*4.7. Gut Flora Analysis*

Participants were provided with a gut testing kit by uBiome (https://en.wikipedia. org/wiki/UBiome, (131 kit numbers from November 2015 (Study 1) and 185 kit numbers June 2017 (Study 2)), which sequenced 16S rRNA of the microbiome from fecal swabs. uBiome provided the sequencing service for registered research groups which provided access to the full suite of 16s RNA raw data sets. Swabs were taken before and after the treatment interventions. Participants were required to swab a fecal sample from toilet paper using the provided swab and place the swab into a provided vial filled with a stabilization liquid. The swab was then swirled in the vial, removed, and discarded. Then, the vial was capped and stored at ambient temperature. Raw data across all taxonomic levels for each participant and time were retrieved.

The raw taxonomic data were downloaded from Ubiome from 122 individual data sets in .json format, converted to .csv using konklone (http://konklone.io/json/, accessed on January to March 2016 for Study 1 and October 2017 to February 2018 for Study 2), and realigned in a single .xlsx database for multivariate analysis. Multivariate analysis was undertaken with PRIMER6+ using multidimensional scaling (MDS) plots to visualize patterns, which were subsequently tested statistically using PERMANOVA. Data were used as untransformed, square root-transformed, and presence/absence data to account for differences in abundance as well as composition. Data were tested using Bray Curtis similarity distance, as these are ecological data of biodiversity with a lot of zeros in the data. Where there were differences in treatment groups, SIMPER analysis was applied to determine which taxonomic group(s) contributed most to this variation. These were then tested in univariate *t*-tests using SPSS (Version 21).

The data were organized for analysis in PRIMER (version 6 with PERMANOVA) and PERMANOVA as per the following Table 11. The level of Genus provided the greatest detail of taxa without losing too much of the quantity, although overall patterns were very similar.


**Table 11.** Data for analysis in PRIMER and PERMANOVA at each taxonomic level.

In the study, 6.5 million gut flora were quantified across close to 200 genera (Table 12 below). The species level of bacteria from the gut is still a field of research in rapid growth in line with the application of genetic tools, so the quantity of over 200 species is an underestimate and is the reason why only 30% of organisms can be identified at this level. Therefore, data that were analyzed at the genus level were selected as a sensitive level of taxonomy but with adequate resolution of diversity accounting for 86% of the total count, while the species level only accounted for 32% of the total count.


**Table 12.** Summary of the taxonomic levels of microorganisms represented in the BioBelly gut flora data from 122 samples from 67 participants.

The data were analyzed at the taxonomic level of genus as this accounted for 86% of the total count, while species level only accounted for 32% of the total count.

Data were prepared as the change in number of normalized counts of genera for each participant. The number of genera were reduced based on the genera that contributed to the change to reduce the noise of the multivariate data or the influence of highly abundant species. Data were visually analyzed in untransformed, square root-transformed, and in presence–absence transformations in multidimensional scaling (MDS) and in principal component analysis (PCA) plots in the PRIMER-E software package (Plymouth Marine Laboratories). Permutational multivariate analysis of variance was used to compare the dissimilarities between groups using the PERMANOVA+ for PRIMER software extension. Data were analyzed in transformed and untransformed formats to determine if shifts were predominantly species abundance or composition changes. Analyses were completed using PCA and PERMANOVA models. Further identification of the genera that contributed most to the differences between groups was undertaken using the SIMPER analysis in the PRIMER-E package.
