Non-Small Cell Lung Cancer Detection and Subtyping by UPLC-HRMS-Based Tissue Metabolomics
Xiaoling Zang,Jie Zhang,Peng Jiao,Xuyan Xue,Zhihua Lv
DOI: https://doi.org/10.1021/acs.jproteome.2c00316
2022-01-01
Journal of Proteome Research
Abstract:Non-small cell lung cancer (NSCLC) is the prevalent histological subtype of lung cancer. In this study, we performed ultraperformance liquid chromatography-high resolution mass spectrometry (UPLC-HRMS)-based metabolic profiling of 227 tissue samples from 79 lung cancer patients with adenocarcinoma (AC) or squamous cell carcinoma (SCC). Orthogonal partial least squares-discriminant analysis (oPLS-DA) analyses showed that AC, SCC, and NSCLC tumors were discriminated from adjacent noncancerous tissue (ANT) and distant noncancerous tissue (DNT) samples with good accuracies (91.3, 100, and 88.3%), sensitivities (85.7, 100, and 83.9%), and specificities (94.3, 100, and 90.7%), using 12, 4, and 7 discriminant metabolites, respectively. The discriminant panel for AC detection included valine, sphingosine, glutamic acid gamma-methyl ester, and lysophosphatidylcholine (LPC) (16:0), levels of which in tumor tissues were significantly altered. Valine, sphingosine, LPC (18:1), and leucine derivatives were used for SCC detection. The discrimination between AC and SCC had 96.8% accuracy, 98.2% sensitivity, and 85.7% specificity using a five-metabolite panel, of which valine and creatine had significant differences. The classification models were further verified with external validation sets, showing a promising prospect for NSCLC tissue detection and subtyping.