Fig. 2: Eigengene modelling using ARIMA augmented with environmental parameters and Fourier terms.
From: Forecasting the dynamics of a complex microbial community using integrated meta-omics
a, The signals S1–17 encapsulate the time-dependent dynamics underlying the microbial community. The scale of the y axis is dimensionless as the eigenvectors. b, The signals S1–17 are explained as ARIMA processes under the influence of the environmental variables. The five blocks of explanatory variables are: Model (ARIMA components), Manual (manually collected environmental variables, directly on the sampling location), Inflow (inflow stream of wastewater in the plant), V1 (first anaerobic tank in the plant) and V2 (second anaerobic tank in the plant). Every circle represents a significant variable according to analysis of variance (Benjamini–Hochberg adjusted P < 0.05) for the corresponding signal among S1–17; the size represents the value of the coefficient, the ring colour its sign and the fill colour the log10(P value). c, The signals are connected by a temporal transfer of information, suggesting a succession of ecological events. The signals with a purple edge are putatively nonlinear, and their relationships have been confirmed with convergent cross-mapping analysis. The dashed lines indicate weak transfer of information, while a full arrow and a hollow one represent an imbalance in information transfer (in favour of the solid arrow).
