大型语言模型智能体用于基于证据的遗传疾病严重程度分类
Large Language Model Agents for Evidence Based Genetic Disease Severity Classification
- UNSW Sydney（新南威尔士大学）
- Strands（23Strands）
- University of Pittsburgh（匹兹堡大学）
- Shenzhen Institute of Advanced Technology（深圳先进技术研究院）
- Chulalongkorn University（朱拉隆功大学）
- New South Wales Health Pathology（新南威尔士州卫生病理局）
- Prince of Wales Hospital（威尔士亲王医院）
- Neuroscience Research Australia (NeuRA)（澳大利亚神经科学研究所）
- University of Pittsburgh Medical Center（匹兹堡大学医学中心）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
本研究提出结合ReAct与RAG的AI智能体，依据ACMG和ACOG指南自动分类遗传疾病严重程度，在表型层面达93.55%准确率，并支持标准化检测组合设计。
AI中文摘要：
遗传疾病的严重程度分类具有主观性和劳动密集型特点，在基因组筛查中造成瓶颈，而商业检测组合在规模和重叠性上差异很大。我们开发了一个自主AI智能体，将推理与行动（ReAct）与检索增强生成（RAG）相结合，对10,211个人类表型本体术语进行分类。它使用美国医学遗传学学院（ACMG）认可的严重程度指南和美国妇产科学会（ACOG）的生活质量标准来检索PubMed文献，生成可解释的推理链，并独立验证声明。在表型层面，使用专家策划的队列，该智能体达到了93.55%的准确率（MCC 0.9237），其中82.6%至91.4%的声明得到直接证据或有效推理的支持。基因层面的严重程度在8,738对中聚合，识别出3,283对具有严重或极重表现型的常染色体隐性遗传对。外部验证显示与Mackenzie's Mission基因列表的一致性为95.2%。该系统通过提供可靠、自动化的分类并辅以直接证据，实现了标准化检测组合设计。
英文摘要：
Disease severity classification for genetic conditions is subjective and labor-intensive, creating bottlenecks in genomic screening, where commercial panels vary widely in size and overlap. We developed an autonomous AI agent integrating Reasoning and Acting (ReAct) with Retrieval-Augmented Generation (RAG) to classify 10,211 Human Phenotype Ontology terms. It uses American College of Medical Genetics (ACMG)-endorsed severity guidelines and American College of Obstetricians and Gynecologists (ACOG) quality-of-life criteria to retrieve PubMed literature, generate interpretable reasoning chains, and independently verify claims. At the phenotype level, using expert-curated cohorts, the agent achieved 93.55% accuracy (MCC 0.9237) with 82.6% to 91.4% of claims supported by direct evidence or valid inferences. Gene-level severity was aggregated across 8,738 pairs, identifying 3,283 autosomal recessive pairs with severe or profound presentations. External validation showed 95.2% concordance with Mackenzie's Mission gene list. This system enables standardized panel design by providing reliable, automated classification supported by direct evidence.
