Universal Dynamics of Microbial Communities in Full-Scale Textile Wastewater Treatment Plants and System Prediction by Machine Learning.

The performance of full-scale biological wastewater treatment plants (WWTPs) depends on the operational and environmental conditions of treatment systems. However, we do not know how much these conditions affect microbial community structures and dynamics across systems over time and predictability of the treatment performance. For over a year, the microbial communities of four full-scale WWTPs processing textile wastewater were monitored. During temporal succession, the environmental conditions and system treatment performance were the main drivers, which explained up to 51% of community variations within and between all plants based on the multiple regression models. We identified the universality of community dynamics in all systems using the dissimilarity-overlap curve method, with the significant negative slopes suggesting that the communities containing the same taxa from different plants over time exhibited a similar composition dynamic. The Hubbell neutral theory and the covariance neutrality test indicated that all systems had a dominant niche-based assembly mechanism, supporting that the communities had a similar composition dynamic. Phylogenetically diverse biomarkers for the system conditions and treatment performance were identified by machine learning. Most of the biomarkers (83%) were classified as generalist taxa, and the phylogenetically related biomarkers responded similarly to the system conditions. Many biomarkers for treatment performance perform functions that are crucial for wastewater treatment processes (e.g., carbon and nutrient removal). This study clarifies the relationships between community composition and environmental conditions in full-scale WWTPs over time.