From mechanism to application: decrypting light-regulated denitrifying microbiome through geometric deep learning

Abstract

Background: Regulation on denitrifying microbiomes is crucial for sustainable industrial biotechnology and ecological nitrogen cycling. The holisticgenetic profiles of microbiomes can be provided by meta-omics. However, precise decryption and further applications of highly complex microbiomes and corresponding meta-omics datasets remain great challenges. Results: Here, we combined optogenetics and geometric deep learning, following the discover-model-learn-advance (DMLA) cycle, that successfully decrypted light-regulated denitrifying microbiomes and validated the model predictions in the wet lab. Graph neural networks (GNNs) exhibited superior performance in integrating gene expression and subcellular information to identify co-expressed gene panels. Enrichment analysis on critical gene panels successfully predicted the co-expression between reactive oxygen species (ROS) and nitrogen metabolism, as well as the divergent secretion system. Yellow light centralized metabolism fluxes to synthesize protein and increased the extracellular protein concentrations by 83.8%. Contrariwise, blue light decentralized the metabolism fluxes to secrete bioactive substances like secondary metabolites, cofactors and vitamins. The topological network of gene panels guided the scientific discovery of nitrate-superoxide co-regulation and biotechnology development that utilize superoxide to facilitate nitrate removal and realized 99.9% enhancement. Conclusions: Overall, our study showcased the potential of GNNs-empowered optogenetic approaches for regulating denitrification and accelerating the mechanistic discovery of microbiomes for in-depth research and versatile applications.