Wooden-Tip Electrospray Ionization Mass Spectrometry Combined With Machine Learning for Differentiating Thyroid Tumours

Abstract

Direct mass spectrometry (MS) analysis of human tissues at the molecular level has great potential for clinical diagnosis and biomarker discovery. However, conventional MS-based analytical methods often require complicated and time-consuming sample preparation, which limits their applicability in rapid clinical analysis. In this study, we developed a rapid analytical strategy by integrating ambient ionization MS with machine learning (ML) for the differentiation of different thyroid tumours. A disposable slim wooden tip (WT) was employed as both a sample holder and an electrospray emitter, enabling direct extraction and ionization of metabolites from tiny thyroid tissue samples under electrospray ionization (ESI) conditions. Using this WT–ESI–MS method, lipid profiles of thyroid tissues could be obtained within minutes without extensive sample preparation. A total of 45 thyroid samples, including 15 healthy tissues, 15 benign tumours and 15 malignant tumours, were analysed. The acquired MS data were further processed using ML-based classification models to distinguish different tumours and identify potential lipid biomarkers. Structural characterization of representative lipids was also performed by MS/MS analysis. The results demonstrated that this WT–ESI–MS combined with ML provides a rapid and effective approach not only for differentiating tumour tissues and healthy samples but also for benign and malignant tumours, highlighting its potential application in clinical diagnosis and intraoperative tissue evaluation.