但是如果取样的时候,对肿瘤病人和正常人分别取他们的外周血,这个时候就很难看到全局的表达量差异了。这个时候无论是表达量检测手段是什么,表达量芯片,转录组测序,蛋白质组学或者代谢组学,哪怕是单细胞转录组,都有人做过,都很难发现什么全局的差异。但是,大家仍然是会强行找差异然后各种机器学习勉强得到一些结果争取发表出来。比如2025年1月的文献:《LcProt: Proteomics-based identification of plasma biomarkers for lung cancer multievent, a multicentre study》,有多个队列:
有意思的是,研究者们其实并不是仅仅是取样了blood,在文献里面写了是:Peripheral blood samples before and after treatment, tumour tissue, pair para-cancerous tissue and biopsy tissue samples were planned to collect based on the treatment protocols.
文献PMID: 19951989:取样是 peripheral blood mononuclear cell (PBMC) ,来源于 137 patients with NSCLC tumors and 91 patient controls with nonmalignant lung conditions
数据集GSE20189 ,取样是 lung tissue and peripheral whole blood (PWB) from adenocarcinoma cases and controls,是153 subjects (73 adenocarcinoma cases, 80 controls)
tcga的luad和lusc数据集,We used 226 samples in which cancer metastases were annotated and divided them into two categories, namely, lung cancer metastasis and non-metastasis. 来源于文献:https://doi.org/10.1016/j.compbiomed.2022.106490