AI Agent in Biology Research: A Survey

Cong Qi, Wenbo Wang, Siqi Jiang, Qin Liu, Xun Song, Hanzhang Fang, Zhi Wei

Abstract: The rapid expansion of biological data and the increasing complexity of experimental workflows have created an urgent need for intelligent systems that are capable of autonomous reasoning, planning, and action. Although current computational models demonstrate strong predictive abilities, they remain primarily passive in scientific research that requires human dominance in objective setting, hypothesis construction, and experimental design. In contrast, AI agents introduce a paradigm shift: they combine reasoning, planning, and adaptive decision-making with the ability to invoke external tools, retrieve knowledge, and iteratively refine actions based on feedback. However, research on AI agents in the biological domain is still in its early stages, with existing work scattered across diverse application contexts. This situation necessitates a comprehensive review to synthesize current advances. By consolidating insights from over one hundred recent papers spanning clinical analytics, molecular modeling, multi-omics computation, and exploratory knowledge discovery, we develop a five-dimensional taxonomy that encompasses task domains, architectural paradigms, interaction modes, evaluation strategies, and resource integration. Based on this, the survey identifies practical challenges such as privacy, reliability, and evaluation, indicating a trend toward more collaborative, robust, accessible, and standardized biological AI agent systems.

External IDs:doi:10.36227/techrxiv.176417742.26205344/v1