Although ultra-high-resolution mass spectrometry (UHRMS) coupled with high-performance liquid chromatography (LC-UHRMS) can identify the isomeric diversity of molecular formulas in organic matter, its application is limited by high operational costs, the requirement for expert handling, and complex data interpretation. Here, the authors report a graph neural network model for inferring isomeric diversity from direct infusion UHRMS (DI-UHRMS) data and demonstrate its high accuracy in predicting the chromatographic isomeric diversity of natural dissolved organic matter, with comparable results achieved across various DI-UHRMS techniques.
- Tongcun Liu
- Yuanbi Yi
- Ding He