contaminant sources implemented in SourceTracker. As expected, the proportion of contaminant bacterial DNA increased with decreasing starting microbial biomass, with 80.1% of the most diluted sample arising from contaminant sequences. Inclusion of contaminant sequences led to overinflated diversity estimates and distorted microbiome composition. All methods for contaminant identification successfully identified some contaminant sequences, which varied depending on the method parameters used and contaminant prevalence. Notably, removing sequences present in a negative control erroneously removed 20% of expected sequences. SourceTracker successfully removed over 98% of contaminants when the experimental environments were well defined. However, SourceTracker misclassified expected sequences and performed poorly when the experimental environment was unknown, failing to remove 97% of contaminants. In contrast, the Decontam frequency method did not remove expected sequences and successfully removed 70 to 90% of the contaminants. IMPORTANCE The relative scarcity of microbes in low-microbial-biomass environments makes accurate determination of community composition challenging. Identifying and controlling for contaminant bacterial DNA are critical steps in understanding microbial communities from these low-biomass environments. Our study introduces the use of a mock community dilution series as a positive control and evaluates four computational strategies that can identify contaminants in 16S rRNA gene sequencing experiments in order to remove them from downstream analyses. The appropriate computational approach for removing contaminant sequences from an experiment depends on prior knowledge about the microbial environment under investigation and can be evaluated with a dilution series of a mock microbial community.
Bibliographical noteFunding Information:
We also thank our funders for their generous support. L.K. was supported as a Scholar of the Oregon BIRCWH K12 Program funded by the National Institutes of Health (NIH) grants NICHD K12HD04348 and NIH/NIDDK K01DK116706. J.B. was supported by grants NIH UL1TR001881 and DK46763. A.J.W. was supported by grant NIH R01DK104718. S.M. was supported by grant NIH/NCATS UL1TR002369. L.K. and M.A. were supported by grant NIH/NEI R01EY029266 and the Rheumatology Research Foundation. M.A. was also supported by the Spondylitis Association of America.
Copyright © 2019 Karstens et al.
- 16S rRNA gene sequencing
- Low microbial biomass