Self-evolving AI agents for protein discovery and directed evolution

2026-03-28

AI Summary

This work addresses the limitations of current protein science discovery, which is hindered by the need for manual coordination between information and algorithms, a task at which general-purpose AI agents typically underperform. To overcome this, the authors propose VenusFactory2, a self-evolving multi-agent framework that enables end-to-end automation of protein discovery and optimization through dynamically synthesized workflows driven solely by natural language prompts. Departing from conventional static tool-calling paradigms, VenusFactory2 introduces a novel mechanism for dynamic workflow composition and natural language–guided experimental design. Evaluated on the VenusAgentEval benchmark, the framework substantially outperforms existing agents and achieves, for the first time, fully automated directed evolution and functional optimization of proteins.

Abstract

Protein scientific discovery is bottlenecked by the manual orchestration of information and algorithms, while general agents are insufficient in complex domain projects. VenusFactory2 provides an autonomous framework that shifts from static tool usage to dynamic workflow synthesis via a self-evolving multi-agent infrastructure to address protein-related demands. It outperforms a set of well-known agents on the VenusAgentEval benchmark, and autonomously organizes the discovery and optimization of proteins from a single natural language prompt.

Problem

• protein discovery
• directed evolution
• AI agents
• scientific discovery
• workflow orchestration

Innovation

• self-evolving multi-agent
• dynamic workflow synthesis
• protein discovery
• directed evolution
• autonomous AI agent