Breast cancer detection using targeted plasma metabolomics

• Published in: Journal of chromatography. B, Analytical technologies in the biomedical and life sciences Analytical technologies in the biomedical and life sciences, 2019-01, Vol.1105, p.26-37
• Main Authors: Jasbi, Paniz, Wang, Dongfang, Cheng, Sunny Lihua, Fei, Qiang, Cui, Julia Yue, Liu, Li, Wei, Yiping, Raftery, Daniel, Gu, Haiwei
• Format: Article
• Language: English
• Subjects: Biomarker discovery; Biomarkers, Tumor - blood; Breast cancer; Breast Neoplasms - blood; Breast Neoplasms - diagnosis; Chromatography, Liquid - methods; Female; Humans; LC-MS/MS; Least-Squares Analysis; Linear Models; Metabolome - physiology; Metabolomics; Metabolomics - methods; Reproducibility of Results; ROC Curve; Sensitivity and Specificity; Tandem Mass Spectrometry - methods; Targeted detection
• ISSN: 1570-0232; 1873-376X
• DOI: 10.1016/j.jchromb.2018.11.029

Breast cancer (BC) is a major cause of human morbidity and mortality, especially among women. Despite the important role of metabolism in the molecular pathogenesis of cancer, robust metabolic markers to enable enhanced screening and disease monitoring of BC are still critically needed. In this study, a targeted liquid chromatography-tandem mass spectrometry (LC-MS/MS) metabolic profiling approach is presented for the identification of metabolic marker candidates that could enable highly sensitive and specific detection of all-stage as well as early-stage BC. In this targeted approach, 105 metabolites from >35 metabolic pathways of potential biological relevance were reliably detected in 201 plasma samples taken from two groups of subjects (102 BC patients and 99 healthy controls). The results of our general linear model and partial least squares-discriminant analysis (PLS-DA) informed the construction of a novel 6-metabolite panel of potential biomarkers. A receiver operating characteristic (ROC) curve generated based on an improved PLS-DA model showed relatively high sensitivity (0.80), specificity (0.75), and area under the receiver-operating characteristic curve (AUROC = 0.89). Similar classification performance of the model was observed for detection of early-stage BC (AUROC = 0.87, sensitivity: 0.86, specificity: 0.75). Bioinformatics analyses revealed significant disturbances in arginine/proline metabolism, tryptophan metabolism, and fatty acid biosynthesis. Our univariate and multivariate results indicate the effectiveness of this metabolomics approach for all-stage as well as early-stage BC diagnosis; our bioinformatics results indicate affected pathways related to tumor growth, metastasis, and immune escape mechanisms. Future studies should validate these results using more samples from different locations.