Gut microbiota ecology: Biodiversity estimated from hybrid neutral-niche model increases with health status and aging

Model selection results (S1 Fig) show that the model that better fits the GM RSA is the 2NB. Specifically, in 4 out of 6 data sets (ELDERMET, Kong, Biagi and Bian) the selected model is the 2NB, while in the other two data sets (Biagi & Schnorr and Odamaki) the performances of the three models are mostly comparable. According to the 2NB model, the GM RSA of each sample is described by two Negative Binomials, that respectively account for “rare” and “abun- dant” species (S2 Fig). Consequently, the two distributions modeling the RSA can be inter- preted as referring to rare and abundant species and the RSA distribution can be rewritten as PRSAðnÞ ¼ a � yrare � NBðbrare; drare; SrareÞ þ ð1 aÞ � yabundant � NBðbabundant; dabundant ; Sabundant Þ ð4Þ In this model, the GM biodiversity is given by the combination of two biodiversity num- bers, θrare and θabundant, that analogously refer to rare and abundant species. As in the case of the biodiversity index θ in the 1NB model (Eq (9)), θrare and θabundant can be derived from the parameters of the two Negative Binomials and are defined as yrare ¼ Nobs ½ð1 brare=drareÞ Srare=brare 1� � GðSrare=brareÞ ð5Þ yabundant ¼ Nobs ½ð1 babundant=dabundantÞ Sabundant =babundant 1� � GðSabundant=babundant Þ ð6Þ In the following subsections we discuss the results obtained from the 2NB model applied to three different biological questions: 1. the relationship between GM biodiversity and aging across several data sets; 2. the discriminative ability of GM biodiversity under important diet and life-style differences

and between healthy and unhealthy aging; 3. the ability of GM biodiversity to predict health status in old age. 1. GM biodiversity increases with aging. We investigated the general relationship between GM biodiversity, i.e. θrare and θabundant, and aging by fitting a non-linear regression model adjusted for sex and total number of reads, as detailed in the Materials and Methods section. For this purpose, we considered healthy control subjects from all data sets except the Italian controls of Schnorr et al., for which age is not available, and the ELDERMET data set samples at times greater than 0, for which young controls are not available. We fitted the model considering samples from all data sets together, while adding the data set of origin as covariate so that to take into account the possible differences related to the individual peculiar- ities of each data set. As illustrated in Fig 1, the model is given by the weighted sum of three basis functions (splines), that represent the behavior of GM biodiversity in three phases of life: spline1 refers to youth years and describes a decrease in biodiversity with age when its coeffi- cient is positive; spline2 indicates a high biodiversity in the middle ages but a lower one in young and elderly people when its coefficient is positive; spline3 refers to old ages and suggests an increase of biodiversity with age when its coefficient is positive. See Materials and Methods for further details on the spline regression model. The splines regression model provides a global result that characterizes the average trend of

biodiversity with age in all data sets (top block of Table 1), and also data set specific results that describe the deviations of each data set from the average behavior (lower blocks of Table 1, Figs 2 and 3). The global trend results show that on average the coefficient related to spline1 is negative, the one of spline2 is negligible and the one associated to spline3 is positive. This result PLOS ONE Gut microbiota biodiversity and healthy aging PLOS ONE | https://doi.org/10.1371/journal.pone.0237207 October 30, 2020 5 / 23