Fig 1.
Sample collection and processing of clinical stool samples for traditional clinical microbiology versus 16S rRNA gene sequencing.
A traditional fecal microbiology test requires collecting a rather large stool sample in a cumbersome process and immediately delivery to the laboratory or clinical practitioner. Specific organisms are cultured from the sample based on the physician’s requests, and processing requires interpretation by extensively trained laboratory personnel. This approach usually focuses on the discovery of culturable pathogens. In contrast, 16S rRNA gene sequencing requires only a fraction of the biological material needed for culture-based techniques (just a swab from toilet paper). In addition, the sample is collected in tube with a buffer that lyses microorganisms and stabilizes DNA, allowing the sample to be mailed at room temperature. Thus, sample collection and delivery are greatly simplified. Sequencing and interpretation can be automated to reduce human labor and error. Finally, this method can detect uncultivable organisms and relative abundances of both pathogenic and commensal organisms.
Fig 2.
Bioinformatics target identification performance metrics.
The 46 preliminary targets identified from literature and available clinical tests are comprised of 15 genera and 31 species. To optimize the bioinformatics pipeline for accurate detection of the maximum number of targets, the following performance metrics were evaluated based on the number of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN) detected in a manually curated amplicon database (described in S1 Doc): specificity = TN / (TN + FP); sensitivity = TP / (TP + FN); precision = TP / (TP + FP); and negative predictive value (NPV) = TN / (TN + FN). After optimization, 28/46 preliminary targets passed our stringent threshold of 90% (red vertical line) for each of the parameters, resulting in the accurate detection of all genera (light blue) except for Pseudoflavonifractor, and 14/31 species (dark blue).
Fig 3.
Reference ranges from a cohort of healthy individuals for 28 clinically relevant species and genera.
Healthy participant stool microbiome data were analyzed to determine the empirical reference ranges for each target. The boxplot displays the relative abundance for each of 897 self-reported healthy individuals, revealing the healthy ranges of abundance for the taxa in the test panel. The healthy distribution is used to define the 99% confidence interval (red line). Boxes indicate the 25th–75th percentile, and the median coverage is indicated by a horizontal line in each box. Even in this healthy cohort, many of the bacteria that are associated with poor health conditions are present at some level. As most taxa are absent in a significant number of individuals most boxes expand to 0%, the healthy lower limit (not shown).
Fig 4.
Experimental validation of the clinical 16S rRNA gene sequencing for pathogens on the screening test panel using verification samples.
Commercially available verification samples (Luminex) containing real or synthetic stool samples positive for at least one control taxon from the target panel were tested using the DNA extraction, amplification and bioinformatics pipeline described in this paper. Of the 35 samples on this panel, 33 yielded 10,000 or more reads. Together, these 33 samples contained the 5 pathogenic taxa in our target list, all of which were accurately identified at a level above the maximum value of the healthy range (red lines). All 33 control samples tested within the healthy range for the remainder of the taxa on our panel (not shown), and thus were considered negative for the pathogenic taxa shown here. Five samples positive for Yersinia, a genus that is not present in our target list, were included as additional negative controls. These samples are visualized for the Escherichia-Shigella genus as they contained DNA for this taxon within the healthy range.
Fig 5.
Human health associations of the 28 targets microorganisms.
All of the 28 taxa on the test have been associated with human health in the gut microbiome. Here we show the associations for 13 specific conditions. 13 of the taxa are associated with health conditions, meaning that these microorganisms have been shown to be elevated in patients suffering from these conditions. The 11 microorganisms that are inversely associated were found to be less abundant in people who have this condition in the scientific literature (S2 Table). 4 taxa are associated with some and inversely associated with other conditions. Interestingly, both elevated and reduced levels of Lactobacillus have been associated with obesity [44–46].