Ml-guided microbial consortia for selective selenium reduction
An AI-driven platform for in-situ selenium bioremediation that prevents arsenic mobilization by using a microbial consortium. This solution integrates a Se/As knowledge graph and neural network predictions to deliver deployment-ready consortia tailored to specific geochemical environments.
Overview
This innovative platform leverages machine learning to design microbial consortia capable of selectively reducing selenium in groundwater without mobilizing arsenic. Traditional bioremediation methods often result in unintended arsenic release due to non-selective biostimulation. Our solution replaces this with precision-engineered consortia that are site-specific. By integrating a selenium/arsenic knowledge graph and a graph neural network, we predict effective microbial combinations, optimizing for desired geochemical interactions.
Technical specifications
Key features:
- Uses a heterogeneous graph neural network for predictive modeling of microbial interactions
- Curates extensive genomic, geochemical, and phenotypic data for accurate model training
- Involves multi-task models predicting selenium reduction efficiency and arsenic mobilization risk
- Includes closed-loop validation through flow-through aquifer column testing
Applications:
- Groundwater bioremediation at sites with selenium and arsenic contamination
- Environmental remediation efforts requiring precision and selectivity
Technology readiness level
The platform is currently at TRL 3, with ongoing developments aimed at advancing to TRL 5 through further validation and field testing. Initial phases involve site characterization and ML-guided consortium design, progressing to column validation and final deliverables preparation for deployment.
About Berkeley Lab
Berkeley Lab is a multi‑program U.S. Department of Energy national laboratory in Berkeley, operated by the University of California. Industry engages through open scientific user facilities, national high‑performance computing, co‑located nanoscience and materials characterization labs, and testbeds, with options for nonproprietary and, where permitted, proprietary work. Proximity to UC Berkeley and the Bay Area innovation ecosystem enables joint appointments, sponsored research, and talent pipelines via CRADAs and Strategic Partnership Projects. Research is funded primarily by the DOE Office of Science, with additional competitive support from other federal agencies and industry. A dedicated technology transfer office manages IP, licensing, and collaboration agreements.
delivered to your inbox.