Naseem, S., Miller, M. A., Martinez-Gomez, N. C., Sun, N., Joachimiak, M. P.
An agentic AI system integrating knowledge graphs, metabolic modeling, and optimal experimental design automates microbial growth media optimization.
Microbial cultivation optimization is labor-intensive and inefficient, with traditional one-factor-at-a-time (OFAT) approaches being particularly ineffective for exploring complex, multidimensional nutrient parameter spaces.
The authors developed MicroGrowAgents, an agent-based system that integrates knowledge graphs (KG-Microbe), metabolic modeling, literature mining, genome-guided design, and the MaxPro algorithm for optimal experimental design. 28 specialized agents and 50 skills query structured biological knowledge, mine literature evidence, and perform multi-objective Pareto optimization.
Applied to Methylorubrum extorquens AM1, the system identified a single stable Pareto-optimal medium (MPOB_058) from 70 designed conditions in quadruplicate. The system provides complete provenance tracking with 90.5% literature citation coverage, advancing reproducible, data-driven approaches to microbial cultivation.