Page 73 - 南京医科大学自然版

第45卷第4期 南京医科大学学报（自然科学版）
2025年4月 Journal of Nanjing Medical University（Natural Sciences） ·509 ·

·临床研究·

基于焦亡相关基因构建多囊卵巢综合征诊断模型

陈 讯 ，陈雯昕 ，张 文 ，俞池园 ，许波群 1*
1，2
1
1，2
1
南京医科大学第二附属医院妇产科，江苏 南京 210011；南京医科大学附属逸夫医院妇产科，江苏 南京 211100
1 2
［摘 要］ 目的：探索焦亡相关基因在多囊卵巢综合征（polycystic ovary syndrome，PCOS）发病机制中的作用，并构建PCOS的
精确预测模型。方法：利用从基因表达公共数据库（Gene Expression Omnibus，GEO）中获取的 3 个微小 RNA（microRNA，
mRNA）表达谱，分析PCOS患者与正常健康女性之间焦亡相关基因（pyroptosis⁃related gene，PRG）的表达差异。采用广义线性
模型（generalized linear model，GLM）、随机森林（random forest，RF）、支持向量机（support vector machine，SVM）和极限梯度提升
（extreme gradient boosting，XGB）这4种机器学习算法来识别疾病特征基因。采用实时定量聚合酶链反应（real⁃time quantitative
polymerase chain reaction，RT⁃qPCR）法检测10例PCOS患者和10例正常健康女性血浆中特征基因的表达量。结果：建立了基
于PRG的PCOS预测模型和列线图。XGB显示出最高的准确性，决策曲线分析进一步支持了这一结果。一致聚类显示PCOS
病例中有两个亚组，组2比组1表现出更多的免疫浸润。差异表达分析鉴定两个亚型之间的差异表达基因，并对基因进行富集
分析。临床验证结果显示，含 CARD 结构域的凋亡相关斑点样蛋白（apoptosis associated speck like protein containing a CARD，
又称 PYD and CARD domain containing，PYCARD）、黑素瘤缺乏因子 2（absent in melanoma 2，AIM2）、染色质修饰蛋白 4B（chro⁃
matin modifying protein 4B，CHMP4B）和 NOD 样受体蛋白 2（NOD⁃like receptor family pyrin domain containing 2，NLRP2）在 PCOS
患者组中的表达量明显高于正常对照组，差异有统计学意义，验证了基于 PRG 的 PCOS 预测模型的准确性。结论：本研究为
PCOS与焦亡之间的关系提供了初步见解，并提出了PCOS的精确预测模型。
［关键词］ 多囊卵巢综合征；细胞焦亡；免疫浸润；机器学习；生物信息学分析
［中图分类号］ R711.75 ［文献标志码］ A ［文章编号］ 1007⁃4368（2025）04⁃509⁃14
doi：10.7655/NYDXBNSN241034

A diagnostic model of polycystic ovary syndrome based on pyroptosis⁃related genes
1，2 1 1，2 1 1*
CHEN Xun ，CHEN Wenxin ，ZHANG Wen ，YU Chiyuan ，XU Boqun
1 Department of Obstetrics and Gynecology，the Second Affiliated Hospital of Nanjing Medical University，Nanjing
210011；Department of Obstetrics and Gynecology，the Affiliated Sir Run Run Hospital of Nanjing Medical
2
University，Nanjing 211100，China

［Abstract］ Objective：To explore the role of genes related to pyroptosis in the pathogenesis of polycystic ovary syndrome（PCOS）
and construct an accurate prediction model for PCOS. Methods：The differential expression of pyroptosis⁃related genes（PRG）between
PCOS patients and normal healthy women was analyzed by using three microRNA（mRNA）expression profiles obtained from the Gene
Expression Omnibus（GEO）database. Four machine learning algorithms，namely the generalized linear model（GLM），random forest
（RF），support vector machine（SVM），and extreme gradient boosting（XGB），were employed to identify the gene characteristics of
PCOS. Real⁃time quantitative PCR（RT⁃qPCR）method was utilized to detect the expression levels of specific genes in the plasma of 10
PCOS patients and 10 normal healthy women. Results：A predictive model and a nomogram were established based on PRG to
accurately predict PCOS. Among the four machine learning algorithms，the XGB method demonstrated the highest accuracy in
validating the performance of the model，which was further supported by decision curve analysis. Consensus clustering revealed two
distinct subgroups within PCOS cases，with Cluster 2 exhibiting higher level of immune infiltration compared with Cluster 1.
Differential expression analysis was then conducted to identify differentially expressed genes between the two subtypes，followed by
pathway enrichment analysis on the genes. Clinical verification showed that the plasma expression levels of the apoptosis associated

［基金项目］ 国家自然科学基金（81873820）；江苏省卫健委医学科研项目（H2023113）
通信作者（Corresponding author），E⁃mail：boqun_xu@njmu.edu.cn（ORCID：0000⁃0002⁃3360⁃528X）
∗