Artificial intelligence-assisted phenotypic monitoring and diagnostic of wastewater microbiome management towards sustainability.

Wastewater resource recovery facilities (WRRFs) rely on functional microbiome to remove pollutants and safeguard water sustainability towards United Nation's Sustainable Development Goals (SDGs). However, conventional DNA-based and operator experience-based monitoring approaches often fail to provide early warning of functional instability, leading to sudden WRRF performance loss and increased risk of regulatory noncompliance, primarily due to the lack of precise, functionality-driven monitoring systems. Here, we develop an artificial intelligence(AI)-assisted single-cell Raman spectroscopy (SCRS) platform and assemble a large Ramanome database (46,892 single cells across 12 WRRF configurations) for high-resolution phenotypic monitoring and diagnostics of wastewater microbiomes for a reliable and sustainable WRRF system. Our results demonstrate that Ramanome-defined operational phenotypic units (OPUs) and their associated phenotypic metrics (e.g., phenotypic diversity, network structures) can serve as robust phenotypic signals for accurate WRRF performance monitoring, complementing the conventional taxonomy-based signals (e.g., 16S rRNA). Moreover, our explainable AI models accurately identify key functional phenotypes and OPUs (accuracy 0.95-1.00) and enable quantitative diagnosis of WRRF health across regulatory compliance levels (accuracy 0.55-1.00). Overall, this function-driven SCRS-AI framework establishes a robust and scalable platform for predictive microbiome monitoring, diagnostics, and management, advancing sustainable wastewater treatment innovations in alignment with the SDGs.