Serum lipidomic profiling for liver cancer screening using surface-assisted laser desorption ionization MS and machine learning

Abstract

The liver is a major organ in metabolism, and alterations in serum lipids are associated with liver disorders. Here, a rapid, easy, and reliable screening technique based on lipidomic profiling was developed using machine learning and surface-assisted laser desorption ionization mass spectrometry (SALDI MS) for liver cancer diagnosis. A graphitized carbon matrix (GCM) was created for serum lipid profiling in SALDI MS and demonstrated a better performance for neutral lipids analysis than conventional organic matrices. The fingerprint of serum lipids, including triacylglycerols (TGs), diacylglycerols (DGs), cholesteryl esters (CEs), glycerophospholipids (GPs), and other components, could be directly obtained by GCM-assisted LDI MS without extraction. Five machine learning methods were applied to distinguish liver cancer (LC) patients from healthy controls (HC) and chronic hepatitis B (CHB) patients. The best diagnostic performance was attained by linear discriminant analysis (LDA), which has a confusion matrix accuracy of 98.3 %. The receiver operating characteristic (ROC) curve for liver cancer exhibited an area under the curve (AUC) of 0.99, indicating a high degree of prediction accuracy. One-way ANOVA analysis revealed that numerous TGs were down-regulated in LC group. The results demonstrated the viability of GCM-assisted LDI MS as a valuable diagnostic tool for liver cancer.