CANCELLED - Harness the power of machine learning and omics to identify microbial functional composition across diverse environments

Background/Question/Methods

Microbial communities regulate soil organic matter (SOM) decomposition processes but are still underrepresented in land carbon cycling models. The key challenges in this effort are the lack of a proper understanding of the spatiotemporal distribution of microbial functional diversity and an explicit parameterization approach to integrate complex microbial taxonomic and functional information into models. To address these challenges, we integrated observed omics data in the US to identify key microbial functions involved in SOM decomposition. We collected metagenomics, metatranscriptomics, and metaproteomics data from online omics databases, including JGI-IMG and MG-RAST, to identify relative abundances of enzymes involved in SOM decomposition processes. We clustered omics-informed enzyme functional information into enzyme functional classes that reduced functional redundancies of omics data and directly linked it with represented SOM decomposition processes in the land surface models. We integrated clustered enzyme functional information with a high-resolution US soil organic carbon dataset and environmental factors data to develop an AI-based prediction model. Applying this model allowed us to identify the spatial distribution of microbial functions for SOM decomposition across the diverse environments in the US.

Results/Conclusions

Our study indicates that general microbial enzyme functions involved in SOM decomposition can be identified across diverse environments in the US. However, the relative abundance of microbial functional composition strongly depends on soil carbon content, soil stoichiometry, and eco-regions and can be predicted by integrating the environmental information with ML-based enzyme functional classes prediction. Our effort develops a microbial functional dataset that can be used to parameterize and benchmark microbial-mediated SOM decomposition in land carbon cycling models. The developed ML model can be used to inform different life-history strategies of microbial communities across diverse environments.