Article Figures & Data
Figures
- FIG 1
Analysis of a mock microbial community dilution series reveals that contaminating bacteria increase with decreasing starting DNA. A mock microbial community consisting of 8 known bacteria was subject to 8 series of a 3-fold dilution (1:3 through 1:6,561), subject to bacterial DNA isolation and amplification, and sequenced with the Illumina MiSeq platform. (A) Number of reads per sample. (B) The proportion of reads from contaminant DNA increased with the amount of dilution. (C) Stacked bar plot representing the bacteria identified in each sample. The expected ASVs from the mock microbial community are displayed in color, while all other bacterial ASVs are in grayscale.
- FIG 2
Impact of contaminants on common alpha-diversity measures. Failure to remove contaminants from the data set leads to increased estimates of alpha diversity evaluated by the number of observed ASVs (Observed) inverse Simpson index (InvSimpson), and Shannon diversity index (Shannon). Expected results were calculated based on the expected mock community ASVs in each dilution sample.
- FIG 3
Classification of ASVs. Red, mock community ASVs correctly classified; light gray, mock community ASVs incorrectly classified; blue, contaminant ASVs correctly classified; dark gray, contaminant ASVs incorrectly classified. (A) Not correcting for contaminants leads to a large proportion of sequences being incorrectly considered mock ASVs across the dilution series. (B) Filtering by removing ASVs present in the negative control incorrectly classified many mock community ASVs as contaminants and misclassified many contaminant ASVs as mock community ASVs in the diluted samples. (C to E) Abundance filtering required an abundance of 1% to classify the majority of contaminant ASVs correctly, but it also removed mock community ASVs (E). (F to J) The Decontam frequency method did not misclassify any mock community ASV sequences and correctly classified the majority of contaminant ASVs. In highly diluted samples (D6 to D8), contaminant ASVs are misclassified as mock community ASVs. (K to N) SourceTracker performs well, correctly classifying the majority of mock community and contaminant ASVs, though some mock community ASVs are classified incorrectly as contaminant ASVs (K and L). However, in the scenario where the experimental environment is not well defined (M and N), many of the ASVs are incorrectly classified as mock community ASVs. S1, scenario 1; S2, scenario 2.
- FIG 4
Accuracy of methods to identify contaminants. The negative-control filter has poor accuracy, regardless of prevalence of contaminants. All other methods have high accuracy when the prevalence of contaminants is <5%. As the prevalence of contaminants increases, the accuracy of most methods drops, with the exception of SourceTracker with well-defined experimental conditions. Thr, threshold.
- FIG 5
Example of the effect of contaminant removal on common microbiome summary measures (relative abundance of ASVs, alpha diversity). (A) Relative abundance of Lactobacillus. (B) Relative abundance of Escherichia/Shigella. (C) Alpha diversity summarized by the inverse Simpson index. For brevity, results are shown for three dilution series samples representing low (less than 5%, D3), moderate (between 10% and 50%; D6), and high (greater than 50%, D8) levels of contaminant ASVs. Expected results shown by solid black line are based on the composition of the expected mock community ASVs in the specified dilution sample. See Fig. S1 and S2 for results across the entire dilution series.
Tables
- TABLE 1
Impact of decreasing starting material for 16S rRNA gene sequencing
Parameter Data by dilution D0 D1 D2 D3 D4 D5 D6 D7 D8 No. of reads 251,419 172,915 250,861 247,581 216,341 136,081 128,053 41,071 40,927 No. of unique ASVs 18 20 114 172 262 312 381 147 193 No. of unique genera 15 16 64 61 80 86 107 58 74 % contaminantsa 0.1 0.1 1.8 4.5 12.0 27.9 64.5 55.8 80.1 ↵a Percent contaminants calculated as the percentage of sequences in each sample that were not an exact match to the mock microbial community reference sequences.
- TABLE 2
Details of computational methods for identifying and removing contaminants
Method Details Parameters evaluated Filter, negative control ASVs present in the negative control are removed None Filter, abundance ASVs below a relative abundance threshold are removed Abundance threshold, 0.01%, 0.10%, or 1.00% Decontam, frequency ASVs with a correlation with DNA concentration are removed Threshold parameter, 0.1, 0.2, 0.3, 0.4, or 0.5 SourceTracker, scenario 1; experimental source environment is well defined Proportions of ASVs predicted to not be from a defined experimental source are removed Source environments; for case 1, mock community profile + contaminant profile + negative-control profile; for case 2, mock community profile + negative-control profile SourceTracker, scenario 2; experimental source environment is not defined Proportion of ASVs predicted to be from a contaminant source are removed Source environments; for case 1, contaminant profile + negative-control profile; for case 2, negative-control profile
TABLE S1
Estimated number of bacterial cells that yielded the DNA used in 16S rRNA gene PCR reaction for the mock microbial dilution series and DNA concentration measured by nanodrop. Download Table S1, CSV file, 0.00 MB.
Copyright © 2019 Karstens et al.
This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.
TABLE S2
Taxonomic classification of contaminant ASVs with at least 1% abundance in at least 1 sample. Download Table S2, CSV file, 0.01 MB.
Copyright © 2019 Karstens et al.
This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.
TABLE S3
Percentage of contaminant ASVs and mock community ASVs removed using each contaminant removal method. Download Table S3, XLSX file, 0.03 MB.
Copyright © 2019 Karstens et al.
This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.
FIG S1
Recovered microbial community profiles after contaminant removal. Download FIG S1, PDF file, 0.02 MB.
Copyright © 2019 Karstens et al.
This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.
