Artificial Intelligence agents for biological research: a survey

Checked 30 Sept 2026. Open license (CC-BY, CC-BY-SA, CC0, public domain): full text indexed and used in synthesis.

The rapid growth of biological data and experimental complexity has motivated increasing interest in artificial intelligence (AI) systems that extend beyond static prediction toward autonomous reasoning and action. While recent computational models achieve strong predictive performance, they largely operate as passive tools within human-driven research workflows. In contrast, AI agents integrate reasoning, planning, tool invocation, and feedback-driven refinement, enabling more adaptive and interactive forms of biological analysis. This survey provides a systematic synthesis of recent progress in biological AI agents by reviewing over 100 representative studies across clinical analytics, molecular and drug design, multi-omics analysis, and knowledge discovery. We introduce a unified 5D taxonomy that organizes existing work along task domains, system architectures, interaction modes, evaluation strategies, and resource integration. Building on this framework, we analyze common design patterns, highlight emerging capabilities enabled by agentic paradigms, and identify key open challenges, including reliability, privacy, scalability, and standardized evaluation. Collectively, this survey clarifies the conceptual and methodological landscape of biological AI agents and outlines directions toward more robust, transparent, and collaborative agent-based systems for biological research. To serve as a living resource for the community, we curated a GitHub repository that includes resources and benchmark summaries, available at https://github.com/MineSelf2016/biological_agents_survey.

• Independent teams have converged on multi-agent, tool-using LLM architectures as the core design pattern for AI-for-science systems: Coscientist (Boiko et al. 2023) pairs GPT-4 with search, code execution and lab automation, Co-Scientist (Gottweis et al. 2026) is a multi-agent system built on Gemini, and the Virtual Lab (Swanson et al. 2024) has an LLM principal investigator guiding specialist agents. Qi et al. 2026's survey argues this convergence is functional, since multi-agent redundancy offers error-checking that can mitigate hallucination and bias. Trace →
• Role-specialized multi-agent teams are a convergent design pattern across systems such as Co-Scientist and the Virtual Lab, and Qi et al. 2026's survey articulates the rationale: planners, executors, validators and critics provide structured redundancy and error-checking that may mitigate hallucination and bias. Trace →

• AI for science and agentic research