基于本体学增强的大型语言模型构建罕见疾病知识图的自动和端到端系统:发展研究研究
Lang Cao1, Jimeng Sun1, Adam Cross2
1Department of Computer Science, University of Illinois Urbana-Champaign, Urbana, IL, United States.
JMIR medical informatics
|December 18, 2024
概括
自动罕见病挖掘 (AutoRD) 通过将大型语言模型与本体学集成,从医学文本中改进罕见病识别. 该系统增强了用于研究和临床试验的患者识别,克服了当前方法的局限性.
科学领域:
- 生物医学信息学 生物医学信息学
- 自然语言处理自然语言处理.
- 罕见疾病研究 罕见疾病研究
背景情况:
- 罕见疾病影响全球数百万人,但缺乏特定的诊断代码 (ICD-9,ICD-10),阻碍了患者的识别和研究.
- 目前的大型语言模型 (LLM) 缺乏专门的医学知识,以有效地管理和提取罕见疾病数据.
- 鉴定罕见疾病的挑战使临床试验招聘和研究工作复杂化.
研究的目的:
- 开发一种自动化系统 (AutoRD) 来从医学文本中提取罕见疾病信息.
- 整合本体学和结构化知识,以优越的罕见病实体和关系提取.
- 超越常见的LLMs和传统方法在罕见疾病采矿中的性能.
主要方法:
- 开发了AutoRD作为一个管道:数据预处理,实体/关系提取,实体校准和知识图构建.
- 利用了GPT-4和来自人类表型和Orphanet本体学的医学知识图表.
- 采用连锁思维推理和快速工程来提高提取效率.
主要成果:
- 在RareDis2023数据集上,AutoRD实现了整体实体提取F1得分为56.1%,关系提取F1得分为38.6%.
- 在整体提取性能方面,与基线LLM相比,表现出14.4%的改善.
- 在罕见疾病实体提取方面获得了83.5%的高F1得分,展示了精度和回忆.
结论:
- AutoRD有效地提取罕见疾病信息并构建知识图,解决LLM在这个领域的局限性.
- 实体学增强的LLM显著改善了罕见疾病的识别和与临床特征的联系.
- 该系统有助于更好地识别患者进行研究和临床试验招聘,促进包容性.
相关概念视频
Classification of Illness
7.3K
The meaning of illness is individualized to each person who experiences an alteration in health. In contrast, disease is a medical term indicating a pathological change in the structure and function of the body or mind. It is a condition that has specific symptoms and boundaries.
An illness is a response to a disease in which the person's level of functioning is changed compared with a previous level. The general classification of illness includes acute and chronic.
Acute illness is severe...
An illness is a response to a disease in which the person's level of functioning is changed compared with a previous level. The general classification of illness includes acute and chronic.
Acute illness is severe...
7.3K
Lysosomal Hydrolases
3.8K
Lysosomes are the site for the degradation of macromolecules and biological polymers released during membrane trafficking events such as secretory, endocytic, autophagic, and phagocytic pathways. The membrane-enclosed area of the lysosome, called the lumen, contains hydrolytic enzymes active in an acidic environment. These acid hydrolases are functional at a pH between 4.5 and 5 and are involved in cellular processes such as cell signaling, energy metabolism, restoration of the plasma membrane,...
3.8K
Genomics
35.8K
Genomics is the science of genomes: it is the study of all the genetic material of an organism. In humans, the genome consists of information carried in 23 pairs of chromosomes in the nucleus, as well as mitochondrial DNA. In genomics, both coding and non-coding DNA is sequenced and analyzed. Genomics allows a better understanding of all living things, their evolution, and their diversity. It has a myriad of uses: for example, to build phylogenetic trees, to improve productivity and...
35.8K
Model Approaches for Pharmacokinetic Data: Distributed Parameter Models
56
Pharmacokinetic models are mathematical constructs that represent and predict the time course of drug concentrations in the body, providing meaningful pharmacokinetic parameters. These models are categorized into compartment, physiological, and distributed parameter models.
The distributed parameter models are specifically designed to account for variations and differences in some drug classes. This model is particularly useful for assessing regional concentrations of anticancer or...
The distributed parameter models are specifically designed to account for variations and differences in some drug classes. This model is particularly useful for assessing regional concentrations of anticancer or...
56
Genome-wide Association Studies-GWAS
12.4K
Genome-wide association studies or GWAS are used to identify whether common SNPs are associated with certain diseases. Suppose specific SNPs are more frequently observed in individuals with a particular disease than those without the disease. In that case, those SNPs are said to be associated with the disease. Chi-square analysis is performed to check the probability of the allele likely to be associated with the disease.
GWAS does not require the identification of the target gene involved in...
GWAS does not require the identification of the target gene involved in...
12.4K
Genetic Lingo
100.9K
Overview
100.9K


