[1]李 莎,田 原,谢晋烨,等.基于临床实验室常规检验数据使用随机森林分类器构建多发性骨髓瘤预测模型[J].现代检验医学杂志,2026,41(02):46-49+69.[doi:10.3969/j.issn.1671-7414.2026.02.008]
　LI Sha,TIAN Yuan,XIE Jinye,et al.Development of a Multiple Myeloma Predictive Model Using Routine Clinical Laboratory Data Based on a Random Forest Classifier[J].Journal of Modern Laboratory Medicine,2026,41(02):46-49+69.[doi:10.3969/j.issn.1671-7414.2026.02.008]
点击复制

基于临床实验室常规检验数据使用随机森林分类器构建多发性骨髓瘤预测模型()

《现代检验医学杂志》[ISSN:/CN:]

卷:
第41卷
期数:
2026年02期
页码:
46-49+69
栏目:
出版日期:
2026-03-15

文章信息/Info

Title:
Development of a Multiple Myeloma Predictive Model Using Routine Clinical Laboratory Data Based on a Random Forest Classifier
文章编号:
1671-7414（2026）02-046-05
作者:
李 莎，田 原，谢晋烨，欧阳能良，吴杏儿，王 娟，李任洲，黄云秀，李 曼，李丽敏
中山市人民医院，广东中山 528400
Author(s):
LI Sha，TIAN Yuan，XIE Jinye，OUYANG Nengliang，WU Xing’er，WANG Juan，LI Renzhou，HUANG Yunxiu，LI Man，LI Limin
Zhongshan City People’s Hospital, Guangdong Zhongshan 528400, China
关键词:
随机森林分类器; 多发性骨髓瘤; 常规检验指标; 预测模型
分类号:
R733.3；R730.43
DOI:
10.3969/j.issn.1671-7414.2026.02.008
文献标志码:
A
摘要:
目的 构建基于常规检验数据的多发性骨髓瘤(MM)预测模型。方法纳入2022~2024年中山市人民医院确诊的224例MM患者作为研究组,纳入同期200例健康体检者作为对照组。收集两组共74项常规检验指标,先通过科尔莫戈罗夫-斯米尔诺夫(K-S)检验判断数据正态性,再采用独立样本t检验或秩和检验比较组间差异。对差异具有统计学意义的指标绘制受试者工作特征(ROC)曲线,筛选曲线下面积(AUC)值>0.7的指标作为有效特征变量,采用随机森林分类器构建预测模型,并通过5折交叉验证对模型性能进行评估。结果共发现两组间差异具有统计学意义的检验指标38项(均P<0.05)。其中,尿液总κ轻链(uκ)和尿液总λ轻链(uλ)的AUC值均为1.0,具有最强的预测能力。5核苷酸酶(5-NT)、类风湿因子(RF)、红细胞分布宽度-变异系数(RDW-CV)、凝血酶时间(TT)升高以及红细胞比容(HCT)、血红蛋白(Hb)降低时的AUC值均超过0.9,具有较强的预测能力。β2微球蛋白(β2-MG)、红细胞分布宽度-标准差(RDW-SD)、胱抑素C(CysC)、C反应蛋白(CRP)、腺苷脱氨酶(ADA)、甘油三酯(TG)、肌酐(Cr)、碱性磷酸酶(ALP)、尿素(Urea)、抗链球菌溶血素(ASO)升高以及白蛋白(Alb)、红细胞(RBC)、估算肾小球滤过率(eGFR)、血小板(PLT)、活化部分凝血活酶时间(APTT)、肌酸激酶同工酶(CK-MB)、平均红细胞血红蛋白浓度(MCHC)、高密度脂蛋白胆固醇(HDL-C)、总蛋白(TP)降低时的AUC值介于0.7~0.9,具有中等的预测能力。其余差异具有统计学意义的指标的AUC值介于0.5~0.7,具有较弱的预测能力。基于AUC值>0.7的指标进行了重要性排序。构建的随机森林模型在预测MM患者与健康个体时,准确度达到100%,且精确率、召回率及F1分数均为1.00。该模型已保存为.pkl文件,完整的Python代码也已上传至GitHub平台。结论基于常规检验指标,使用随机森林分类器能够有效构建MM预测模型,不仅提高了MM的早期识别效率,而且显著增强了常规检验数据的利用价值。
Abstract:
Objective To develop a predictive model for multiple myeloma (MM) based on routine laboratory test data. Methods A total of 224 confirmed MM patients diagnosed at Zhongshan City People’s Hospital between 2022 and 2024 were enrolled as the study group, and 200 healthy individuals undergoing routine physical examinations during the same period were selected as the control group. Seventy-four routine laboratory indicators were collected from both groups. Data normality was assessed by the Kolmogorov-Smirnov (K-S) test, followed by independent samples t-test or rank-sum test to compare differences between groups. Receiver operating characteristic (ROC) curves were plotted for statistically significant indicators. Variables with area under the curve (AUC)>0.7 were selected as effective features for model construction. A random forest classifier was used to con-struct the predictive model, and its performance was evaluated by 5-fold cross-validation. Results Thirty-eight laboratory indica-tors showed statistically significant differences between the two groups (all P<0.05). Among these, total urinary κ and λ light chains exhibited the highest predictive performance with AUC values of 1.0. Elevated 5-nucleotidase (5-NT), rheumatoid factor (RF), red cell distribution width-coefficient of variation (RDW-CV), and thrombin time (TT), as well as decreased hematocrit (HCT) and hemoglobin (Hb) levels, showed strong predictive power with AUCs>0.9. Variables such as β2-microglobulin(β2-MG), red cell distribution width-standard deviation (RDW-SD), Cystatin C (CysC), C-reactive protein (CRP), adenosine de-aminase (ADA), triglycerides (TG), creatinine (Cr), alkaline phosphatase (ALP), urea, anti-streptolysin O (ASO) elevation, and while albumin (Alb), red blood cells (RBC), estimated glomerular filtration rate (eGFR), platelets (PLT), activated partial throm-boplastin time (APTT), creatine kinase isoenzyme (CK-MB), mean corpuscular hemoglobin concentration (MCHC), high-density lipoprotein cholesterol (HDL-C), and total protein (TP) were decreased with AUCs ranging from 0.7 to 0.9, indicating moderate predictive capability. The remaining statistically significant indicators had AUCs between 0.5 and 0.7, indicating weaker predic-tive ability. Indicators with AUC values around 0.7 were prioritized based on importance. The constructed random forest model achieved an accuracy of 100%, with precision, recall, and F1-score all at 1.00 in distinguishing MM patients from healthy con-trols. The model was saved as a .pkl file, and the complete Python code has been uploaded to GitHub. Conclusions Using a ran-dom forest classifier, a predictive model for MM was effectively established based on routine laboratory indicators, significantly improving early identification of MM and enhancing the clinical utility of routine laboratory data.

参考文献/References:

[1] JIA Y , ZHOU J , TAN T K, et al. Enhancer-mediated upregulation of HJURP promotes growth and survival of t(4;14)-positive Multiple Myeloma [J]. Cancer Research, 2022, 82(3): 406-418.
[2] KUMAR S K, DISPENZIERI A, LACY M Q, et al. Continued improvement in survival in Multiple Myeloma: changes in early mortality and outcomes in older patients[J]. Leukemia, 2014, 28(5): 1122-1128.
[3] 刘志伟,李晓婷,逢婷,等.常规检验项目在多发性骨髓瘤诊断中的临床价值[J].临床血液学杂志,2021,34(1):48-52. LIU Z W, LI X T, FENG T, et al. The clinical value of routine tests in the diagnosis of Multiple Myeloma[J]. Journal of Clinical Hematology, 2021, 34(1): 48-52.
[4] 中国医师协会血液科医师分会,中华医学会血液学分会.中国多发性骨髓瘤诊治指南(2024年修订)[J].中华内科杂志,2024,63(12):1186-1195. Chinese Hematology Association, Chinese Society of Hematology. Guidelines for the diagnosis and management of Multiple Myeloma in China (2024 revision)[J]. Chinese Journal of Internal Medicine, 2024, 63(12): 1186-1195.
[5] CAI J Y, LIU Z H, WANG Y Y, et al. Construction of the prediction model for Multiple Myeloma based on machine learning[J]. International Journal of Laboratory Hematology, 2024, 46(5): 918-926.
[6] WALLINGTON-BEDDOE C T, MYNOTT R L. Prognostic and predictive biomarker developments in Multiple Myeloma [J]. Journal of Hematology &Oncology, 2021, 14(1): 151.
[7] 霍豆,秦爽,吴永昌,等.血清总轻链与游离轻链定量检测在多发性骨髓瘤诊断中的临床价值探讨[J].现代检验医学杂志, 2020, 35(4): 87-88, 157. HUO D, QIN S, WU Y C, et al. Clinical value of quantitative detection of sTLC and sFLC in diagnosis of Multiple Myeloma[J]. Journal of Modern Laboratory Medicine, 2020, 35(4): 87-88, 157.
[8] 郭进京,胡林辉,陶千山,等.红细胞分布宽度在多发性骨髓瘤患者预后分期中的价值[J].现代检验医学杂志, 2017, 32(3): 34-36, 39. GUO J J, HU L H, TAO Q S, et al. Value of red cell distribution width in the prognosis of patients with Multiple Myeloma[J]. Journal of Modern Laboratory Medicine, 2017, 32(3): 34-36, 39.
[9] 李东英,岳园芳,陈琳,等.红细胞分布宽度标准差对143例初治多发性骨髓瘤患者预后预测价值[J].中华肿瘤防治杂志, 2020, 27(6): 475-479, 498. LI D Y, YUE Y F, CHEN L, et al. Standard deviation of red blood cell distribution width in predicting the prognosis of 143 patients with newly treated Multiple Myeloma[J]. Chinese Journal of Cancer Prevention and Treatment, 2020, 27(6): 475-479, 498.
[10] BAN C M, YANG F Y, WEI M, et al. Integrative analysis of gene expression through one-class logistic regression machine learning identifies stemness features in Multiple Myeloma [J]. Frontiers in Genetics, 2021, 12: 666561.
[11] 杨文丽,李晓明.多发性骨髓瘤患者血脂的检测及临床意义[J].现代肿瘤医学,2021,29(19):3458-3461. YANG W L, LI X M. Detection of blood lipids in patients with Multiple Myeloma and its clinical significance[J]. Journal of Modern Oncology, 2021, 29(19): 3458-3461.
[12] CARC? D, MARKOVIC U, CASTORINA P, et al. C-reactive protein monitoring and clinical presentation of fever as predictive factors of prolonged febrile neutropenia and blood culture positivity after autologous hematopoietic stem cell transplantation-single-center real-life experience [J]. Journal of Clinical Medicine, 2022, 11(2): 312.

相似文献/References:

[1]韩秀蕊,杨娣娣,李艳春,等.多发性骨髓瘤患者骨髓涂片与骨髓活检同步检查的比较分析[J].现代检验医学杂志,2015,30(03):129.[doi:10.3969/j.issn.1671-7414.2015.03.039]
　HAN Xiu-rui,YANG Di-di,LI Yan-chun,et al.Comparative Analysis of Bone Marrow Smears and Biopsies Synchronous Check for Myeloma Patients[J].Journal of Modern Laboratory Medicine,2015,30(02):129.[doi:10.3969/j.issn.1671-7414.2015.03.039]
[2]邱 爽,孟瑞芳,蒋筱漪,等.血清免疫固定电泳、蛋白电泳、免疫球蛋白及轻链定量在诊断多发性骨髓瘤中的临床应用[J].现代检验医学杂志,2015,30(02):61.[doi:10.3969/j.issn.1671-7414.2015.02.019]
　QIU Shuang,MENG Rui-fang,JIANG Xiao-yi,et al.Clinical Application of Immunofixtion Electrophoresis, Serum Protein Electrophoresis and Immunoglobulins and Light Chain Quantitative Analysis in the Diagnosis of Multiple Myeloma[J].Journal of Modern Laboratory Medicine,2015,30(02):61.[doi:10.3969/j.issn.1671-7414.2015.02.019]
[3]李 瑛,李 军.多发性骨髓瘤患者血清中IL-6与IL-27水平监测的临床应用[J].现代检验医学杂志,2016,31(04):87.[doi:10.3969/j.issn.16717-414.2016.04.023]
　LI Ying,LI Jun.Clinical Application of Monitoring IL-6 and IL-27 Levels in Patients with Multiple Myeloma[J].Journal of Modern Laboratory Medicine,2016,31(02):87.[doi:10.3969/j.issn.16717-414.2016.04.023]
[4]郭进京,胡林辉,等.红细胞分布宽度在多发性骨髓瘤患者预后分期中的价值[J].现代检验医学杂志,2017,32(03):34.[doi:10.3969/j.issn.1671-7414.2017.03.009]
　GUO Jin-jing,HU Lin-hui,TAO Qian-shan,et al.Value of Red Cell Distribution Width in the Prognosis of Patients with Multiple Myeloma[J].Journal of Modern Laboratory Medicine,2017,32(02):34.[doi:10.3969/j.issn.1671-7414.2017.03.009]
[5]谭 奎,沈婵娟,张 玲,等.多发性骨髓瘤患者骨髓CD269和CD317基因的差异性表达研究[J].现代检验医学杂志,2017,32(06):64.[doi:10.3969/j.issn.1671-7414.2017.06.001]
　TAN Kui,SHEN Chan-juan,ZHANG Ling,et al.Differential Expression of CD269 and CD317 Genes in Bone Marrow of Patients with Multiple Myeloma[J].Journal of Modern Laboratory Medicine,2017,32(02):64.[doi:10.3969/j.issn.1671-7414.2017.06.001]
[6]盘国雄,谭才燕,何嘉颖,等.多发性骨髓瘤患者血清中lncRNA PCAT-1的表达水平与临床预后研究[J].现代检验医学杂志,2018,33(01):72.[doi:10.3969/j.issn.1671-7414.2018.01.001]
　PAN Guo-xiong,TAN Cai-yan,HE Jia-ying,et al.Serum LncRNA PCAT-1 Expression Level of Patients with Multiple Myeloma and Clinical Value[J].Journal of Modern Laboratory Medicine,2018,33(02):72.[doi:10.3969/j.issn.1671-7414.2018.01.001]
[7]刘玉霞,胡国瑜,袁朝晖,等.CD269和CD317在多发性骨髓瘤中的表达及临床意义[J].现代检验医学杂志,2018,33(02):58.[doi:10.3969/j.issn.1671-7414.2018.02.001]
　LIU Yu-xia,HU Guo-yu,YUAN Chao-hui,et al.Expression of CD269 and CD317 in Multiple Myeloma and Its Clinical Significance[J].Journal of Modern Laboratory Medicine,2018,33(02):58.[doi:10.3969/j.issn.1671-7414.2018.02.001]
[8]张 蓉,李国辉,刘小五,等.初诊多发性骨髓瘤患者外周血淋巴细胞绝对值/ 单核细胞绝对值比值在预测临床预后的价值研究[J].现代检验医学杂志,2020,35(01):101.[doi:10.3969/j.issn.1671-7414.2020.01.027]
　ZHANG Rong,LI Guo-hui,LIU Xiao-wu,et al.Value of Peripheral Blood Lymphocyte Absolute Value/Monocyte Absolute Value Ratio in Predicting Clinical Prognosis of Newly Diagnosed MultipleMyeloma Patients[J].Journal of Modern Laboratory Medicine,2020,35(02):101.[doi:10.3969/j.issn.1671-7414.2020.01.027]
[9]霍 豆,秦 爽,吴永昌,等.血清总轻链与游离轻链定量检测在多发性骨髓瘤诊断中的临床价值探讨[J].现代检验医学杂志,2020,35(04):87.[doi:10.3969/j.issn.1671-7414.2020.04.021]
　HUO Dou,QIN Shuang,WU Yong-chang,et al.Clinical Value of Quantitative Detection of sTLC and sFLC in Diagnosis of Multiple Myeloma[J].Journal of Modern Laboratory Medicine,2020,35(02):87.[doi:10.3969/j.issn.1671-7414.2020.04.021]
[10]何　进,张　艳,申娴娟,等.多发性骨髓瘤患者血清可溶性PD-L1水平在辅助诊断及临床分型的价值研究[J].现代检验医学杂志,2021,36(02):15.[doi:doi:10.3969/j.issn.1671-7414.2021.02.004]
　HE Jin,ZHANG Yan,SHEN Xian-juan,et al.Value of Blood Soluble PD-L1 in the Auxiliary Diagnosis and Clinical Subtype of Multiple Myeloma[J].Journal of Modern Laboratory Medicine,2021,36(02):15.[doi:doi:10.3969/j.issn.1671-7414.2021.02.004]

备注/Memo

备注/Memo:
基金项目： 广东省中山市科技局重大项目（项目编号：K2021B3014），广东省中山市卫健局一般项目（项目编号：2021A020523），中山市社会公益科技研究项目（2022B1080）。
作者简介：李莎（1989-），女，学士，副主任技师，研究方向：临床生物化学检验，E-mail：1024234271@qq.com。
通讯作者：李丽敏（1994-），女，硕士，主管技师，研究方向：临床生物化学检验，E-mail：841909331@qq.com。
更新日期/Last Update: 2026-03-15