文献速递传染病防控与疫情报告免疫规划

利用机器学习方法分析单细胞多组学数据识别COVID-19患者康复后免疫系统长期效应相关的基因与蛋白特征

Identification of gene and protein signatures associated with long-term effects of COVID-19 on the immune system after patient recovery by analyzing single-cell multi-omics data using a machine learning approach.
PubMed 文献 · Vaccine · 2024-08-24

该研究采用LASSO、MCFS、RF、CATBoost和XGBoost五种特征排序算法,结合增量特征选择、合成少数类过采样技术及决策树与k近邻两种分类算法,分析COVID-19康复患者与健康对照者在流感疫苗接种后的单细胞多组学数据(表面蛋白与转录组)。研究发现COVID-19康复患者具有独特的免疫状态,特征为核糖体蛋白RPS26低表达和免疫细胞表面蛋白CD33、CD48高表达,且TMEM176B在康复患者单核细胞中高表达。

要点

  • COVID-19康复患者呈现独特免疫状态:RPS26低表达,CD33与CD48高表达。
  • TMEM176B在COVID-19康复患者单核细胞中高表达。
  • 研究构建了有效分类器并识别了多组学标志物,有助于理解COVID-19对免疫系统的长期影响。

核心结论

  • COVID-19康复患者存在独特的免疫状态,以核糖体蛋白RPS26低表达和免疫细胞表面蛋白CD33、CD48高表达为特征。
  • TMEM176B在COVID-19康复患者单核细胞中高表达。
  • 这些发现有助于辨别免疫细胞亚型间的分子差异,并促进理解COVID-19对免疫系统的长期影响。

研究设计与样本

  • 研究分析COVID-19康复患者与健康对照者在流感疫苗接种后的单细胞多组学数据(表面蛋白与转录组)。
  • 数据集分为B细胞、CD4+ T细胞、CD8+ T细胞、单核细胞和自然杀伤细胞五个免疫细胞亚型子集。
  • 每个细胞由28,402个scRNA-seq(RNA)特征、3个Hash Tag Oligo(HTO)特征、138个CITE特征和23,569个Single Cell Transform(SCT)特征表示。
  • 采用LASSO、MCFS、RF、CATBoost和XGBoost五种特征排序算法,结合增量特征选择、合成少数类过采样技术及决策树与k近邻两种分类算法进行分析。

主要发现

  • COVID-19康复患者呈现核糖体蛋白RPS26低表达和免疫细胞表面蛋白CD33、CD48高表达。
  • TMEM176B在COVID-19康复患者单核细胞中高表达。
  • 识别了一些多组学标志物并构建了有效分类器。

局限性

  • 原文未报告局限性。

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

Viral infections significantly impact the immune system, and impact will persist until recovery. However, the influence of severe acute respiratory syndrome coronavirus 2 infection on the homeostatic immune status and secondary immune response in recovered patients remains unclear. To investigate these persistent alterations, we employed five feature-ranking algorithms (LASSO, MCFS, RF, CATBoost, and XGBoost), incremental feature selection, synthetic minority oversampling technique and two classification algorithms (decision tree and k-nearest neighbors) to analyze multi-omics data (surface proteins and transcriptome) from coronavirus disease 2019 (COVID-19) recovered patients and healthy controls post-influenza vaccination. The single-cell multi-omics dataset was divided into five subsets corresponding to five immune cell subtypes: B cells, CD4+ T cells, CD8+ T cells, Monocytes, and Natural Killer cells. Each cell was represented by 28,402 scRNA-seq (RNA) features, 3 Hash Tag Oligo (HTO) features, 138 Cellular indexing of transcriptomes and epitopes by sequencing (CITE) features and 23,569 Single Cell Transform (SCT) features. Some multi-omics markers were identified and effective classifiers were constructed. Our findings indicate a distinct immune status in COVID-19 recovered patients, characterized by low expression of ribosomal protein (RPS26) and high expression of immune cell surface proteins (CD33, CD48). Notably, TMEM176B, a membrane protein, was highly expressed in monocytes of COVID-19 convalescent patients. These observations aid in discerning molecular differences among immune cell subtypes and contribute to understanding the prolonged effects of COVID-19 on the immune system, which is valuable for treating infectious diseases like COVID-19.

来源:Vaccine,查看原文。版权归原作者所有。

引用

  1. Ren J, Gao Q, Zhou X, et al. Identification of gene and protein signatures associated with long-term effects of COVID-19 on the immune system after patient recovery by analyzing single-cell multi-omics data using a machine learning approach. Vaccine. 2024. PMID: 39182316. doi:10.1016/j.vaccine.2024.126253

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。