基于蛋白质组学的慢阻肺风险预测与药物靶点识别
32蛋白评分在测试集中C指数为0.826(95%CI 0.803-0.849),优于多基因风险评分,与临床模型相当;简化10蛋白面板保持稳健性能。孟德尔随机化识别出ADM、SCGB1A1为保护性蛋白,MMP12、TNFRSF10A增加风险。
要点
- 32蛋白风险评分对慢阻肺发病预测准确性高(C指数0.826),优于多基因风险评分。
- 简化10蛋白面板性能稳健(C指数0.816),更具临床可行性。
- 蛋白评分联合临床因素可改善风险重分类(NRI 0.251-0.318)。
核心结论
- 蛋白质生物标志物在慢阻肺预测中优于遗传风险评分,并能补充临床因素。
- 简化10蛋白面板具有临床可行性。
- 研究识别了新的生物学通路和因果治疗靶点,常规临床应用前需进一步验证。
研究设计与样本
- 采用最小绝对收缩和选择算子回归在训练集中识别预测蛋白。
- 模型性能采用Harrell C指数、净重分类改善(NRI)和综合判别改善指数(IDI)评估。
- 通过通路分析和孟德尔随机化分析探索生物学机制和因果效应。
主要发现
- 测试集中32蛋白风险评分强烈预测慢阻肺发病(C指数0.826,95%CI 0.803-0.849)。
- 32蛋白评分优于多基因风险评分(C指数0.510,95%CI 0.478-0.542),与临床模型相当(C指数0.845,95%CI 0.823-0.867)。
- 简化10蛋白面板保持稳健性能(C指数0.816,95%CI 0.792-0.840)。
- 蛋白评分加入临床因素改善风险重分类(NRI 0.251-0.318;IDI 0.042-0.064)。
- 孟德尔随机化分析识别ADM和SCGB1A1为保护性蛋白,MMP12和TNFRSF10A增加风险。
关键数值与推荐
- 32蛋白风险评分预测慢阻肺的C指数为0.826(95%CI 0.803-0.849)。
- 10蛋白面板C指数为0.816(95%CI 0.792-0.840)。
- 多基因风险评分C指数为0.510(95%CI 0.478-0.542)。
- 临床模型C指数为0.845(95%CI 0.823-0.867)。
- 蛋白评分加入临床因素的NRI为0.251-0.318,IDI为0.042-0.064。
临床怎么用
- 蛋白质风险评分可用于慢阻肺风险预测,优于遗传风险评分并补充临床因素。
- 简化10蛋白面板更具临床落地可行性。
- ADM、SCGB1A1、MMP12、TNFRSF10A等为潜在因果治疗靶点,但需进一步验证后用于常规临床。
局限性
- 常规临床实施前需进一步验证。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: Chronic obstructive pulmonary disease (COPD) is a leading cause of global mortality, yet existing risk prediction models remain limited. This study aimed to develop and validate a protein-based risk score for COPD, comparing its performance against COPD polygenic risk scores (PRSs) and clinical risk factors, while exploring underlying biological pathways and causal protein-disease associations. METHODS: The study analysed 27 796 UK Biobank participants from England (70% training and 30% testing set) and 3534 from Scotland/Wales (validation cohort). Least absolute shrinkage and selection operator regression identified predictive proteins in the training set, with model performance assessed using Harrell's C-index, Net Reclassification Improvement (NRI) and Integrated Discrimination Improvement Index (IDI). Pathway and Mendelian randomisation (MR) analyses explored biological mechanisms and causal effects. RESULTS: In the testing set, a developed 32-protein risk score strongly predicted incident COPD with high accuracy (C-index 0.826, 95% CI 0.803 to 0.849). It outperformed PRS (C-index 0.510, 95% CI 0.478 to 0.542) and matched clinical models (C-index 0.845, 95% CI 0.823 to 0.867). A simplified 10-protein panel retained robust performance (C-index 0.816, 95% CI 0.792 to 0.840). Adding the protein scores to clinical factors improved risk reclassification (NRI 0.251-0.318; IDI: 0.042-0.064). MR analysis identified ADM and SCGB1A1 as protective, while MMP12 and TNFRSF10A increased risk. Pathway analysis implicated inflammation and extracellular remodelling. Chitinase-3-like protein 1 and matrix metalloproteinase-9 were central players in the protein-protein interaction network. Similar results were found in the validation cohort. CONCLUSION: Protein biomarkers outperform genetic risk scores and complement clinical factors for COPD prediction, with a streamlined 10-protein panel offering clinical feasibility. The study identifies novel pathways and causal therapeutic targets. Further validation is needed prior to routine clinical implementation.
引用
- Zhang Y, Ye Z, Yang S, et al. Proteomics-based risk prediction and drug targets identification for chronic obstructive pulmonary disease. Thorax. 2026. PMID: 40992936. doi:10.1136/thorax-2024-222397
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。