Blood metabolites–based MCTarg for multi-cancer screening and clinical diagnosis.

Abstract

Listen

Translate article

e15028 Background: Multi-cancer early detection (MCED) tests have emerged as promising tools for reducing cancer-related healthcare costs and mortality. As metabolome is closely linked to the phenotype of bio-individual, metabolomics analysis enables the high throughput investigating of the pathophysiological conditions. Based on metabolomics and machine learning, we developed the Multiple Cancer Target (MCTarg) analysis for multiple cancer screening and diagnosis by a single blood draw. Methods: A total of 1153 individuals (172 healthy participants, 623 patients with lung cancer, gastric cancer, and colorectal cancer, and 358 these organ-related benign disease patients) were enrolled from three independent clinical sites in China. Plasma samples were collected and measured by mass spectrometry (MS)-based platforms (LC-MS polar and lipid). MCTarg was developed with three machine learning modules for three scenarios (community screening for non-healthy individuals, clinical diagnosis for classifying malignant and benign diseases, and tracing tumor of origin). Results: MCTarg exhibited an Area Under the Curve (AUC) of 0.96 with 92.82% sensitivity at 80% specificity to differentiate the non-healthy individuals from healthy controls, and an AUC of 0.86 with 75.41% sensitivity at 79.66% specificity in discriminating multi-cancer from benign diseases. Notably, the classification accuracy for distinguishing between malignant and benign conditions in the colorectal organ was reached 90.74%, with a sensitivity of 83.87% and a specificity of 100%. Furthermore, the overall accuracy for detecting the cancer origin reached 84.78% with an average sensitivity and specificity of 78.08% and 90.7%, respectively. Conclusions: The outstanding results indicated that blood-metabolites-based MCTarg is poised to be a cost-effective, precise, and versatile tool for diagnosing multi-cancer and discriminating tumor of origin. Its potential applications encompass health checkups and supplementary diagnostic scenarios. More deadly cancer types will be extended and the performance of MCTarg will be validated by prospective and population-scale cohorts with longitudinal follow-up.