遗传

• 研究报告 •    

基于关键词MCC共现网络的GWAS文献计量学研究

舒安镕1,李鸿1,袁哲明1,2   

  1. 1.湖南农业大学农业大数据分析与决策工程技术研究中心,长沙 410128

    2.岳麓山实验室,长沙 410125
  • 发布日期:2026-11-05
  • 基金资助:
    湖南省科技创新计划项目(编号:2023NK2001)资助

Bibliometric analysis of genome-wide association studies based on keyword co-occurrence networks using the Matthews correlation coefficient

Anrong Shu1, Hong Li1, Zheming Yuan1,2   

  1. 1. Hunan Engineering and Technology Research Center for Agricultural Big Data Analysis and Decision-Making, Hunan Agricultural University, Changsha 410128, China

    2. Yuelushan Laboratory, Changsha 410125, China
  • Online:2026-11-05

摘要: 全基因组关联分析(genome-wide association studies,GWAS)自2007年规模化应用以来,已积累大量文献,系统梳理其演进脉络具有重要意义。然而,传统关键词共现分析易受高频词稀释效应影响,细分主题在网络中的辨识度不足。为此,本研究引入马修斯相关系数(Matthews correlation coefficient,MCC)量化关键词间的二元关联强度;基于2007~2025年Web of Science核心合集收录的56,375篇文献,结合PELT断点检测、CiteSpace突现词分析与共被引分析,解析GWAS领域的主题演进与知识结构。结果表明:(1)PELT算法识别出2011年、2021年两个断点,将领域划分为萌芽期、扩张期与快速增长期,其中快速增长期发文量占41.3%;(2)基于MCC构建的无加权网络模块度更高(Q=0.510,传统余弦相似度共现网络Q=0.349),加权与无加权网络的Q值相对偏差为14.66%(余弦相似度网络为41.42%),稳定性更好;(3)主题演进呈现由早期疾病易感位点识别,经多组学功能注释与精细定位,向以孟德尔随机化(Mendelian randomization,MR)为核心的因果推断转变的轨迹;(4)共被引网络识别出4个核心知识聚类:MR方法学与因果推断、GWAS分析工具与数据资源、混合线性模型与动植物关联作图,以及早期GWAS与遗传流行病学。综上,基于MCC的关键词共现分析可减弱高频词干扰,得到更清晰的主题模块,可为GWAS领域的主题演进梳理与知识图谱构建提供方法学支撑。突现词分析提示,未来前沿可能集中于多组学整合的因果推断、单细胞分辨率的遗传效应解析,以及因果变异的精细定位与功能解释。

关键词: 全基因组关联分析, 文献计量学, 马修斯相关系数, 关键词共现网络, 孟德尔随机化, 因果推断

Abstract: Since their large-scale application in 2007, genome-wide association studies (GWAS) have accumulated a substantial body of literature, and systematically tracing their evolution is therefore of considerable importance. Conventional keyword co-occurrence analysis, however, is susceptible to dilution by high-frequency terms, resulting in poor distinguishability of subthemes in the network. To address this limitation, this study introduces the Matthews correlation coefficient (MCC) to quantify the binary association strength between keywords. Based on 56,375 publications indexed in the Web of Science Core Collection from 2007 to 2025, and combining PELT changepoint detection, CiteSpace burst analysis, and co-citation analysis, this study examines the thematic evolution and knowledge structure of the GWAS field. The results show that: (1) the PELT algorithm identified 2011 and 2021 as two changepoints, dividing the field into an emerging stage, an expansion stage, and a rapid-growth stage, with the rapid-growth stage accounting for 41.3% of all publications; (2) the unweighted MCC network exhibited higher modularity (Q=0.510 vs 0.349 for the conventional cosine-similarity co-occurrence network), with a relative deviation of 14.66% between the weighted and unweighted Q values (41.42% for the cosine-similarity network), indicating greater network stability; (3) the thematic evolution followed a trajectory from the early identification of disease-susceptibility loci, through multi-omics functional annotation and fine-mapping, toward causal inference centered on Mendelian randomization (MR); and (4) the co-citation network identified four core knowledge clusters, namely MR methodology and causal inference, GWAS analytical tools and data resources, mixed linear models and association mapping in plants and animals, and early GWAS and genetic epidemiology. Overall, MCC-based keyword co-occurrence analysis helps mitigate interference from high-frequency terms and yields clearer thematic modules, thereby providing methodological support for tracing thematic evolution and constructing knowledge maps in the GWAS field. Burst analysis suggests that future frontiers may focus on multi-omics-integrated causal inference, the dissection of genetic effects at single-cell resolution, and the fine-mapping and functional interpretation of causal variants.

Key words: genome-wide association studies, bibliometrics, Matthews correlation coefficient, keyword co-occurrence network, Mendelian randomization, causal inference