Department of Pediatrics, University of California San Diego, La Jolla, California, USADepartment of Computer Science and Engineering, University of California San Diego, La Jolla, California, USACenter for Microbiome Innovation, University of California San Diego, San Diego, California, USA
A small number of bacteria showed high levels of growth in the storage studies. (A to D) The log2 fold change of an OTU relative to time zero within the Mayo Clinic fecal stability study and the preservation study by Song et al. (1). A minimum read threshold of 10 reads per sample was used, and OTUs present below this threshold in both samples are not shown. (A and B) Mayo Clinic fecal stability study (2) at day 1 (A) and day 4 (B). (C and D) Results of fecal preservation study by Song et al. (1) at day 7 (C) and day 14 (D). The numbers (and percentages of total bacteria) showing at least 4-fold change in the room-temperature storage compared to day 0 were 11 (1.6%), 21 (3.1%), 49 (7.1%), and 144 (19%) for days 1, 2, 7, and 14, respectively. (E) A topology plot of the maximal (max.) fold change in the two storage studies whose results are shown in panels A to D (x axis) and minimal (min.) fold change in the AGP fecal samples compared to data reported in references 6 and 7 and PGP (unpublished), studies in which samples were immediately frozen (y axis). Red circles denote 14 bacterial samples selected as potentially blooming (>2-fold change in both axes or >50-fold change in the storage studies). Six bacterial samples showing >2-fold change in AGP compared to results of all studies using fresh-frozen samples that were not present in the storage studies are also added to the potential blooming list (Table S1).
Effect of bloom filtering on American Gut data. (A and B) PCoA of Bray-Curtis distances from a random subset of 200 American Gut Project samples (AGP [unpublished]; colored pink) compared to 3 studies containing fresh-frozen fecal samples: Personal Genome Project (PGP [unpublished]; colored green); whole-grain feces (EWF ; colored orange); and UK Twins (; colored purple), respectively. The PCoA data shown represent results obtained before (A) and after (B) applying the filter for blooms to all samples. The size of a sphere is scaled by the amount of candidate bloom bacteria in a sample prior to filtering. (C and D) Mean taxonomy distribution for the same studies before (C) and after (D) filtering for blooms. (E and F) The well-known effect of age on alpha diversity and how the effect is observed only after the removal of bloom reads. The Kruskal-Wallis test statistic (E) and corresponding –log(P value) (F) are shown for different numbers of bacteria used for the filtering before applying the test. A value of 0 on the x axis indicates no filtering. The x axis is ordered by decreasing severity score of the bloom where bloom 1 represents greater severity than bloom 2, and each point on the x axis includes the prior blooms (e.g., position 5 includes bloom sOTUs 1 through 5).
Effect of number of candidate blooming bacteria used for filtering. Data represent mean Bray-Curtis distances between different experiments following removal of 0 to 20 candidate blooming bacteria (x axis). A total of 1,000 random pairs of samples were chosen for each level of filtering. Distances shown correspond to comparisons of American Gut to UK Twins (red), American Gut to PGP (blue), American Gut to whole-grain feces (yellow), UK Twins to PGP (dashed green), UK Twins to whole-grain feces (dashed black), and whole-grain feces to PGP (dashed light blue). Download FIG S1, PDF file, 0.02 MB.