A new multi-agent system automates gene discovery and experimental design, bringing expert-level analytical workflows to crop breeding research.

Gene discovery in crops is a painstaking process. Researchers must sift through mountains of literature, cross-reference genomic data, and design experiments that test increasingly complex hypotheses. Each step demands deep domain expertise—and even then, the path from question to insight is rarely straight.

Now, a research team has introduced GeneScientist, an AI-powered multi-agent system built to automate this entire pipeline.

How It Works

GeneScientist integrates domain-specific tools, predictive models, and hypothesis-guided reasoning into a single coordinated framework. At its core are two specialized agents:

• Planner — decomposes user objectives into executable sub-goals
• Executor — parameterizes and runs the corresponding tool calls

These agents collaborate iteratively, forming a closed-loop reasoning cycle that links task planning, evidence retrieval, predictive inference, and hypothesis refinement until the objective is fulfilled.

To manage complex analytical workflows, the team added an orchestration layer based on a directed acyclic graph, which structures task dependencies and ensures each step executes in the right order.

Four modules