Extended Data Fig. 1: Distribution of daily abundance changes for human and mouse gut microbiota.
Daily abundance changes for an OTU were calculated as the logarithm of the ratio of successive abundances, \(\mu = {\mathrm{log}}(X\left( {t + 1} \right)/X\left( t \right))\), where X(t) is the relative abundance of the OTU on day t. The distributions of abundance changes for the analyzed bacterial communities closely follow the Laplace distribution: \(p\left( \mu \right) = (1/2b){\mathrm{exp}}( - |\mu |/b)\) in a, humans (b = 0.83 ± 0.1, 0.67 ± 0.1, 0.71 ± 0.07, 0.73 ± 0.05, human A, B, M3, F4, respectively; mean ± s.d., across n = 6 equal subsamples of the data, see Methods) and b, mice (b = 0.82 ± 0.1, 0.67 ± 0.03, mice on the LFPP and HFHS diets, respectively; mean ± s.d., across n = 3 individual mice). c, Daily abundance changes of gut microbiota calculated at different taxonomic resolutions. Microbiota abundance changes were calculated as the logarithm of the ratio of successive abundances, \(\mu = {\mathrm{log}}(X\left( {t + 1} \right)/X\left( t \right))\), where X(t) corresponds to the sum of abundances on day t for all OTUs within the same taxonomic group. OTUs were defined at the level of 97% sequence similarity of 16S rRNA. d, Distribution of normalized daily abundance changes in the human gut microbiota. To obtain the distribution, daily abundance changes for individual OTUs were first normalized by their corresponding standard deviations. The distributions of resulting normalized abundance changes were then combined across all OTUs and across humans A, B, M3, and F4. Barplot shows the Akaike Information Criterion (AIC) calculated based on the maximum likelihood estimate (MLE) fits to the data using the Gaussian and Laplace distributions. The bars show the mean AIC calculated across 10,000 bootstrap samples of the abundance changes data, errors represent the standard deviation, and the triple asterisk *** represents p<10-4; across the 104 bootstrap samples, the AIC were always smaller for the Laplace distributions, indicating better fits to the data. In all panels, solid lines show MLE fits to the data.
