南方医科大学学报

南方医科大学学报 ›› 2026, Vol. 46 ›› Issue (3): 559-569.doi: 10.12122/j.issn.1673-4254.2026.03.10

• 基础研究 • 上一篇    

酒精暴露与股骨头坏死的潜在关联：基于机器学习构建诊断模型

陶红成1(), 梁富凯1, 黄文波1, 范思奇2(), 曾平2,3()   

  1. 1.广西中医药大学，广西 南宁 530200
    2.广西中医药大学第一附属医院，广西 南宁 530023
    3.广西中医药大学广西中医基础研究重点实验室，广西 南宁 530200
  • 收稿日期:2025-08-09 出版日期:2026-03-20 发布日期:2026-03-26
  • 通讯作者: 范思奇,曾平 E-mail:taohongcheng2021@stu.gxtcmu.edu.cn;fansiqi932828@163.com;zengp@gxtcmu.edu.cn
  • 作者简介:陶红成，在读博士研究生，E-mail: taohongcheng2021@stu.gxtcmu.edu.cn
  • 基金资助:
    国家自然科学基金地区科学基金项目(82160913);国家自然科学基金地区科学基金项目(81960876);广西壮族自治区科学技术厅广西青年科学基金项目（中医药壮瑶医药专项-广西中医药大学）(2025GXNSFBA069217);广西中医骨伤临床医学研究中心(GXZYYXYJZX-2025-01)

Analysis of potential association of alcohol exposure with femoral head osteonecrosis and construction of a diagnostic model using machine learning

Hongcheng TAO1(), Fukai LIANG1, Wenbo HUANG1, Siqi FAN2(), Ping ZENG2,3()   

  1. 1.Guangxi University of Chinese Medicine, Nanning 530200, China
    2.First Affiliated Hospital of Guangxi University of Chinese Medicine, Nanning 530023, China
    3.Guangxi Key Laboratory of Chinese Medicine Basic Research, Guangxi University of Chinese Medicine, Nanning 530200, China
  • Received:2025-08-09 Online:2026-03-20 Published:2026-03-26
  • Contact: Siqi FAN, Ping ZENG E-mail:taohongcheng2021@stu.gxtcmu.edu.cn;fansiqi932828@163.com;zengp@gxtcmu.edu.cn

RichHTML

22

PDF

74

摘要：

目的 运用113种机器学习方法并基于酒精暴露相关基因构建并验证股骨头坏死诊断模型。 方法 从GEO数据库中获取酒精暴露和股骨头坏死相关的转录组学数据，进行系统分析以开发和验证股骨头坏死诊断模型。合并数据集并去除批次效应后，进一步进行差异基因分析，确定酒精暴露和股骨头坏死的差异表达基因（DEGs）并取交集。对取得的交集DEGs 进行功能富集分析。采用单样本基因集富集分析（ssGSEA）对免疫浸润情况进行量化分析。在训练集上探索12种机器学习算法的113种组合，并采用10折交叉验证，构建ONFH的诊断模型，在测试集上进行验证。将MC3T3-E1细胞分为对照组和酒精组，通过qRT-PCR检测两组细胞中构建模型的差异基因表达，以验证构建模型的可靠性。利用“Enrichr”平台来识别ONFH的潜在药物。 结果 总共鉴定出21个酒精暴露和股骨头坏死均密切相关的交集DEGs，基于富集分析，这些基因亦有参与免疫过程。通过免疫浸润分析显示，与健康对照组相比，酒精暴露和股骨头坏死患者体内免疫细胞浸润情况存在明显差异。接着，构建了由8个基因（SOAT1、GMCL1、GMPR、CISD2、ST3GAL6、AHSP、UBL3、PTPN12）组成的可靠的诊断模型，并进一步通过qRT-PCR验证发现，这8个基因在对照组和酒精组MC3T3-E1细胞中表达存在显著差异（P<0.05）。最后，预测了10个可能靶向治疗股骨头坏死的潜在药物。 结论 通过整合生物信息学分析和机器学习方法，并基于酒精暴露和股骨头坏死共有的DEGs，构建出可用于诊断股骨头坏死的可靠的模型。

关键词: 酒精暴露, 股骨头坏死, 生物信息学分析, 机器学习, 诊断模型

Abstract:

Objective To construct and validate a diagnostic model for osteonecrosis of the femoral head (ONFH) based on alcohol exposure-related genes using machine learning methods. Methods The transcriptomic data related to alcohol exposure and ONFH were obtained from the GEO database for construction of a diagnostic model for ONFH. The differentially expressed genes (DEGs) of alcohol exposure and ONFH were identified, and functional enrichment analysis of the intersecting genes was performed. Single sample gene set enrichment analysis (ssGSEA) was used to quantify immune infiltration. A total of 113 combinations of 12 machine learning algorithms were tested on the training set, and 10-fold cross-validation was used to construct the diagnostic model of ONFH, which was validated on the test set. The expressions of the DEGs in the model were detected by qRT-PCR in alcohol-treated MC3T3-E1 cells to verify the reliability of the constructed model. The Enrichr platform was used to identify potential drugs for ONFH. Results Twenty-one intersecting DEGs closely related to alcohol exposure and ONFH were identified, which were also involved in immune processes. Immune infiltration analysis showed that the patients with alcohol exposure and ONFH had significant differences in immune cell infiltration compared with the healthy controls. The diagnostic model was constructed based on 8 genes (SOAT1, GMCL1, GMPR, CISD2, ST3GAL6, AHSP, UBL3, and PTPN12), which showed significant differential expressions in alcohol-treated MC3T3-E1 cells. Ten potential drugs for ONFH treatment were predicted. Conclusion By integrating bioinformatics analysis and machine learning methods, a reliable model for diagnosing ONFH has been successfully constructed based on the DEGs shared by alcohol exposure and ONFH.

Key words: alcohol exposure, osteonecrosis of the femoral head, bioinformatics analysis, machine learning, diagnostic model