Microbial Keystone Taxa Identification in Global Wastewater Treatment Plants Based on Deep Learning

Microorganisms play a vital role in maintaining the stability of the activated sludge (AS) ecosystem. Recent studies indicate that certain keystone taxa impact both composition and function, but independent of their abundance. However, an effective framework for identifying these taxa from numerous high-throughput sequencing data is still lacking, particularly without the challenging task of reconstructing the detailed microbial correlation network. We developed a deep learning framework that quantifies microbial impact values across samples, bypassing network reconstruction. This algorithm effectively avoids the tricky issue of differing keystoneness across species in various samples, making it applicable to AS assemblage samples from various environmental and climatic conditions. In this work, we applied this framework to the high-throughput sequencing of the global wastewater treatment sample and identified 61 candidate taxa as the keystones for the wastewater treatment process. We found that the temperature and dissolved oxygen are the primary factors influencing the keystone taxa. Moreover, we found that the increased connectivity of keystone taxa with other members promotes tighter integration within the activated sludge microbial community. In summary, this study introduces an expeditious framework for efficient keystone taxa identification and reveals their role in enhancing microbial community integration in wastewater treatment systems.