麻豆精品视频在线_韩国三级av_亚洲一区中文_涩涩的网址_91极品尤物_日本猛少妇色xxxxx_在线性视频_亚洲第一综合_大牛影视剧免费播放在线_被触手肉干高h潮文

MAESD: A Unified Multi-Agent Evolutionary Framework for Protein Sequence Design

• 1、School of Artificial Intelligence and Computer Science(School of Software),Jiangnan University，Wu xi，214122

Abstract：Computational protein design traditionally requires expert-level biological inputs to specify structural and functional constraints, creating significant barriers for non-specialists. To bridge this gap, we leverage recent advances in large language models (LLMs) that demonstrate their capability to perform complex reasoning in specialized domains by leveraging knowledge bases to produce expert-grade outputs. In this paper, we present MAESD(Multi-Agent Evolutionary framework for protein Sequence Design), a framework for protein design via natural language instructions. The system first employs large language models to interpret biological requirements and extract actionable design constraints, then establishes an evolutionary optimization loop , and finally iteratively refines sequences through generation and protein screening. Our framework establishes a dynamic mapping between natural language specifications and executable protein design constraints via collaborative multi-agent reasoning, with evolutionary optimization ensuring biological plausibility at each iteration. By integrating natural language understanding with evolutionary computation, MAESD significantly lowers the expertise barrier for protein design while achieving performance comparable to expert-crafted solutions.

Keywords： Protein Design, Multi-Agent Systems, Evolutionary Loop