上海中医药杂志2026，Vol.60Issue(3)：1-10,10.DOI:10.16305/j.1007-1334.2026.z20250825002
基于机器学习的初诊食管鳞状细胞癌患者脾虚/湿热的证候变化及代谢特征研究
Machine learning‑based study on syndrome transformation and metabolic characteristics of spleen deficiency/dampness‑heat in patients with newly diagnosed esophageal squamous cell carcinoma
摘要
Abstract
Objective To explore the syndrome transformation and metabolic characteristics of spleen deficiency/dampness-heat syndromes of traditional Chinese medicine(TCM)in patients with newly diagnosed esophageal squamous cell carcinoma(ESCC)based on machine learning algorithms.Methods Two independent ESCC cohorts were enrolled,namely the discovery cohort and validation cohort,and healthy volunteers were recruited as controls.The Spleen Deficiency/Dampness-Heat Syndrome Patient-Reported Outcome Scale(hereinafter referred to as the PRO Scale)was used to collect and score syndrome information of newly diagnosed ESCC patients.Machine learning K‐means clustering analysis was applied to determine TCM syndromes,and patients were classified into the spleen deficiency syndrome(SDS)group,dampness-heat syndrome(DHS)group,combined syndrome(CS)group(with both spleen deficiency and dampness-heat syndromes),and non-spleen deficiency/dampness-heat syndrome(NSD/DHS)group.Postoperative PRO Scale data were collected from the discovery cohort and compared with preoperative data to analyze the syndrome transformation and dynamic changes in TCM syndrome manifestations.Fasting peripheral venous blood samples were collected from subjects in the discovery cohort and healthy controls for metabolomic detection and analysis.Additionally,an ESCC cohort with survival information and the corresponding serum metabolomic data from the preliminary study of our research group were included for correlation analysis.Results ①A total of 252 newly diagnosed ESCC patients were enrolled,including 147 cases in the discovery cohort and 105 cases in the validation cohort,with 75 healthy controls recruited separately.Complete preoperative PRO Scale data were obtained from all 252 patients.For the discovery cohort,serum samples were collected from 75 patients before treatment,and PRO Scale data on the 10th day after radical surgery were followed up and acquired from 73 of these patients.②K-means clustering analysis effectively identified TCM syndromes in newly diagnosed ESCC patients.The proportions of spleen deficiency/dampness-heat related syndromes were 40.8%in the discovery cohort and 41.9%in the validation cohort.③Surgery disturbed spleen deficiency and dampness-heat syndromes in patients and exacerbated the manifestation of yellow,thick and greasy tongue coating.④Metabolomic analysis showed that the metabolic profiles of the DHS group and CS group were significantly deviated from those of the NSD/DHS group.Syndrome-specific metabolites were identified:indoxyl sulfate and pseudouridine for spleen deficiency syndrome;obacunone and threonine for dampness-heat syndrome;ranaconitine and salicylic acid for combined syndrome.⑤Survival analysis indicated that high expressions of indoxyl sulfate,pseudouridine and ranaconitine predicted poor overall survival,whereas high expression of obacunone indicated favorable prognosis.Conclusions Machine learning-assisted analysis of PRO Scale data can accurately identify TCM syndromes and their dynamic changes in newly diagnosed ESCC patients.Each syndrome corresponds to specific metabolic profiles and characteristic metabolites,and some of these metabolites have independent predictive value for prognosis.关键词
食管癌/机器学习/脾虚证/湿热证/代谢组学/中医证候/外源代谢Key words
esophageal cancer/machine learning/spleen deficiency syndrome/dampness-heat syndrome/metabolomics/traditional Chinese medicine syndrome/exogenous metabolite引用本文复制引用
高玲,秦辰泰,王思亮,郝苗秀,张铭,石玉琳,冯利,刘丹,杨晞,陈文连,刘红,季光,徐汉辰,刘雷,张杰,金星,董昌盛,郑苗苗..基于机器学习的初诊食管鳞状细胞癌患者脾虚/湿热的证候变化及代谢特征研究[J].上海中医药杂志,2026,60(3):1-10,10.基金项目
国家重点研发计划"中医药现代化"重点专项(2023YFC3503200,2023YFC3503201,2022YFC3500200,2022YFC3500201) （2023YFC3503200,2023YFC3503201,2022YFC3500200,2022YFC3500201）
国家中医药管理局中医药创新团队及人才支持计划项目(ZYYCXTD-C-202208) （ZYYCXTD-C-202208）
国家自然科学基金项目(82304780) （82304780）
上海市自然科学基金项目(25ZR1402486) （25ZR1402486）
